feat(cmd): serve mounts audio upload when enabled (ADR-033)

Wires buildUploads into cmdServe: Intake and limits into the web App, the
worker on the server's ctx. Logs the two facts an operator must not miss: the
queue is process-local (single replica), and a configured fallback is a cloud
endpoint that private audio can leave the homelab through.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

.gitea/build-image.sh

@@ -0,0 +1,121 @@
+#!/usr/bin/env bash
+# Build an OCI image with buildctl and push it to the local registry.
+#
+# CANONICAL COPY: mathias/infra ci-templates/build-image.sh (infra#309).
+# Every repo carries a byte-identical copy at .gitea/build-image.sh, written by
+# `task ci:sync` and held in place by the drift gate. DO NOT EDIT THE COPIES --
+# edit here, re-sync, and the gate will tell you which repos still disagree.
+#
+# Why this exists: this step was hand-copied into a dozen repos, so migrating it
+# off rootless buildah (infra#241) meant eleven near-identical commits, eleven
+# verifications, and three separate rediscoveries of the same fix in the two
+# repo templates and the gitea-ci skill. The differences between the copies were
+# incidental -- a label here, a build-arg there -- never intentional. Those are
+# inputs now, not forks of the implementation.
+#
+# Why buildctl and not buildah: the gitea-runner pod runs with
+# `capabilities: drop [ALL]` and `allowPrivilegeEscalation: false`, which
+# unconditionally blocks the setuid-root exec that rootless buildah's uid/gid
+# remapping needs, and the pod has no OCI runtime at all. buildctl is a thin
+# client -- it hands the build to the host's buildkitd over a socket, so the
+# privileged layer work happens on the host and this pod's securityContext is
+# not in the picture. buildah is absent from the runner image as of baked-v14
+# and its absence is asserted at pod start (infra#308).
+#
+# Inputs, all via environment:
+#
+#   IMAGE            (required) image name, e.g. cobalt-dingo
+#   REGISTRY         default localhost:5000
+#   SHA_TAG          default `git rev-parse --short HEAD`
+#   VERSION_TAG      optional extra tag, e.g. v1.2.3 on a tag build
+#   PUSH_LATEST      default 1; set 0 to skip the :latest tag
+#   CONTEXT          default . -- build context directory
+#   DOCKERFILE_DIR   default $CONTEXT -- directory holding the Dockerfile
+#   DOCKERFILE_NAME  optional, when the file is not named `Dockerfile`
+#   TARGET           optional, multi-stage build target (--opt target=)
+#   BUILD_ARGS       optional, space-separated K=V pairs
+#   BUILD_SECRETS    optional, space-separated buildkit secret specs, each
+#                    `id=NAME,src=PATH`. Pass the SPEC only -- the caller
+#                    writes the secret to PATH itself. A secret VALUE must
+#                    never be passed here: it would land in this script's
+#                    argv and therefore in the CI log.
+#   EXTRA_LABELS     optional, space-separated K=V pairs
+#   GIT_SHA          optional, stamped as org.opencontainers.image.revision
+#   IMAGE_SOURCE     optional, stamped as org.opencontainers.image.source
+#
+# Note on shell style below: every conditional is a full `if` block, never
+# `[ -n "$X" ] && cmd`. Under `set -e` the && form exits the script whenever the
+# test is false, which is a real bug this estate has already shipped once.
+
+set -euo pipefail
+
+if [ -z "${IMAGE:-}" ]; then
+  echo "build-image: IMAGE is required" >&2
+  exit 2
+fi
+
+REGISTRY="${REGISTRY:-localhost:5000}"
+BUILDKIT_HOST="${BUILDKIT_HOST:-unix:///run/buildkit/buildkitd.sock}"
+CONTEXT="${CONTEXT:-.}"
+DOCKERFILE_DIR="${DOCKERFILE_DIR:-$CONTEXT}"
+PUSH_LATEST="${PUSH_LATEST:-1}"
+VERSION_TAG="${VERSION_TAG:-}"
+
+if [ -z "${SHA_TAG:-}" ]; then
+  SHA_TAG="$(git rev-parse --short HEAD)"
+fi
+
+REF="${REGISTRY}/${IMAGE}:${SHA_TAG}"
+
+TAR_FILE="$(mktemp)"
+trap 'rm -f "$TAR_FILE"' EXIT
+
+args=(
+  --frontend dockerfile.v0
+  --local "context=${CONTEXT}"
+  --local "dockerfile=${DOCKERFILE_DIR}"
+)
+
+if [ -n "${DOCKERFILE_NAME:-}" ]; then
+  args+=(--opt "filename=${DOCKERFILE_NAME}")
+fi
+if [ -n "${TARGET:-}" ]; then
+  args+=(--opt "target=${TARGET}")
+fi
+for spec in ${BUILD_SECRETS:-}; do
+  # Opaque id=NAME,src=PATH -- buildkit reads the file, nothing is echoed.
+  args+=(--secret "${spec}")
+done
+if [ -n "${GIT_SHA:-}" ]; then
+  args+=(--opt "label:org.opencontainers.image.revision=${GIT_SHA}")
+fi
+if [ -n "${IMAGE_SOURCE:-}" ]; then
+  args+=(--opt "label:org.opencontainers.image.source=${IMAGE_SOURCE}")
+fi
+for kv in ${BUILD_ARGS:-}; do
+  args+=(--opt "build-arg:${kv}")
+done
+for kv in ${EXTRA_LABELS:-}; do
+  args+=(--opt "label:${kv}")
+done
+
+echo "build-image: building ${REF}"
+buildctl --addr "$BUILDKIT_HOST" build "${args[@]}" --output "type=oci,dest=${TAR_FILE}"
+
+# buildctl has no registry-push equivalent to `buildah push`, so skopeo copies
+# the OCI archive up -- once per tag, from the same archive, so every tag points
+# at bit-identical content.
+push_tag() {
+  echo "build-image: pushing ${1}"
+  skopeo copy --dest-tls-verify=false "oci-archive:${TAR_FILE}" "docker://${1}"
+}
+
+push_tag "$REF"
+if [ "$PUSH_LATEST" = "1" ]; then
+  push_tag "${REGISTRY}/${IMAGE}:latest"
+fi
+if [ -n "$VERSION_TAG" ]; then
+  push_tag "${REGISTRY}/${IMAGE}:${VERSION_TAG}"
+fi
+
+echo "✓ build-image: ${REF} pushed"

.gitea/workflows/ci.yml

@@ -61,33 +61,163 @@ jobs:
             echo "version-tag=${REF#refs/tags/}" >> "$GITHUB_OUTPUT"
           fi
 
+      # infra#241: rootless buildah inside the gitea-runner pod can't remap the
+      # UID/GID range newuidmap/newgidmap need (the pod drops ALL capabilities
+      # and sets allowPrivilegeEscalation:false — both deliberate, it also holds
+      # CI secrets), so extracting a base-image layer that chowns outside the
+      # single available mapping fails with `ApplyLayer ... requested 0:42 for
+      # /etc/gshadow`. Main has been red on exactly this since 2026-08-13.
+      # buildctl instead talks to koala's host buildkitd over the socket the
+      # runner pod already mounts: the privileged extraction happens inside
+      # buildkitd on the host, never in this pod. Same migration agentsquad
+      # (infra#233) and segment-embedder (infra#243) already made — zero
+      # security tradeoff, unlike loosening the pod's capabilities.
+      #
+      # buildctl has no registry-push flag of its own: it writes an OCI archive
+      # and skopeo pushes that, which is also why every tag below is one copy of
+      # a single build rather than a rebuild.
       - name: Build and push to local registry
-        run: |
+      # infra#309: implementation lives in .gitea/build-image.sh, a
-          REGISTRY="localhost:5000"
+      # byte-identical copy of mathias/infra's ci-templates/build-image.sh.
-          REF="${REGISTRY}/${{ env.IMAGE }}:${{ steps.meta.outputs.sha-tag }}"
+      # Do not edit it here -- edit the canonical file, run `task ci:sync`,
-          buildah build \
+      # and infra's drift gate names any repo still carrying an old copy.
-            --label "org.opencontainers.image.revision=${{ github.sha }}" \
+        env:
-            --label "org.opencontainers.image.source=${{ github.repositoryUrl }}" \
+          SHA_TAG: ${{ steps.meta.outputs.sha-tag }}
-            -t ${REF} \
+          VERSION_TAG: ${{ steps.meta.outputs.version-tag }}
-            -t ${REGISTRY}/${{ env.IMAGE }}:latest \
+          GIT_SHA: ${{ github.sha }}
-            .
+          IMAGE_SOURCE: ${{ github.repositoryUrl }}
-          buildah push --tls-verify=false ${REF}
+        run: bash .gitea/build-image.sh
-          buildah push --tls-verify=false ${REGISTRY}/${{ env.IMAGE }}:latest
-          [[ -n "${{ steps.meta.outputs.version-tag }}" ]] && \
-            buildah push --tls-verify=false \
-              ${REGISTRY}/${{ env.IMAGE }}:${{ steps.meta.outputs.version-tag }} || true
-          echo "Image pushed to ${REF}"
 
-      - name: Smoke test
+      # The buildah-based smoke test that used to sit here is gone with the
-        run: |
+      # buildctl migration above: `buildah from ${REF}` would pull the image
-          REGISTRY="localhost:5000"
+      # back into the pod's buildah store and hit the very uid/gid wall the
-          REF="${REGISTRY}/${{ env.IMAGE }}:${{ steps.meta.outputs.sha-tag }}"
+      # migration exists to avoid — it would fail the job for a reason that has
-          CNAME="smoke-${{ steps.meta.outputs.sha-tag }}"
+      # nothing to do with the image being good. The `Verify rollout` step in
-          sudo k3s ctr images pull --plain-http ${REF}
+      # the deploy job below is the stronger check anyway: it requires the real
-          OUTPUT=$(timeout 5 sudo k3s ctr run --rm ${REF} ${CNAME} /tapir 2>&1 || true)
+      # pod to reach Ready in-cluster, not just the entrypoint to print its own
-          sudo k3s ctr containers delete ${CNAME} 2>/dev/null || true
+      # name for 5s. Tag builds lose their only image-level check (deploy is
-          echo "$OUTPUT" | grep -q "tapir" \
+      # main-push only) — accepted; a tag is always built from a main commit
-            && echo "Smoke test passed" \
+      # that already deployed.
-            || echo "Smoke test inconclusive: $OUTPUT"
 
-  # ── 3. Mirror to GitHub — skipped for now (SSH key rotation pending) ─
+  # ── 3. Deploy via infra repo + Flux ─────────────────────────────────────────
+  # Flux native image-automation can't scan localhost:5000 from inside k3s pods
+  # (mathias/infra k3s/flux/flux-system/image-automation.yaml) — this job
+  # mirrors cobalt-dingo's proven pattern instead: patch the infra repo's
+  # manifest directly on every push to main, then annotate Flux for a fast
+  # reconcile. Fixes infra#111 (image built+pushed but manifest bump was
+  # manual, so merged features silently didn't deploy).
+  deploy:
+    name: Deploy via GitOps
+    needs: build
+    runs-on: self-hosted
+    if: github.ref == 'refs/heads/main' && github.event_name == 'push'
+    environment: staging
+    steps:
+      - name: Update image tag in infra repo
+        env:
+          IMAGE_TAG: ${{ needs.build.outputs.image-tag }}
+          DEPLOY_KEY: ${{ secrets.INFRA_DEPLOY_KEY }}
+        run: |
+          set -euo pipefail
+          # infra#311: fail on an empty tag BEFORE patching anything. The
+          # grep verification further down CANNOT catch it -- with IMAGE_TAG
+          # empty its pattern matches the very corruption it exists to reject.
+          # The result is `image: <repo>:` in the manifest, which is invalid
+          # YAML, so Flux fails the ENTIRE apps Kustomization and every app
+          # stops reconciling, not just this one. That happened on 2026-08-25.
+          [ -n "${IMAGE_TAG:-}" ] || { echo "IMAGE_TAG is empty - refusing to patch"; exit 1; }
+
+          # INFRA_DEPLOY_KEY is a Gitea org secret (mathias org), already
+          # configured per docs/cd-pipeline.md in the infra repo — same key
+          # cobalt-dingo and brain-gardener use, no new secret needed.
+          mkdir -p ~/.ssh
+          echo "$DEPLOY_KEY" > ~/.ssh/id_infra
+          chmod 600 ~/.ssh/id_infra
+          ssh-keyscan -p 30022 10.0.1.20 >> ~/.ssh/known_hosts 2>/dev/null
+
+          export GIT_SSH_COMMAND="ssh -i ~/.ssh/id_infra -o IdentitiesOnly=yes"
+          rm -rf /tmp/infra
+          git clone -b main ssh://git@10.0.1.20:30022/mathias/infra.git /tmp/infra
+          cd /tmp/infra
+
+          DEPLOYMENT="k3s/apps/tapir/deployment.yaml"
+          # In-place update of the image tag. sed (not yq) so we don't
+          # depend on additional tooling on the runner — same as cobalt-dingo.
+          sed -i "s|image: localhost:5000/tapir:.*|image: localhost:5000/tapir:${IMAGE_TAG}|" "$DEPLOYMENT"
+
+          # Verify the patch took effect.
+          grep -q "localhost:5000/tapir:${IMAGE_TAG}" "$DEPLOYMENT" \
+            || { echo "✗ image tag patch failed"; exit 1; }
+
+          if git diff --quiet "$DEPLOYMENT"; then
+            echo "ℹ image tag unchanged — skipping push"
+          else
+            git -c user.name="tapir CI" \
+                -c user.email="ci@tapir.local" \
+                commit -m "chore(deploy): tapir → ${IMAGE_TAG}" "$DEPLOYMENT"
+            git push origin main
+            echo "✓ pushed to infra repo"
+          fi
+
+          shred -u ~/.ssh/id_infra
+
+      - name: Trigger Flux reconcile (immediate)
+        run: |
+          # Without these annotations, Flux would still pick up the change
+          # within 30s (the apps Kustomization interval). The annotations
+          # cut latency to ~1s.
+          kubectl -n flux-system annotate gitrepository flux-system \
+            reconcile.fluxcd.io/requestedAt="$(date +%s)" --overwrite
+          kubectl -n flux-system annotate kustomization apps \
+            reconcile.fluxcd.io/requestedAt="$(date +%s)" --overwrite
+
+      - name: Wait for Flux to apply new image
+        env:
+          IMAGE_TAG: ${{ needs.build.outputs.image-tag }}
+        run: |
+          # Poll the Deployment spec until it reflects the new tag.
+          # Bound to 60s so a stuck Flux doesn't hang CI.
+          EXPECTED="localhost:5000/tapir:${IMAGE_TAG}"
+          for i in $(seq 1 60); do
+            CURRENT=$(kubectl get deploy tapir -n tapir \
+              -o jsonpath='{.spec.template.spec.containers[0].image}' 2>/dev/null || echo "")
+            if [ "$CURRENT" = "$EXPECTED" ]; then
+              echo "✓ Flux applied new image after ${i}s"
+              break
+            fi
+            sleep 1
+          done
+          # Final assertion (in case the loop exited without matching).
+          kubectl get deploy tapir -n tapir \
+            -o jsonpath='{.spec.template.spec.containers[0].image}' \
+            | grep -qx "$EXPECTED" \
+            || { echo "✗ Flux did not apply new image within 60s"; exit 1; }
+
+      - name: Verify rollout
+        run: |
+          kubectl rollout status deployment/tapir \
+            --namespace tapir \
+            --timeout=120s \
+          || {
+            echo "── pod status ──"
+            kubectl get pods -n tapir -o wide
+            echo "── events ──"
+            kubectl get events -n tapir --sort-by='.lastTimestamp' | tail -20
+            echo "── describe ──"
+            kubectl describe pods -n tapir -l app=tapir | tail -40
+            exit 1
+          }
+
+      - name: Confirm pod running new image
+        env:
+          IMAGE_TAG: ${{ needs.build.outputs.image-tag }}
+        run: |
+          kubectl get pods -n tapir \
+            -l app=tapir \
+            --field-selector=status.phase=Running \
+            -o jsonpath='{.items[*].spec.containers[0].image}' \
+          | grep -q "localhost:5000/tapir:${IMAGE_TAG}" \
+          && echo "✓ pod running new image" \
+          || { echo "✗ pod image mismatch"; exit 1; }
+
+  # ── 4. Mirror to GitHub — skipped for now (SSH key rotation pending) ─

.gitignore

@@ -11,3 +11,14 @@
 .env.*
 !.env.example
 *.local
+
+# Spike media (#28): real recordings, their transcripts and derived analyses are
+# private third-party content and this repo is public. Only synthetic fixtures
+# are committed — see scripts/spike-media/README.md.
+/scripts/spike-media/*.mov
+/scripts/spike-media/*.mp4
+/scripts/spike-media/*.wav
+/scripts/spike-media/*.srt
+/scripts/spike-media/*.json
+/scripts/spike-media/*.html
+!/scripts/spike-media/fixtures/

DECISIONS.md

@@ -123,7 +123,9 @@ secret surface (see VISION Stage 2); it gets the existing, vetted secret path.
 ## ADR-007 — Captions-first; audio-download + speech-to-text deferred
 
 **Status:** Accepted (2026-06-02). Acquisition mechanism superseded by ADR-010 (Data API
-`captions.download` → player/timedtext baseUrl); captions-first stance and STT deferral stand.
+`captions.download` → player/timedtext baseUrl); captions-first stance and STT deferral stand
+**for YouTube**. STT deferral superseded **for user-supplied audio only** by ADR-033; the
+YouTube audio-download rejection is unchanged.
 
 **Context.** YouTube's Data API does not expose transcripts. Options: official captions
 (clean, limited coverage) vs. audio download + local Whisper (broad coverage, ToS-grey,
@@ -1314,6 +1316,191 @@ roll back. No schema, no structural change.
 
 ---
 
+## ADR-033 — Upload audio, get a private Swedish transcript, discuss it (user-supplied STT)
+
+**Status:** Proposed (2026-10-01). **Decider: Mathias — pending.** Issue #33 (audio-only subset of
+#28). **Supersedes ADR-007's speech-to-text deferral for user-supplied audio only.** The
+YouTube audio-download rejection in ADR-007 **stands untouched**. Builds on ADR-012 (isolation),
+ADR-018 (single replica), ADR-021 (**the constraint**), ADR-022 (endpoint chain), ADR-025
+(honest status), ADR-027 (chat). ADR-014 item 4 ("Whisper reconsideration is a future ADR,
+gated on caption data") is a different question — whether to transcribe YouTube videos with no
+captions — and this ADR does not answer it. Measurements: `docs/research/audio-length-ceiling-2026-10-01.md`.
+
+**Context.** The maintainer has a concrete need: a Swedish Teams voice-only export, 30–90
+minutes, to upload and then *discuss* (what was decided, what is open, who owes what). Every
+piece exists — a Whisper endpoint on the gateway (infra#274), chat-with-transcript (ADR-027) —
+and nothing connects them.
+
+ADR-007 deferred STT for three reasons: yt-dlp fragility, ToS-grey scraping, and GPU contention
+with the JEPA PoC. Upload-STT shares **none** of them: the user owns the file, nothing is
+scraped, there is no caption rate gate, and the contention moved from koala to iguana. What it
+does share is landing Whisper in the codebase, which ADR-007 deferred — so a new ADR is required,
+and ADR-017's process note records the one time a guardrail-reversing change shipped without one.
+
+**The constraint that shapes everything.** ADR-021 made `transcripts` a shared, **non-RLS**
+table on the explicit grounds that it holds only public caption content and nothing
+user-identifying. A recorded internal meeting is the opposite: private third-party content owned
+by one user. The existing isolation test asserts `transcripts` *is* non-RLS, so a naive reuse
+would leak across tenants **and pass CI green**.
+
+**What was measured, not argued** (research doc above):
+- **No transcription ceiling up to 90 minutes.** One 174 MB request through the LiteLLM NodePort
+  took 440 s (~12× realtime). The first failure is the public ingress `llm-api.d-ma.be`
+  returning `413` above roughly 8–10 MB (~5 min of WAV) — not whisper, and not a Metal OOM.
+  The 8:20 failure recorded in #33 is consistent with that cap.
+- **`response_format=srt` survives LiteLLM, wrapped** in a JSON envelope (`{"text": "<srt>"}`).
+- **The two backends differ.** `iguana/kb-whisper` returns whole-second segments;
+  `berget/kb-whisper` returns sub-second times and a different `language` token.
+- **The existing chat path truncates at 18,000 chars (~22 min of speech), keeping the head.**
+  #33 assumed chat "retrieves rather than generates" and so has no length ceiling. It does
+  not retrieve; it stuffs the transcript into the prompt and cuts the tail.
+
+**Unknowns, stated.** Real-speech accuracy; compressed-input behaviour; concurrent long jobs on
+iguana; whether the models chat offers can hold an hour of Swedish (~50 k chars); and the Metal
+OOM, deliberately not provoked.
+
+**Decision.**
+
+1. **Audio only; transcript is the only thing kept.** An authenticated user uploads an audio
+   file; Tapir transcribes it and stores the transcript. The audio itself is deleted after
+   transcription (temp file, removed in a `defer`, swept at start-up) — no custody problem to
+   solve, and nothing to hand to #31. Video is rejected with a message naming #28. YouTube
+   audio-download stays rejected. No public/anonymous upload (ADR-008).
+
+2. **Private store: a separate RLS-scoped table, `uploaded_transcripts`** (resolves #28's
+   decision 1). `user_id NOT NULL REFERENCES users ON DELETE CASCADE`, `ENABLE` + `FORCE ROW LEVEL
+   SECURITY`, the same `tapir.current_user_id` policy as every other user-owned table. It carries
+   the transcript text and its timestamped segments (re-transcribing 90 minutes costs ~7
+   minutes; discarding the timestamps is irreversible), `status`, the `endpoint` that served
+   it, `audio_seconds`, and a typed `error_kind`. **Not** a `user_id` column on `transcripts`
+   with a partial policy, even though that would reuse more code: ADR-021's exception is
+   precisely that *that table holds no private row*, the isolation test asserts it has no
+   `user_id`, and a partial policy would turn "single exception" into "single exception, with a
+   private lane inside it" — the shape under which a future query drops the policy clause and
+   leaks. Two tables keep both boundaries independently assertable.
+
+3. **The isolation test asserts both boundaries and the set.** (a) User A cannot read, update or
+   delete user B's uploaded transcript; no GUC set → deny-all. (b) `transcripts` is still
+   non-RLS with no `user_id`. (c) **The set of public-schema tables with row security off is
+   asserted exactly** — so a new table added without RLS fails CI instead of passing it. (c) is
+   new; the existing test only checks the named tables.
+
+4. **Validate by probing, reject clearly.** Format and duration cannot be read from a filename
+   or MIME type — a Teams "voice-only" recording is typically an `.mp4` container with no video
+   stream, which is indistinguishable from video by extension. The upload is probed
+   (`ffprobe`): accept any container with an audio stream and **no** video stream; reject video
+   with "video is #28"; reject unreadable files. Limits, all configurable, all rejected
+   synchronously with a specific message rather than by timeout: **200 MB** upload and
+   **90 minutes** duration. 90 minutes is *the longest length measured*, not a ceiling that was
+   found; raising it is a measurement, not an edit.
+
+5. **Transcode server-side, always, to 16 kHz mono PCM WAV** — even when the input already is,
+   rather than assuming. whisper-server's `--convert` is rejected (its `--tmp-dir` is created at
+   process start and rots under macOS periodic cleanup). The transcoder sits behind a port; the
+   tests use a fake. **This puts `ffmpeg` and `ffprobe` in the runtime image**, which is
+   distroless-static today. **Provisional: copy static ffmpeg/ffprobe binaries into the existing
+   distroless image** (no shell, no package manager retained). This is a heavy new dependency
+   and is **the maintainer's call** — see Open items.
+
+6. **Transcription adapter via the gateway, never the public hostname.** A `Transcriber` port;
+   the implementation calls `POST {TAPIR_GATEWAY_URL}/audio/transcriptions` where the gateway
+   URL is the in-cluster NodePort (`koala:30401/v1`). Primary `iguana/kb-whisper`, fallback
+   `berget/kb-whisper` (immune to iguana contention, which can wedge permanently — infra#273).
+   - **Fallback triggers on any primary failure** (transport error, timeout, non-2xx). It never
+     branches on the body: `{"error":"failed to process audio"}` has had at least five distinct
+     causes.
+   - **`response_format=verbose_json`**, normalised to one segment shape: times clamped to the
+     audio duration (the final segment can end past it), one language token. Not `srt`: the
+     envelope adds a parse step and iguana's segment timing is the same either way.
+   - **No chunking in v1.** Nothing failed to 90 minutes (measured). Chunk-and-stitch is the
+     answer *if* the cap is raised or smaller requests become necessary — see triggers.
+   - **The serving endpoint is recorded per row**, and surfaced, because the two backends
+     differ in timestamp granularity: which one served a transcript is a quality fact.
+   - **Berget is a cloud endpoint.** On fallback the audio of a private meeting leaves the
+     homelab. #33 asks for this fallback; it is configurable (`TAPIR_TRANSCRIBE_FALLBACK_MODEL`,
+     empty disables), recorded per row, and called out in Open items rather than defaulted
+     silently.
+   - Swedish only (`language=sv`). Language detection and routing (#28 decision 2) are not
+     touched.
+   - `-H 'Expect:'`, multipart field `file` — encoded in the adapter and its tests, not rediscovered.
+
+7. **Jobs run in-process, one at a time (ADR-018 stands).** Transcription is minutes (~7.5 min
+   for 90) but a single goroutine behind a one-slot semaphore: two simultaneous 90-minute jobs on
+   iguana were not measured, so concurrency is 1 until they are. No worker, no new deployable,
+   no second replica. Honest status per ADR-025: `queued → transcribing → done | failed`, polled
+   by the same `/status` pattern, every terminal state a rendered card, never a stuck spinner.
+   Audio is not retained, so a restart mid-job **cannot resume**: at start-up any
+   `queued`/`transcribing` row is marked `failed` with `error_kind = interrupted`, and the user
+   re-uploads. The single-replica assumption gains a second dependent (the semaphore is
+   process-local) and is recorded here.
+
+8. **Chat reuses `chat.Service`; it is not forked.** The uploaded-transcript chat handlers
+   look the row up through the RLS-scoped store, so a non-owner gets 404 by the same mechanism
+   as ADR-027's summary gate. ADR-027's decisions (ephemeral, switchable models, trust-the-model)
+   carry over unchanged. **One thing must differ:** the 18,000-char default cuts a one-hour
+   meeting to its first third, so uploaded transcripts take their own cap,
+   `TAPIR_MAX_UPLOADED_TRANSCRIPT_CHARS`. **Provisional default 120,000 chars (~2.4 h at
+   measured density)**, with the existing `Truncated` flag surfaced honestly. The default is a
+   guess until the chat slice measures what the offered models actually accept; it is not
+   evidence that they can.
+
+9. **Not decided here.** Rich `AVSNITT`/`ALTERNATIV` analysis (gated on #30 — `analyze_srt.py`
+   hardcodes `max_tokens: 6000`); video, keyframes, synced playback (#28); language detection;
+   per-transcript deletion UX (account deletion cascades; see Open items); retention of audio.
+
+**On the sequencing question #28 asks of itself** — *is this the thing that best serves the
+Stage-0 gate, or the most interesting thing to build?* Recorded answer: it proceeds **ahead of
+#32** (the demand test) as the maintainer's explicit, knowing call (#28, comment of 2026-10-01),
+on a concrete need rather than speculation. It is not a return-nudge (ADR-020/027): it adds
+nothing that prompts a user to come back, so it does not contaminate the unprompted-return signal
+ADR-016 measures. The honest risk is the one ADR-016 names: it is larger than anything ADR-020
+deferred. The scope cut — chat first, no rich analysis, no video — is what keeps it bounded.
+
+**Consequences.**
+- **Positive.** The user's request is met end to end with existing parts. ADR-021's exception stays
+  singular and now provably so. The measured ceiling removes chunking from v1. No audio is stored.
+- **Negative / watch.**
+  - Tapir's runtime image gains `ffmpeg`: a larger image and a new CVE surface to patch.
+  - iguana becomes load-bearing for a user-facing feature (#28 decision 3; infra#273 is still the
+    dependency that matters) — a Metal OOM now means *every upload fails over to berget*, i.e.
+    private audio leaving the homelab, silently unless the per-row endpoint is surfaced.
+  - A 90-minute job holds the single slot for ~7.5 minutes; a second user's upload queues behind it.
+  - Chat over a long transcript is bounded by model context, not by anything in Tapir.
+  - Tapir's own ingress body limit for a 200 MB upload is **unverified** (it lives in `infra`).
+
+**Validation / resolution triggers** (each observable by someone other than the author):
+- *Chunking*: build it when the duration cap is raised above 90 min, or when a measured
+  request at the current cap fails with whisper-side (not ingress) errors.
+- *Berget-first*: if more than a quarter of uploads in any week are served by the fallback,
+  swap the primary/fallback order and revisit infra#273.
+- *Retrieval chat*: if the chat slice finds an offered model cannot answer a question whose
+  answer is in the last third of a 60-minute transcript, chunked retrieval replaces the larger cap.
+- *Worker*: if a transcription job ever degrades the reading UI, or Tapir runs more than one
+  replica, ADR-018 is revisited.
+- *The decision itself*: if the maintainer has not used the feature unprompted within the
+  ADR-016 window, it was the interesting thing, not the useful one — remove it.
+
+**Open items this ADR creates.**
+- **DECISION NEEDED — ffmpeg in the runtime image.** *Options:* (A) static ffmpeg/ffprobe copied
+  into distroless **(recommended — the only option that satisfies "accept m4a/mp3/opus/mp4,
+  validate duration, reject video"; pure-Go decoders cover wav/mp3 only)**; (B) accept WAV
+  only, convert client-side (fails "I just want to upload it"); (C) a transcode sidecar
+  (second deployable, ADR-018's reasoning against). *Blocks:* the Dockerfile slice, not the
+  tests, which use the port.
+- **DECISION NEEDED — berget fallback for private audio.** *Options:* (A) enabled by default as
+  #33 specifies, with the serving endpoint shown **(recommended, matches the issue)**; (B)
+  disabled by default, a failed iguana job fails honestly and the user retries. *Blocks:*
+  the production default only; the adapter ships either way.
+- Verify Tapir's ingress `proxy-body-size` against the 200 MB cap (`infra`).
+- Per-transcript deletion is not built; account deletion cascades. A private-content store
+  without a delete button is a gap worth a ticket, not a silent omission.
+
+**Reversibility.** Additive: one table (down-migration drops it), one adapter, one upload
+surface, one image change. Removing the upload route removes the feature; `transcripts`, `videos`
+and ADR-021 are untouched.
+
+---
+
 ## Rejected alternatives
 
 Approaches considered during the 2026-06-02 planning + grill session and **deliberately not
@@ -1338,6 +1525,9 @@ maps to the ADR that settles it.
 | Reverse the Dex-write invite flow (Google OIDC only) | Some intended Future-B users won't use Google; OIDC-only leaves them with no onboarding path — invite flow is load-bearing | ADR-017 |
 | k8s CronJob for scheduled discovery (vs in-process) | At Future-B scale the in-process scheduler is simpler to deploy; CronJob's failure-isolation benefit was weighed and traded away knowingly (revisit if >1 replica or load grows) | ADR-018 |
 | Cached-transcript-first onboarding burst (instant, zero-fetch picks) | Live pilot DB: ~3% cross-user video overlap, 0 cached-and-unsummarized, a new user's newest-20 are 20/20 uncached — newest-first and cached-first are structurally incompatible. Empty lever at pilot scale | ADR-028 |
+| Uploaded transcripts as a `user_id` column + partial RLS policy on `transcripts` | Turns ADR-021's single exception into an exception with a private lane inside it; the isolation test asserts `transcripts` has no `user_id`. A separate RLS table keeps both boundaries independently assertable | ADR-033 |
+| whisper-server `--convert` for server-side transcode | Its `--tmp-dir` is created at process start and rots under macOS periodic cleanup | ADR-033 |
+| Calling the public `llm-api.d-ma.be` from in-cluster Tapir | Its ingress returns 413 above ~8–10 MB (~5 min of WAV); the NodePort has no such cap | ADR-033 |
 
 If a future case genuinely reopens one of these, that's a new ADR superseding the relevant one —
 not a silent reversal.

cmd/tapir/main.go

@@ -207,6 +207,27 @@ func cmdServe(ctx context.Context, log *slog.Logger) error {
 		log.Info("web chat enabled (stored-transcript only)", "models", chatModels(cfg))
 	}
 
+	// Audio upload -> private transcript (ADR-033). Off unless TAPIR_UPLOADS_ENABLED:
+	// the image must carry ffmpeg first. The worker is IN-PROCESS and one job at a
+	// time, so it inherits the single-replica assumption below (its queue is
+	// process-local); it stops with the same ctx as the server.
+	intake, worker, err := buildUploads(ctx, cfg, st, log)
+	if err != nil {
+		return err
+	}
+	if intake != nil {
+		app.Uploads = intake
+		app.UploadLimits = uploadLimits(cfg)
+		go worker.Run(ctx)
+		log.Info("audio upload enabled", "model", cfg.TranscribeModel, "fallback", cfg.TranscribeFallbackModel,
+			"max_mb", cfg.UploadMaxBytes>>20, "max_minutes", int(cfg.UploadMaxDuration.Minutes()))
+		log.Warn("audio upload assumes a SINGLE replica: the transcription queue is process-local (ADR-033, ADR-018)")
+		if cfg.TranscribeFallbackModel != "" {
+			log.Warn("audio upload fallback is a CLOUD endpoint: if the primary fails, private audio leaves the homelab",
+				"fallback", cfg.TranscribeFallbackModel)
+		}
+	}
+
 	// User onboarding is handled by the IdP (Authentik invite flow), not Tapir —
 	// the Dex local-password provisioning path was removed (ADR-019). An
 	// authenticated subject with no Tapir user is routed to /register.

cmd/tapir/uploads.go

@@ -0,0 +1,85 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log/slog"
+	"net/http"
+	"os"
+	"path/filepath"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/audio"
+	"git.d-ma.be/mathias/tapir/internal/adapters/transcribe"
+	"git.d-ma.be/mathias/tapir/internal/config"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+// uploadsStore is the store surface the upload wiring needs. *store.Store
+// satisfies it.
+type uploadsStore interface {
+	uploads.Store
+	uploads.WorkerStore
+	FailInterruptedUploads(ctx context.Context) (int, error)
+}
+
+// buildUploads wires audio upload and private transcription (ADR-033). It returns
+// nils when uploads are disabled — never a typed-nil Intake, which would mount the
+// /uploads routes over nothing. When enabled it, in order: prepares a private spool
+// directory, deletes any audio a crashed run left in it, fails the rows such a run
+// left queued or transcribing (the audio is not retained, so they can never
+// resume), and only then builds the intake and the worker. The caller runs the
+// worker. The fallback endpoint is a cloud service; an empty
+// TAPIR_TRANSCRIBE_FALLBACK_MODEL keeps private audio on the homelab.
+func buildUploads(ctx context.Context, cfg config.Config, st uploadsStore, log *slog.Logger) (*uploads.Intake, *uploads.Worker, error) {
+	if !cfg.UploadsEnabled {
+		return nil, nil, nil
+	}
+	if cfg.GatewayURL == "" {
+		return nil, nil, fmt.Errorf("uploads: TAPIR_UPLOADS_ENABLED is set but there is no TAPIR_GATEWAY_URL to transcribe through")
+	}
+
+	if err := os.MkdirAll(cfg.UploadTmpDir, 0o700); err != nil {
+		return nil, nil, fmt.Errorf("uploads: spool dir: %w", err)
+	}
+	removeSpooledAudio(cfg.UploadTmpDir, log)
+
+	n, err := st.FailInterruptedUploads(ctx)
+	if err != nil {
+		return nil, nil, fmt.Errorf("uploads: fail interrupted rows: %w", err)
+	}
+	if n > 0 {
+		log.Warn("uploads: marked interrupted by a restart (audio is not retained; they must be re-uploaded)", "count", n)
+	}
+
+	// No overall client timeout: a 90 minute recording takes minutes. The deadline
+	// is the worker's per-job context.
+	client := &http.Client{}
+	primary := transcribe.NewGateway(cfg.GatewayURL, cfg.GatewayKey, cfg.TranscribeModel, client)
+	var fallback transcribe.Backend
+	if cfg.TranscribeFallbackModel != "" && cfg.TranscribeFallbackModel != cfg.TranscribeModel {
+		fallback = transcribe.NewGateway(cfg.GatewayURL, cfg.GatewayKey, cfg.TranscribeFallbackModel, client)
+	}
+
+	worker := uploads.NewWorker(st, &audio.FFmpeg{Bin: cfg.FFmpegBin}, transcribe.NewChain(primary, fallback),
+		cfg.UploadTmpDir, cfg.TranscribeTimeout)
+	intake := uploads.NewIntake(st, &audio.FFprobe{Bin: cfg.FFprobeBin}, worker,
+		uploadLimits(cfg), cfg.UploadTmpDir)
+	return intake, worker, nil
+}
+
+func uploadLimits(cfg config.Config) uploads.Limits {
+	return uploads.Limits{MaxBytes: cfg.UploadMaxBytes, MaxDuration: cfg.UploadMaxDuration}
+}
+
+// removeSpooledAudio deletes the spool files this process family creates
+// (upload-*, wav-*) and nothing else.
+func removeSpooledAudio(dir string, log *slog.Logger) {
+	for _, pat := range []string{"upload-*", "wav-*"} {
+		matches, _ := filepath.Glob(filepath.Join(dir, pat))
+		for _, m := range matches {
+			if err := os.Remove(m); err != nil {
+				log.Warn("uploads: remove leftover audio", "path", m, "err", err)
+			}
+		}
+	}
+}

cmd/tapir/uploads_test.go

@@ -0,0 +1,127 @@
+package main
+
+import (
+	"context"
+	"errors"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/config"
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// fakeUploadsStore satisfies the store surface buildUploads needs, and counts the
+// start-up sweep.
+type fakeUploadsStore struct {
+	sweeps   int
+	sweepErr error
+}
+
+func (f *fakeUploadsStore) CreateUpload(context.Context, string, string) (string, error) {
+	return "id", nil
+}
+
+func (f *fakeUploadsStore) FailUpload(context.Context, string, string, domain.UploadErrorKind) error {
+	return nil
+}
+
+func (f *fakeUploadsStore) StartTranscribing(context.Context, string, string) error { return nil }
+
+func (f *fakeUploadsStore) CompleteUpload(context.Context, string, string, domain.Transcription) error {
+	return nil
+}
+
+func (f *fakeUploadsStore) FailInterruptedUploads(context.Context) (int, error) {
+	f.sweeps++
+	return 2, f.sweepErr
+}
+
+func uploadCfg(t *testing.T, enabled bool) config.Config {
+	t.Helper()
+	var cfg config.Config
+	cfg.GatewayURL = "http://gateway.invalid/v1"
+	cfg.UploadsEnabled = enabled
+	cfg.UploadMaxBytes = 200 << 20
+	cfg.UploadTmpDir = filepath.Join(t.TempDir(), "spool")
+	cfg.TranscribeModel = "iguana/kb-whisper"
+	cfg.TranscribeFallbackModel = "berget/kb-whisper"
+	return cfg
+}
+
+func TestBuildUploads_DisabledMountsNothingAndSweepsNothing(t *testing.T) {
+	st := &fakeUploadsStore{}
+
+	in, w, err := buildUploads(context.Background(), uploadCfg(t, false), st, quietLog())
+
+	require.NoError(t, err)
+	assert.Nil(t, in, "a typed-nil Intake must never reach web.App.Uploads: it would mount the routes over nothing")
+	assert.Nil(t, w)
+	assert.Zero(t, st.sweeps)
+}
+
+func TestBuildUploads_EnabledFailsInterruptedRowsBeforeAnythingElseStarts(t *testing.T) {
+	st := &fakeUploadsStore{}
+
+	in, w, err := buildUploads(context.Background(), uploadCfg(t, true), st, quietLog())
+
+	require.NoError(t, err)
+	require.NotNil(t, in)
+	require.NotNil(t, w)
+	assert.Equal(t, 1, st.sweeps,
+		"audio is not retained, so rows left queued/transcribing by a restart can never resume: fail them at start-up, not never")
+}
+
+func TestBuildUploads_SweepFailureStopsStartup(t *testing.T) {
+	st := &fakeUploadsStore{sweepErr: errors.New("db down")}
+
+	_, _, err := buildUploads(context.Background(), uploadCfg(t, true), st, quietLog())
+
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "db down")
+}
+
+func TestBuildUploads_CreatesTheSpoolDirectoryPrivately(t *testing.T) {
+	cfg := uploadCfg(t, true)
+
+	_, _, err := buildUploads(context.Background(), cfg, &fakeUploadsStore{}, quietLog())
+
+	require.NoError(t, err)
+	info, err := os.Stat(cfg.UploadTmpDir)
+	require.NoError(t, err)
+	assert.True(t, info.IsDir())
+	assert.Equal(t, os.FileMode(0o700), info.Mode().Perm(), "spooled audio is private third-party content")
+}
+
+func TestBuildUploads_EnabledWithoutAGatewayIsAnError(t *testing.T) {
+	cfg := uploadCfg(t, true)
+	cfg.GatewayURL = ""
+
+	_, _, err := buildUploads(context.Background(), cfg, &fakeUploadsStore{}, quietLog())
+
+	assert.Error(t, err, "uploads on with nowhere to transcribe must fail loudly at start-up, not at the first upload")
+}
+
+func TestBuildUploads_RemovesAudioLeftBehindByACrash(t *testing.T) {
+	cfg := uploadCfg(t, true)
+	require.NoError(t, os.MkdirAll(cfg.UploadTmpDir, 0o700))
+	leftovers := []string{"upload-123", "wav-456"}
+	for _, n := range leftovers {
+		require.NoError(t, os.WriteFile(filepath.Join(cfg.UploadTmpDir, n), []byte("audio"), 0o600))
+	}
+	unrelated := filepath.Join(cfg.UploadTmpDir, "someone-elses-file")
+	require.NoError(t, os.WriteFile(unrelated, []byte("keep"), 0o600))
+
+	_, _, err := buildUploads(context.Background(), cfg, &fakeUploadsStore{}, quietLog())
+
+	require.NoError(t, err)
+	for _, n := range leftovers {
+		_, err := os.Stat(filepath.Join(cfg.UploadTmpDir, n))
+		assert.True(t, os.IsNotExist(err), "%s: audio is never retained, so a spooled file from a crashed run must not survive start-up", n)
+	}
+	_, err = os.Stat(unrelated)
+	assert.NoError(t, err, "only our own spool files are removed")
+}

docs/research/audio-length-ceiling-2026-10-01.md

@@ -0,0 +1,110 @@
+# Audio length ceiling — measured before designing chunking (tapir#33, 2026-10-01)
+
+Question (#33, "the main technical risk"): at what recording length does Swedish
+transcription through the LiteLLM gateway first fail, and with what failure mode?
+Measured **before** any chunking design, with iguana not holding a large ollama model,
+so the result is the ceiling and not the Metal OOM of infra#273.
+
+## Answer
+
+**No transcription ceiling was found at or below 90 minutes.** A single request carrying
+90 minutes of 16 kHz mono PCM (174 MB) transcribed in one pass in 440 s. The first thing
+that actually fails is **not whisper**: it is the public gateway ingress, which rejects the
+upload with `413` at roughly 5.5 minutes of WAV.
+
+| Route | First failure | Failure mode |
+|---|---|---|
+| `https://llm-api.d-ma.be` (openresty ingress) | between 8.0 MB and 10.5 MB body (~4–5.5 min of WAV) | `413 Request Entity Too Large`, HTML body from `openresty`, returned instantly |
+| `http://koala:30401` (LiteLLM NodePort) → `iguana/kb-whisper` | **none up to 90 min / 174 MB** | — |
+| `http://koala:30401` → `berget/kb-whisper` | **none up to 22 min / 42 MB** (largest tried) | — |
+
+The ingress limit is the reason the 8:20 / 15 MB recording in #33 "failed repeatedly" and a
+30-second slice succeeded. That failure was attributed to a Metal OOM; at least in the
+size-dependent form described, a body-size cap explains it without one. (The OOM
+mechanism in infra#273 is separately real; this run did not provoke it.)
+
+The exact ingress cap was bracketed (8.0 MB passed, 10.5 MB failed), not read from the
+ingress configuration, which lives in `infra`. "About 10 MB" is an inference.
+
+## Method
+
+- **Audio.** No known-good Swedish recording is available that is safe to use: the only real
+  one (`IMG_1233.wav`) is the private `/bygge` recording (#31 custody still open) and was
+  deliberately not touched. Instead: macOS `say -v Alva` (Swedish) on iguana, a 1,171-char
+  five-paragraph meeting script → 82.5 s of 16 kHz mono PCM WAV (`ffmpeg -ar 16000 -ac 1`).
+  Concatenated ×4, ×8, ×16, ×32, ×44, ×66 with Python `wave` (the issue's own recipe).
+  Nothing here is a real transcript; none of it is committed.
+- **Request.** `POST /v1/audio/transcriptions`, multipart `file=`, `model=iguana/kb-whisper`,
+  `response_format=verbose_json`, `-H 'Expect:'` (per #33's traps), master key supplied only
+  through `op run` and never on argv.
+- **Confound control.** `ollama /api/ps` on iguana was read before every step. Resident:
+  nothing, or `nomic-embed-text` (0.37 GB). Never a large model.
+
+## Results (`iguana/kb-whisper`, NodePort)
+
+| Audio | WAV size | HTTP | Wall | Realtime factor | Segments | Chars |
+|---|---|---|---|---|---|---|
+| 0:30 | 0.96 MB | 200 | 2.8 s | 11× | — | — |
+| 2:30 / 3:20 / 4:10 | 4.8 / 6.4 / 8.0 MB | 200 (ingress) | 12.9 / 16.0 / 19.9 s | 12× | — | — |
+| 5:30 (×4) | 10.5 MB | **413 ingress**, 200 NodePort in 27 s | 27 s | 12× | 121 | 4,557 |
+| 11:00 (×8) | 21.1 MB | 200 | 53 s | 12× | 241 | 9,094 |
+| 22:00 (×16) | 42.3 MB | 200 | 105 s | 13× | 482 | 18,201 |
+| 44:01 (×32) | 84.5 MB | 200 | 209 s | 13× | 966 | 36,438 |
+| 60:32 (×44) | 116.2 MB | 200 | 291 s | 12× | 1,332 | 50,145 |
+| 90:48 (×66) | 174.3 MB | 200 | 440 s | 12× | 2,005 | 75,165 |
+
+- Throughput is flat at ~12× realtime. 90 minutes costs ~7.5 minutes of wall time. This is
+  a minutes-long job, not a request/response one (see ADR-033 on ADR-018).
+- Segments per source repetition stay at 30.1–30.4 from ×4 to ×66, and segment times are
+  monotonic with `last_end` within 0.2 s of `duration`. No drift and no repetition-loop
+  hallucination appeared across 90 minutes. This is a weak check — the input is a repeated
+  loop and synthetic voice — not evidence of accuracy.
+- **Only `iguana/kb-whisper`'s `x4` run reported `last_end` 360.0 s against `duration`
+  330.2 s.** Whisper's final segment can end past the audio; clamp to `duration`.
+
+## Findings that change the design
+
+1. **Use the NodePort route, not the public hostname.** Tapir runs in-cluster beside the
+   gateway; `llm-api.d-ma.be` is the external door and has a body cap. #33 said "use the
+   gateway, never the host" and named the public URL; the gateway is right, that URL is not.
+   `TAPIR_GATEWAY_URL` already points at `koala:30401/v1`.
+2. **`response_format=srt` survives the LiteLLM hop, but wrapped.** The response is JSON
+   `{"text": "<the SRT>", "usage": …}`, not a bare SRT body — a client that parses the body
+   as SRT gets a parse error on `{`. #33 question 2 is answered: it passes through, inside
+   an envelope.
+3. **The two backends return different timestamp granularity.** `iguana/kb-whisper`
+   segments are **whole seconds** (`0→1`, `1→4`, `4→7`); `berget/kb-whisper` returns real
+   sub-second times (`0.051→1.192`) and a different `language` token (`swedish` vs `sv`).
+   `verbose_json` and the default JSON were byte-identical on iguana. An adapter that
+   falls back between the two must normalise both, and "which endpoint served this" is
+   a quality fact, not just a log line.
+4. **The existing chat path cannot answer about most of a long recording.** At
+   828 chars/min (this fixture; real speech is plausibly a little denser),
+   `TAPIR_MAX_TRANSCRIPT_CHARS`'s default of **18,000** is ~21.7 minutes. A 60-minute
+   transcript is 50 k chars, a 90-minute one 75 k. `chat.Service` truncates by keeping
+   the head (`internal/adapters/chat/chat.go`), so questions about the last two-thirds of
+   a one-hour meeting get an answer from a transcript that does not contain it. #33 says
+   chat "retrieves rather than generates" and therefore has no ceiling; it **stuffs and
+   truncates**, and its ceiling is the model context. See ADR-033.
+5. **Chunking is not needed for v1 on length.** Nothing failed up to 90 minutes. The
+   chunking-and-stitching recipe in #33 (offset timestamps, renumber cues, strip `\r`) is
+   not built. It stays the answer *if* a longer cap is wanted or the OOM mechanism forces
+   smaller requests — see ADR-033's resolution triggers.
+
+## What this did not measure
+
+- **Metal OOM under contention.** Deliberately avoided. A big resident ollama model
+  during a long transcription is still the known way to wedge iguana permanently.
+- **Real speech.** TTS is clean, single-voice, no crosstalk or room noise. WER and
+  speaker overlap in a Teams export are untested. Speed is probably representative;
+  accuracy is not claimed.
+- **Compressed input.** Only WAV was sent. Whether `m4a`/`mp3`/`opus` can reach whisper
+  without a server-side transcode is not established here (ADR-033 treats it as needing one).
+- **Concurrency.** One request at a time. Two simultaneous 90-minute jobs on iguana were
+  not tried.
+- **`berget/kb-whisper` beyond 22 min / 42 MB.** Enough to know the fallback is not
+  size-capped at meeting length for 22 minutes; not proven at 90.
+- **Time-of-day / other tenants.** One afternoon, one run per length.
+
+Scripts were throwaway shell/Python in a session scratchpad and are not committed; the
+method above reproduces them.

internal/adapters/audio/ffmpeg.go

@@ -0,0 +1,45 @@
+package audio
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"os/exec"
+	"strings"
+)
+
+// FFmpeg converts any audio container to the strict format whisper wants: 16 kHz
+// mono signed 16-bit PCM WAV (ADR-033). It ALWAYS transcodes, even when the input
+// already is that format — a Teams export often is, and the transcode is then
+// nearly free; assuming it would be the bug. This replaces whisper-server's
+// --convert, whose --tmp-dir is created at process start and rots under macOS
+// periodic cleanup.
+type FFmpeg struct {
+	// Bin is the ffmpeg executable ("ffmpeg" resolves via PATH).
+	Bin string
+}
+
+// ToWAV writes src, converted, to dst. Video streams are dropped (-vn): by now the
+// file has already been probed as audio-only, but a cover-art stream must not end
+// up in the WAV.
+func (f *FFmpeg) ToWAV(ctx context.Context, src, dst string) error {
+	bin := f.Bin
+	if bin == "" {
+		bin = "ffmpeg"
+	}
+	cmd := exec.CommandContext(ctx, bin,
+		"-y", "-v", "error", "-i", src, "-vn", "-ac", "1", "-ar", "16000", "-c:a", "pcm_s16le", "-f", "wav", dst)
+	cmd.WaitDelay = killGrace
+	if out, err := cmd.CombinedOutput(); err != nil {
+		if ctxErr := ctx.Err(); ctxErr != nil {
+			return fmt.Errorf("audio: ffmpeg: %w", ctxErr)
+		}
+		return fmt.Errorf("audio: ffmpeg: %w: %s", err, strings.TrimSpace(string(out)))
+	}
+	// A zero exit that wrote nothing is still a failure: never hand whisper a
+	// missing file and let the error surface three layers later.
+	if info, err := os.Stat(dst); err != nil || info.Size() == 0 {
+		return fmt.Errorf("audio: ffmpeg wrote no output at %s", dst)
+	}
+	return nil
+}

internal/adapters/audio/ffmpeg_test.go

@@ -0,0 +1,60 @@
+package audio_test
+
+import (
+	"context"
+	"errors"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/audio"
+)
+
+func TestFFmpegToWAV_AsksForSixteenKilohertzMonoPCMAlways(t *testing.T) {
+	// Always transcode, even when the input already is 16 kHz mono: a Teams export
+	// often is, and the point is to not ASSUME (ADR-033). Records the argv.
+	rec := filepath.Join(t.TempDir(), "argv")
+	bin := stub(t, `printf '%s\n' "$@" > `+rec+`
+for a; do last="$a"; done
+printf data > "$last"`)
+
+	err := (&audio.FFmpeg{Bin: bin}).ToWAV(context.Background(), "/spool/in", filepath.Join(t.TempDir(), "out.wav"))
+
+	require.NoError(t, err)
+	raw, err := os.ReadFile(rec)
+	require.NoError(t, err)
+	argv := strings.Split(strings.TrimSpace(string(raw)), "\n")
+	joined := " " + strings.Join(argv, " ") + " "
+	for _, want := range []string{" -ac 1 ", " -ar 16000 ", " -c:a pcm_s16le ", " -vn ", " -i /spool/in "} {
+		assert.Contains(t, joined, want)
+	}
+	assert.True(t, strings.HasSuffix(strings.TrimSpace(joined), "out.wav"), "the output path is the last argument")
+}
+
+func TestFFmpegToWAV_NonzeroExitIsAnErrorCarryingStderr(t *testing.T) {
+	bin := stub(t, `echo "moov atom not found" >&2; exit 1`)
+
+	err := (&audio.FFmpeg{Bin: bin}).ToWAV(context.Background(), "/in", filepath.Join(t.TempDir(), "o.wav"))
+
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "moov atom not found")
+}
+
+func TestFFmpegToWAV_ZeroExitWithNoOutputIsAnError(t *testing.T) {
+	bin := stub(t, `exit 0`)
+
+	err := (&audio.FFmpeg{Bin: bin}).ToWAV(context.Background(), "/in", filepath.Join(t.TempDir(), "never-written.wav"))
+
+	require.Error(t, err)
+}
+
+func TestFFmpegToWAV_MissingBinaryIsAnError(t *testing.T) {
+	err := (&audio.FFmpeg{Bin: filepath.Join(t.TempDir(), "no-ffmpeg")}).ToWAV(context.Background(), "/in", "/out.wav")
+
+	require.Error(t, err)
+	assert.False(t, errors.Is(err, context.Canceled))
+}

internal/adapters/audio/ffprobe.go

@@ -0,0 +1,95 @@
+// Package audio holds the adapters that inspect and convert user-supplied audio
+// with ffmpeg's tools (ADR-033). They sit behind small ports so the use case is
+// tested with fakes and only this package knows a subprocess is involved.
+package audio
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"os/exec"
+	"strconv"
+	"time"
+
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+// FFprobe implements uploads.Prober by running ffprobe on the spooled file.
+type FFprobe struct {
+	// Bin is the ffprobe executable ("ffprobe" resolves via PATH).
+	Bin string
+}
+
+// killGrace bounds how long Probe waits for a cancelled ffprobe's pipes to close.
+const killGrace = 2 * time.Second
+
+// Probe runs ffprobe and classifies the outcome. A non-zero exit means ffprobe
+// looked at the file and could not read it: that is the FILE's fault and returns
+// uploads.ErrUnreadable. A missing binary, a cancelled context or unparsable
+// output is OURS and is returned as an ordinary error, so it is never reported to
+// the user as their file being bad.
+func (f *FFprobe) Probe(ctx context.Context, path string) (uploads.Probe, error) {
+	bin := f.Bin
+	if bin == "" {
+		bin = "ffprobe"
+	}
+	cmd := exec.CommandContext(ctx, bin,
+		"-v", "error", "-print_format", "json", "-show_format", "-show_streams", path)
+	cmd.WaitDelay = killGrace
+	out, err := cmd.Output()
+	if err != nil {
+		if ctxErr := ctx.Err(); ctxErr != nil {
+			return uploads.Probe{}, fmt.Errorf("audio: ffprobe: %w", ctxErr)
+		}
+		var exit *exec.ExitError
+		if errors.As(err, &exit) {
+			return uploads.Probe{}, fmt.Errorf("%w: ffprobe: %s", uploads.ErrUnreadable, exit.Stderr)
+		}
+		return uploads.Probe{}, fmt.Errorf("audio: run ffprobe: %w", err)
+	}
+	p, err := ParseProbe(out)
+	if err != nil {
+		return uploads.Probe{}, fmt.Errorf("audio: ffprobe output: %w", err)
+	}
+	return p, nil
+}
+
+// probeJSON is the subset of `ffprobe -print_format json` output that matters.
+type probeJSON struct {
+	Streams []struct {
+		CodecType   string `json:"codec_type"`
+		Disposition struct {
+			AttachedPic int `json:"attached_pic"`
+		} `json:"disposition"`
+	} `json:"streams"`
+	Format struct {
+		Duration string `json:"duration"`
+	} `json:"format"`
+}
+
+// ParseProbe turns ffprobe's JSON into a uploads.Probe. A video stream that is
+// only attached cover art (an mp3/m4a with an album picture) is not video. A
+// missing or "N/A" duration yields zero, which uploads.Validate rejects as
+// unreadable rather than accepting an empty recording.
+func ParseProbe(data []byte) (uploads.Probe, error) {
+	var j probeJSON
+	if err := json.Unmarshal(data, &j); err != nil {
+		return uploads.Probe{}, fmt.Errorf("decode: %w", err)
+	}
+	var p uploads.Probe
+	for _, s := range j.Streams {
+		switch s.CodecType {
+		case "audio":
+			p.HasAudio = true
+		case "video":
+			if s.Disposition.AttachedPic == 0 {
+				p.HasVideo = true
+			}
+		}
+	}
+	if secs, err := strconv.ParseFloat(j.Format.Duration, 64); err == nil && secs > 0 {
+		p.Duration = time.Duration(secs * float64(time.Second))
+	}
+	return p, nil
+}

internal/adapters/audio/ffprobe_test.go

@@ -0,0 +1,129 @@
+package audio_test
+
+import (
+	"context"
+	"errors"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/audio"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+func TestParseProbe(t *testing.T) {
+	tests := []struct {
+		name string
+		json string
+		want uploads.Probe
+	}{
+		{
+			name: "audio-only m4a",
+			json: `{"streams":[{"codec_type":"audio","disposition":{"attached_pic":0}}],"format":{"duration":"2701.5"}}`,
+			want: uploads.Probe{Duration: 2701*time.Second + 500*time.Millisecond, HasAudio: true},
+		},
+		{
+			name: "mp4 with a real video stream",
+			json: `{"streams":[{"codec_type":"video","disposition":{"attached_pic":0}},{"codec_type":"audio"}],"format":{"duration":"240.0"}}`,
+			want: uploads.Probe{Duration: 240 * time.Second, HasAudio: true, HasVideo: true},
+		},
+		{
+			name: "mp3 whose only video stream is attached cover art is NOT video",
+			json: `{"streams":[{"codec_type":"audio"},{"codec_type":"video","disposition":{"attached_pic":1}}],"format":{"duration":"60"}}`,
+			want: uploads.Probe{Duration: 60 * time.Second, HasAudio: true},
+		},
+		{
+			name: "no streams at all",
+			json: `{"streams":[],"format":{"duration":"10"}}`,
+			want: uploads.Probe{Duration: 10 * time.Second},
+		},
+		{
+			name: "duration absent means zero, which Validate rejects as unreadable",
+			json: `{"streams":[{"codec_type":"audio"}],"format":{}}`,
+			want: uploads.Probe{HasAudio: true},
+		},
+		{
+			name: "duration reported as N/A means zero",
+			json: `{"streams":[{"codec_type":"audio"}],"format":{"duration":"N/A"}}`,
+			want: uploads.Probe{HasAudio: true},
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got, err := audio.ParseProbe([]byte(tt.json))
+
+			require.NoError(t, err)
+			assert.Equal(t, tt.want, got)
+		})
+	}
+}
+
+func TestParseProbe_GarbageIsAnError(t *testing.T) {
+	_, err := audio.ParseProbe([]byte("not json"))
+	assert.Error(t, err)
+}
+
+// stub writes an executable shell script standing in for ffprobe. This is a fake
+// at the process boundary: the real binary is not on every host (it is not on
+// koala), and what is under test is how Probe drives and interprets a process.
+func stub(t *testing.T, body string) string {
+	t.Helper()
+	p := filepath.Join(t.TempDir(), "ffprobe")
+	require.NoError(t, os.WriteFile(p, []byte("#!/bin/sh\n"+body+"\n"), 0o755))
+	return p
+}
+
+func TestFFprobeProbe_ParsesWhatTheProcessPrints(t *testing.T) {
+	bin := stub(t, `echo '{"streams":[{"codec_type":"audio"}],"format":{"duration":"90"}}'`)
+
+	got, err := (&audio.FFprobe{Bin: bin}).Probe(context.Background(), "/ignored")
+
+	require.NoError(t, err)
+	assert.Equal(t, uploads.Probe{Duration: 90 * time.Second, HasAudio: true}, got)
+}
+
+func TestFFprobeProbe_PassesTheFilePathAsTheLastArgument(t *testing.T) {
+	out := filepath.Join(t.TempDir(), "args")
+	bin := stub(t, `for a; do last="$a"; done; printf '%s' "$last" > `+out+`
+echo '{"streams":[{"codec_type":"audio"}],"format":{"duration":"1"}}'`)
+
+	_, err := (&audio.FFprobe{Bin: bin}).Probe(context.Background(), "/spool/upload-123")
+
+	require.NoError(t, err)
+	got, err := os.ReadFile(out)
+	require.NoError(t, err)
+	assert.Equal(t, "/spool/upload-123", string(got))
+}
+
+func TestFFprobeProbe_NonzeroExitMeansTheFileIsUnreadable(t *testing.T) {
+	bin := stub(t, `echo "Invalid data found when processing input" >&2; exit 1`)
+
+	_, err := (&audio.FFprobe{Bin: bin}).Probe(context.Background(), "/x")
+
+	assert.ErrorIs(t, err, uploads.ErrUnreadable)
+}
+
+func TestFFprobeProbe_MissingBinaryIsOurFaultNotAnUnreadableFile(t *testing.T) {
+	_, err := (&audio.FFprobe{Bin: filepath.Join(t.TempDir(), "no-such-ffprobe")}).Probe(context.Background(), "/x")
+
+	require.Error(t, err)
+	assert.False(t, errors.Is(err, uploads.ErrUnreadable),
+		"a missing ffprobe must not be reported to the user as their file being bad")
+}
+
+func TestFFprobeProbe_StopsWhenTheContextIsCancelled(t *testing.T) {
+	bin := stub(t, `sleep 30`)
+	ctx, cancel := context.WithTimeout(context.Background(), 100*time.Millisecond)
+	defer cancel()
+
+	start := time.Now()
+	_, err := (&audio.FFprobe{Bin: bin}).Probe(ctx, "/x")
+
+	require.Error(t, err)
+	assert.Less(t, time.Since(start), 5*time.Second)
+	assert.False(t, errors.Is(err, uploads.ErrUnreadable), "a timeout is not a verdict on the file")
+}

internal/adapters/store/migrate_test.go

@@ -3,6 +3,7 @@ package store_test
 import (
 	"context"
 	"database/sql"
+	"errors"
 	"os"
 	"testing"
 
@@ -34,6 +35,30 @@ func fileMigrator(t *testing.T) *migrate.Migrate {
 	return m
 }
 
+// headVersion reports the current (HEAD) schema version so a test can restore
+// to it after stepping down, without hard-coding what HEAD is. Adding a
+// migration on top changes HEAD but no test that uses this needs editing.
+func headVersion(t *testing.T, m *migrate.Migrate) uint {
+	t.Helper()
+	v, dirty, err := m.Version()
+	require.NoError(t, err)
+	require.False(t, dirty, "schema must not be dirty")
+	return v
+}
+
+// migrateTo drives the schema to an exact version *by version number*, not by
+// step count. This is the whole point of the migrate-test design: a migration
+// added above the target does not shift any count here, so unrelated tests stay
+// green (see issue #8). ErrNoChange (already at that version) is not a failure.
+func migrateTo(t *testing.T, m *migrate.Migrate, version uint) {
+	t.Helper()
+	err := m.Migrate(version)
+	if errors.Is(err, migrate.ErrNoChange) {
+		return
+	}
+	require.NoError(t, err)
+}
+
 // loginEventsExists reports whether the login_events relation is present.
 func loginEventsExists(t *testing.T) bool {
 	t.Helper()
@@ -53,25 +78,15 @@ func TestMigration010LoginEventsUpDown(t *testing.T) {
 	require.True(t, loginEventsExists(t), "login_events must exist at latest migration")
 
 	m := fileMigrator(t)
-	// 011..016 sit above 010; step them down first so 010 is exercised in isolation.
+	head := headVersion(t, m)
-	require.NoError(t, m.Steps(-1), "down 016 drops channel_caption_state, login_events intact")
-	require.True(t, loginEventsExists(t), "016 down leaves login_events intact")
-	require.NoError(t, m.Steps(-1), "down 015 reshapes transcripts, login_events intact")
-	require.True(t, loginEventsExists(t), "015 down leaves login_events intact")
-	require.NoError(t, m.Steps(-1), "down 014 drops channel_title, login_events intact")
-	require.True(t, loginEventsExists(t), "014 down leaves login_events intact")
-	require.NoError(t, m.Steps(-1), "down 013 drops channel_errors, login_events intact")
-	require.True(t, loginEventsExists(t), "013 down leaves login_events intact")
-	require.NoError(t, m.Steps(-1), "down 012 is a no-op, login_events intact")
-	require.True(t, loginEventsExists(t), "012 down leaves login_events intact")
-	require.NoError(t, m.Steps(-1), "down 011 must not touch login_events")
-	require.True(t, loginEventsExists(t), "011 down leaves login_events intact")
 
-	require.NoError(t, m.Steps(-1), "down 010 must drop login_events")
+	migrateTo(t, m, 9) // just below 010 — everything above steps down
 	require.False(t, loginEventsExists(t), "login_events must be gone after the down migration")
 
-	require.NoError(t, m.Steps(7), "up must recreate 010 then re-apply 011..016")
+	migrateTo(t, m, 10) // up 010
 	require.True(t, loginEventsExists(t), "login_events must be restored after the up migration")
+
+	migrateTo(t, m, head) // restore to HEAD for sibling tests
 }
 
 // autoSummarizeDefault reads the users.auto_summarize column default as text
@@ -93,21 +108,15 @@ func TestMigration011AutoSummarizeDefaultUpDown(t *testing.T) {
 	require.Equal(t, "true", autoSummarizeDefault(t), "011 sets the default to TRUE")
 
 	m := fileMigrator(t)
-	require.NoError(t, m.Steps(-1), "down 016 drops channel_caption_state")
+	head := headVersion(t, m)
-	require.NoError(t, m.Steps(-1), "down 015 reshapes transcripts")
+
-	require.NoError(t, m.Steps(-1), "down 014 drops channel_title")
+	migrateTo(t, m, 10) // just below 011 — reverts the column default
-	require.NoError(t, m.Steps(-1), "down 013 drops channel_errors")
-	require.NoError(t, m.Steps(-1), "down 012 is a no-op")
-	require.NoError(t, m.Steps(-1), "down 011 reverts the column default")
 	require.Equal(t, "false", autoSummarizeDefault(t), "default is FALSE after the down migration")
 
-	require.NoError(t, m.Steps(1), "up 011 re-applies the TRUE default")
+	migrateTo(t, m, 11) // up 011 re-applies the TRUE default
 	require.Equal(t, "true", autoSummarizeDefault(t))
-	require.NoError(t, m.Steps(1), "up 012 runs clean (no FORCE RLS on fresh schema)")
+
-	require.NoError(t, m.Steps(1), "up 013 creates channel_errors")
+	migrateTo(t, m, head) // restore to HEAD for sibling tests
-	require.NoError(t, m.Steps(1), "up 014 recreates channel_title")
-	require.NoError(t, m.Steps(1), "up 015 reshapes transcripts to shared")
-	require.NoError(t, m.Steps(1), "up 016 recreates channel_caption_state (HEAD)")
 }
 
 // channelTitleExists reports whether videos.channel_title is present.
@@ -127,17 +136,15 @@ func TestMigration014VideoChannelTitleUpDown(t *testing.T) {
 	require.True(t, channelTitleExists(t), "channel_title exists at latest migration")
 
 	m := fileMigrator(t)
-	require.NoError(t, m.Steps(-1), "down 016 drops channel_caption_state, channel_title intact")
+	head := headVersion(t, m)
-	require.True(t, channelTitleExists(t), "016 down leaves channel_title intact")
+
-	require.NoError(t, m.Steps(-1), "down 015 reshapes transcripts, channel_title intact")
+	migrateTo(t, m, 13) // just below 014 — drops channel_title
-	require.True(t, channelTitleExists(t), "015 down leaves channel_title intact")
-	require.NoError(t, m.Steps(-1), "down 014 must drop channel_title")
 	require.False(t, channelTitleExists(t), "channel_title must be gone after the down migration")
 
-	require.NoError(t, m.Steps(1), "up 014 must recreate channel_title")
+	migrateTo(t, m, 14) // up 014 recreates channel_title
 	require.True(t, channelTitleExists(t), "channel_title must be restored after the up migration")
-	require.NoError(t, m.Steps(1), "up 015 restores the shared transcripts shape")
+
-	require.NoError(t, m.Steps(1), "up 016 recreates channel_caption_state (HEAD)")
+	migrateTo(t, m, head) // restore to HEAD for sibling tests
 }
 
 // TestMigration012FixAutoSummarizeRLS proves 012 runs cleanly and flips any
@@ -148,7 +155,11 @@ func TestMigration012FixAutoSummarizeRLS(t *testing.T) {
 
 	// Round-trip: down 012, then up 012 — must be idempotent.
 	m := fileMigrator(t)
-	require.NoError(t, m.Steps(-1), "down 012 must not error")
+	head := headVersion(t, m)
-	require.NoError(t, m.Steps(1), "up 012 must re-apply cleanly")
+
+	migrateTo(t, m, 11) // down 012 must not error
+	migrateTo(t, m, 12) // up 012 must re-apply cleanly
 	require.Equal(t, "true", autoSummarizeDefault(t), "default still TRUE after 012 re-applied")
+
+	migrateTo(t, m, head) // restore to HEAD for sibling tests
 }

internal/adapters/store/migrations/017_uploaded_transcripts.down.sql

@@ -0,0 +1,3 @@
+-- Drops the private uploaded-transcript store and every transcript in it. There
+-- is no revert of the data: the original audio was never kept (ADR-033).
+DROP TABLE uploaded_transcripts;

internal/adapters/store/migrations/017_uploaded_transcripts.up.sql

@@ -0,0 +1,48 @@
+-- Migration 017: uploaded_transcripts — PRIVATE transcripts of audio a user
+-- uploaded themselves (ADR-033, tapir#33).
+--
+-- This is deliberately NOT the shared `transcripts` table (migration 015,
+-- ADR-021). That table is non-RLS on the explicit grounds that it holds only
+-- public caption content and nothing user-identifying. A recorded meeting is the
+-- opposite: private third-party content owned by exactly one user. Reusing
+-- `transcripts` — or adding a user_id to it with a partial policy — would put a
+-- private row in the one table the isolation test asserts is shared, and would
+-- pass CI green. A separate table keeps ADR-021's exception singular.
+--
+-- Row-level security follows migration 003 exactly: ENABLE + FORCE (the app
+-- connects as the table owner, which bypasses RLS unless forced), and a policy
+-- keyed off the per-request GUC tapir.current_user_id. With no GUC set the
+-- predicate is NULL and the table is deny-all.
+--
+-- The original audio is never stored (ADR-033): this row is all that is kept.
+-- segments holds the timestamped cues; re-transcribing 90 minutes costs ~7
+-- minutes, so discarding the timestamps would be irreversible.
+CREATE TABLE uploaded_transcripts (
+    id            UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    user_id       UUID NOT NULL REFERENCES users(id) ON DELETE CASCADE,
+    title         TEXT NOT NULL,
+    status        TEXT NOT NULL DEFAULT 'queued'
+                  CHECK (status IN ('queued', 'transcribing', 'done', 'failed')),
+    content       TEXT,
+    segments      JSONB,
+    endpoint      TEXT,             -- which backend served it (iguana | berget); a quality fact
+    audio_seconds DOUBLE PRECISION,
+    error_kind    TEXT,             -- typed, never a provider's error string (ADR-033)
+    created_at    TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    finished_at   TIMESTAMPTZ,
+    CONSTRAINT uploaded_transcripts_done_has_content
+        CHECK (status <> 'done' OR content IS NOT NULL),
+    CONSTRAINT uploaded_transcripts_failed_has_error_kind
+        CHECK (status <> 'failed' OR error_kind IS NOT NULL)
+);
+
+CREATE INDEX idx_uploaded_transcripts_user ON uploaded_transcripts (user_id, created_at DESC);
+
+ALTER TABLE uploaded_transcripts ENABLE ROW LEVEL SECURITY;
+ALTER TABLE uploaded_transcripts FORCE ROW LEVEL SECURITY;
+CREATE POLICY uploaded_transcripts_isolation ON uploaded_transcripts
+    FOR ALL
+    USING (user_id = current_setting('tapir.current_user_id', true)::uuid);
+
+COMMENT ON TABLE uploaded_transcripts IS
+    'PRIVATE per-user transcripts of user-supplied audio. RLS-scoped (ADR-012). Never merge into the shared transcripts table (ADR-021, ADR-033).';

internal/adapters/store/uploaded_transcripts.go

@@ -0,0 +1,199 @@
+package store
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"time"
+
+	"github.com/jackc/pgx/v5"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// ErrUploadNotFound is returned when an uploaded transcript does not exist OR is
+// not the caller's. The two are deliberately indistinguishable: a non-owner learns
+// nothing about whether someone else's upload exists (ADR-033).
+var ErrUploadNotFound = errors.New("store: uploaded transcript not found")
+
+// Uploaded transcripts are PRIVATE per-user content, stored in their own
+// RLS-scoped table (migration 017) — never in the shared `transcripts` (ADR-021).
+// Every method goes through withUser, so row-level security confines it to the
+// caller's rows; each statement ALSO names user_id in its predicate, so the
+// isolation does not rest on a single mechanism and stays testable under the
+// superuser connection the unit tests use.
+
+// CreateUpload records a new upload in the queued state and returns its id.
+func (s *Store) CreateUpload(ctx context.Context, userID, title string) (string, error) {
+	var id string
+	if err := s.withUser(ctx, userID, func(tx pgx.Tx) error {
+		return tx.QueryRow(ctx,
+			`INSERT INTO uploaded_transcripts (user_id, title) VALUES ($1, $2) RETURNING id`,
+			userID, title).Scan(&id)
+	}); err != nil {
+		return "", fmt.Errorf("store: create upload: %w", err)
+	}
+	return id, nil
+}
+
+// StartTranscribing moves an upload to transcribing.
+func (s *Store) StartTranscribing(ctx context.Context, userID, id string) error {
+	return s.updateUpload(ctx, userID, id, "start transcribing",
+		`UPDATE uploaded_transcripts SET status = 'transcribing'
+		  WHERE id = $1 AND user_id = $2`)
+}
+
+// CompleteUpload stores the finished transcription and moves the upload to done.
+func (s *Store) CompleteUpload(ctx context.Context, userID, id string, t domain.Transcription) error {
+	segs, err := json.Marshal(t.Segments)
+	if err != nil {
+		return fmt.Errorf("store: complete upload: marshal segments: %w", err)
+	}
+	return s.updateUpload(ctx, userID, id, "complete upload",
+		`UPDATE uploaded_transcripts
+		    SET status = 'done', content = $3, segments = $4, endpoint = $5,
+		        audio_seconds = $6, error_kind = NULL, finished_at = NOW()
+		  WHERE id = $1 AND user_id = $2`,
+		t.Text, segs, t.Endpoint, t.AudioSeconds)
+}
+
+// FailUpload moves an upload to failed with a typed reason.
+func (s *Store) FailUpload(ctx context.Context, userID, id string, kind domain.UploadErrorKind) error {
+	return s.updateUpload(ctx, userID, id, "fail upload",
+		`UPDATE uploaded_transcripts
+		    SET status = 'failed', error_kind = $3, finished_at = NOW()
+		  WHERE id = $1 AND user_id = $2`,
+		string(kind))
+}
+
+// updateUpload runs one UPDATE keyed on ($1 = id, $2 = userID) plus extra args,
+// and maps "no row touched" to ErrUploadNotFound — which is what a write to another
+// user's upload looks like under RLS (the row is invisible, not forbidden).
+func (s *Store) updateUpload(ctx context.Context, userID, id, op, sql string, args ...any) error {
+	all := append([]any{id, userID}, args...)
+	return s.withUser(ctx, userID, func(tx pgx.Tx) error {
+		ct, err := tx.Exec(ctx, sql, all...)
+		if err != nil {
+			return fmt.Errorf("store: %s: %w", op, err)
+		}
+		if ct.RowsAffected() == 0 {
+			return ErrUploadNotFound
+		}
+		return nil
+	})
+}
+
+// GetUpload returns one of the caller's uploads, including the transcript text.
+func (s *Store) GetUpload(ctx context.Context, userID, id string) (*domain.UploadedTranscript, error) {
+	var u domain.UploadedTranscript
+	err := s.withUser(ctx, userID, func(tx pgx.Tx) error {
+		row := tx.QueryRow(ctx, uploadSelect+` WHERE id = $1 AND user_id = $2`, id, userID)
+		return scanUpload(row, &u, true)
+	})
+	if errors.Is(err, pgx.ErrNoRows) {
+		return nil, ErrUploadNotFound
+	}
+	if err != nil {
+		return nil, fmt.Errorf("store: get upload: %w", err)
+	}
+	return &u, nil
+}
+
+// ListUploads returns the caller's uploads, newest first, WITHOUT transcript text
+// or segments: a list view must not drag every transcript along.
+func (s *Store) ListUploads(ctx context.Context, userID string) ([]domain.UploadedTranscript, error) {
+	var out []domain.UploadedTranscript
+	err := s.withUser(ctx, userID, func(tx pgx.Tx) error {
+		rows, err := tx.Query(ctx, uploadSelect+` WHERE user_id = $1 ORDER BY created_at DESC, id`, userID)
+		if err != nil {
+			return err
+		}
+		defer rows.Close()
+		for rows.Next() {
+			var u domain.UploadedTranscript
+			if err := scanUpload(rows, &u, false); err != nil {
+				return err
+			}
+			out = append(out, u)
+		}
+		return rows.Err()
+	})
+	if err != nil {
+		return nil, fmt.Errorf("store: list uploads: %w", err)
+	}
+	return out, nil
+}
+
+// FailInterruptedUploads is the start-up sweep. The audio of an upload is never
+// retained (ADR-033), so a job that was queued or transcribing when the process
+// stopped can never resume; leaving it would be a stuck spinner. Every such row,
+// for every user, becomes failed/interrupted and the user re-uploads. It
+// enumerates users from the un-RLS'd identity map and scopes each pass with
+// withUser, exactly as scheduled discovery does (ADR-018). Returns rows failed.
+func (s *Store) FailInterruptedUploads(ctx context.Context) (int, error) {
+	users, err := s.ListAllUsers(ctx)
+	if err != nil {
+		return 0, fmt.Errorf("store: fail interrupted uploads: %w", err)
+	}
+	total := 0
+	for _, u := range users {
+		err := s.withUser(ctx, u.UserID, func(tx pgx.Tx) error {
+			ct, err := tx.Exec(ctx,
+				`UPDATE uploaded_transcripts
+				    SET status = 'failed', error_kind = $2, finished_at = NOW()
+				  WHERE user_id = $1 AND status IN ('queued', 'transcribing')`,
+				u.UserID, string(domain.UploadErrInterrupted))
+			total += int(ct.RowsAffected())
+			return err
+		})
+		if err != nil {
+			return total, fmt.Errorf("store: fail interrupted uploads: user %s: %w", u.UserID, err)
+		}
+	}
+	return total, nil
+}
+
+const uploadSelect = `SELECT id, user_id, title, status, content, segments, endpoint,
+       audio_seconds, error_kind, created_at, finished_at FROM uploaded_transcripts`
+
+// scanUpload scans one uploadSelect row. withText controls whether the transcript
+// body and segments are kept (GetUpload) or dropped (ListUploads).
+func scanUpload(row pgx.Row, u *domain.UploadedTranscript, withText bool) error {
+	var (
+		status, errKind, endpoint, content *string
+		segments                           []byte
+		seconds                            *float64
+		finished                           *time.Time
+	)
+	if err := row.Scan(&u.ID, &u.UserID, &u.Title, &status, &content, &segments, &endpoint,
+		&seconds, &errKind, &u.CreatedAt, &finished); err != nil {
+		return err
+	}
+	u.Status = domain.UploadStatus(deref(status))
+	u.Endpoint = deref(endpoint)
+	u.ErrorKind = domain.UploadErrorKind(deref(errKind))
+	if seconds != nil {
+		u.AudioSeconds = *seconds
+	}
+	if finished != nil {
+		u.FinishedAt = *finished
+	}
+	if !withText {
+		return nil
+	}
+	u.Content = deref(content)
+	if len(segments) > 0 {
+		if err := json.Unmarshal(segments, &u.Segments); err != nil {
+			return fmt.Errorf("unmarshal segments: %w", err)
+		}
+	}
+	return nil
+}
+
+func deref(s *string) string {
+	if s == nil {
+		return ""
+	}
+	return *s
+}

internal/adapters/store/uploaded_transcripts_isolation_test.go

@@ -0,0 +1,196 @@
+package store_test
+
+import (
+	"context"
+	"sort"
+	"testing"
+
+	"github.com/jackc/pgx/v5/pgxpool"
+	"github.com/stretchr/testify/require"
+)
+
+// Isolation proof for ADR-033: an uploaded transcript is PRIVATE content owned by
+// one user, so it lives in its own RLS-scoped table — never in the shared,
+// non-RLS `transcripts` (ADR-021). These tests drive the database directly (raw
+// SQL as the non-superuser `app` role) so they observe row-level security itself,
+// not any application-layer WHERE clause. The superuser pool is used only to seed
+// and to read pg_catalog; it bypasses RLS and so must never be the pool whose
+// answers the assertions rest on.
+
+// seedUploadedTranscript inserts one finished upload for userID as the superuser
+// (RLS bypassed so both users' rows land regardless of any GUC) and returns its id.
+// The text is synthetic — this repo is public, and a real transcript must never
+// appear in it (#33).
+func seedUploadedTranscript(t *testing.T, super *pgxpool.Pool, userID, text string) string {
+	t.Helper()
+	ctx := context.Background()
+	_, err := super.Exec(ctx, `INSERT INTO users (id) VALUES ($1) ON CONFLICT DO NOTHING`, userID)
+	require.NoError(t, err)
+	var id string
+	require.NoError(t, super.QueryRow(ctx,
+		`INSERT INTO uploaded_transcripts (user_id, title, status, content)
+		 VALUES ($1, 'synthetic fixture', 'done', $2) RETURNING id`,
+		userID, text).Scan(&id))
+	return id
+}
+
+func TestUploadedTranscriptsRLS_ScopedReadsSeeOnlyOwnRows(t *testing.T) {
+	newStore(t)
+	super := rawPool(t)
+	resetDB(t, super)
+	seedUploadedTranscript(t, super, userA, "text owned by A")
+	seedUploadedTranscript(t, super, userB, "text owned by B")
+	app := appPool(t, super)
+
+	require.Equal(t, 0, scopedCount(t, app, "", "uploaded_transcripts"),
+		"no tapir.current_user_id must be deny-all: this proves RLS is live, not bypassed")
+	require.Equal(t, 1, scopedCount(t, app, userA, "uploaded_transcripts"),
+		"user A must see exactly its own upload")
+	require.Equal(t, 1, scopedCount(t, app, userB, "uploaded_transcripts"),
+		"user B must see exactly its own upload")
+
+	// Count alone cannot distinguish "A's row" from "B's row" — read the content.
+	var got string
+	ctx := context.Background()
+	tx, err := app.Begin(ctx)
+	require.NoError(t, err)
+	defer tx.Rollback(ctx) //nolint:errcheck
+	_, err = tx.Exec(ctx, `SELECT set_config('tapir.current_user_id', $1, true)`, userA)
+	require.NoError(t, err)
+	require.NoError(t, tx.QueryRow(ctx, `SELECT content FROM uploaded_transcripts`).Scan(&got))
+	require.Equal(t, "text owned by A", got, "A's scoped read must return A's text, never B's")
+}
+
+func TestUploadedTranscriptsRLS_CrossUserWritesAffectZeroRows(t *testing.T) {
+	newStore(t)
+	super := rawPool(t)
+	resetDB(t, super)
+	seedUploadedTranscript(t, super, userA, "text owned by A")
+	bID := seedUploadedTranscript(t, super, userB, "text owned by B")
+	app := appPool(t, super)
+
+	require.Equal(t, int64(0), scopedRowsAffected(t, app, userA,
+		`UPDATE uploaded_transcripts SET content = 'hacked' WHERE id = $1`, bID),
+		"user A scoped UPDATE must touch zero of B's uploads")
+	require.Equal(t, int64(0), scopedRowsAffected(t, app, userA,
+		`DELETE FROM uploaded_transcripts WHERE id = $1`, bID),
+		"user A scoped DELETE must touch zero of B's uploads")
+
+	var content string
+	require.NoError(t, super.QueryRow(context.Background(),
+		`SELECT content FROM uploaded_transcripts WHERE id = $1`, bID).Scan(&content))
+	require.Equal(t, "text owned by B", content, "B's upload must survive A's writes unchanged")
+}
+
+func TestUploadedTranscriptsRLS_InsertForAnotherUserIsRejected(t *testing.T) {
+	newStore(t)
+	super := rawPool(t)
+	resetDB(t, super)
+	seedUploadedTranscript(t, super, userA, "text owned by A")
+	_, err := super.Exec(context.Background(), `INSERT INTO users (id) VALUES ($1)`, userB)
+	require.NoError(t, err)
+	app := appPool(t, super)
+
+	ctx := context.Background()
+	tx, err := app.Begin(ctx)
+	require.NoError(t, err)
+	defer tx.Rollback(ctx) //nolint:errcheck
+	_, err = tx.Exec(ctx, `SELECT set_config('tapir.current_user_id', $1, true)`, userA)
+	require.NoError(t, err)
+
+	// Scoped as A, plant a row that claims to belong to B. The policy's WITH CHECK
+	// (defaulting to USING under FOR ALL) must refuse it.
+	_, err = tx.Exec(ctx,
+		`INSERT INTO uploaded_transcripts (user_id, title, status, content)
+		 VALUES ($1, 'planted', 'done', 'planted')`, userB)
+	require.Error(t, err, "a row scoped as A must not be insertable on behalf of B")
+	require.Contains(t, err.Error(), "row-level security")
+}
+
+func TestUploadedTranscripts_RowSecurityIsEnabledForcedAndKeyedByUser(t *testing.T) {
+	newStore(t)
+	super := rawPool(t)
+	ctx := context.Background()
+
+	var enabled, forced bool
+	require.NoError(t, super.QueryRow(ctx,
+		`SELECT relrowsecurity, relforcerowsecurity FROM pg_class
+		  WHERE oid = 'public.uploaded_transcripts'::regclass`).Scan(&enabled, &forced))
+	require.True(t, enabled, "uploaded_transcripts must have row-level security enabled")
+	require.True(t, forced,
+		"FORCE is what binds the table-owner role tapir; without it the policy is dead in production")
+
+	var userIDNotNull bool
+	require.NoError(t, super.QueryRow(ctx,
+		`SELECT is_nullable = 'NO' FROM information_schema.columns
+		  WHERE table_name = 'uploaded_transcripts' AND column_name = 'user_id'`).Scan(&userIDNotNull))
+	require.True(t, userIDNotNull, "an upload with no owner would match no policy and be unreachable")
+}
+
+func TestUploadedTranscripts_DeleteUserRemovesTheirUploads(t *testing.T) {
+	s := newStore(t)
+	super := rawPool(t)
+	resetDB(t, super)
+	seedUploadedTranscript(t, super, userA, "text owned by A")
+	bID := seedUploadedTranscript(t, super, userB, "text owned by B")
+
+	require.NoError(t, s.DeleteUser(context.Background(), userA))
+
+	var aRows, bRows int
+	ctx := context.Background()
+	require.NoError(t, super.QueryRow(ctx,
+		`SELECT count(*) FROM uploaded_transcripts WHERE user_id = $1`, userA).Scan(&aRows))
+	require.NoError(t, super.QueryRow(ctx,
+		`SELECT count(*) FROM uploaded_transcripts WHERE id = $1`, bID).Scan(&bRows))
+	require.Equal(t, 0, aRows, "deleting a user must remove their private transcripts (ON DELETE CASCADE)")
+	require.Equal(t, 1, bRows, "deleting A must not touch B's transcripts")
+}
+
+// rowSecurityOffTables are the public-schema tables that deliberately have row
+// security OFF. The list is exact on purpose: a NEW table added without RLS must
+// fail this test and force a decision, instead of passing CI by omission — the
+// failure mode ADR-033 names, where a naive reuse "leaks private content across
+// tenants and passes CI green".
+//
+//   - transcripts: shared public caption content, the single exception for CONTENT
+//     (ADR-021). It is the one that must never gain a private row.
+//   - user_identities: the Dex-subject -> user_id map, read at login before any
+//     user is known (migration 004). Holds no user content.
+//   - invitations: exists before the user does, so there is no user to scope by;
+//     the single-use token is the capability (migration 009).
+//   - schema_migrations: golang-migrate bookkeeping.
+//
+// #33 describes `transcripts` as "the only non-RLS table". That is true of content
+// and false of tables; this test asserts the real set.
+var rowSecurityOffTables = []string{
+	"invitations",
+	"schema_migrations",
+	"transcripts",
+	"user_identities",
+}
+
+func TestRowSecurityOffTablesAreExactlyTheKnownSet(t *testing.T) {
+	newStore(t)
+	super := rawPool(t)
+
+	rows, err := super.Query(context.Background(),
+		`SELECT c.relname FROM pg_class c
+		   JOIN pg_namespace n ON n.oid = c.relnamespace
+		  WHERE n.nspname = 'public' AND c.relkind IN ('r', 'p') AND NOT c.relrowsecurity
+		  ORDER BY c.relname`)
+	require.NoError(t, err)
+	defer rows.Close()
+	var got []string
+	for rows.Next() {
+		var name string
+		require.NoError(t, rows.Scan(&name))
+		got = append(got, name)
+	}
+	require.NoError(t, rows.Err())
+
+	want := append([]string{}, rowSecurityOffTables...)
+	sort.Strings(want)
+	require.Equal(t, want, got,
+		"every public table must have row security ON unless it is in rowSecurityOffTables; "+
+			"adding a table here is a decision that needs an ADR")
+}

internal/adapters/store/uploaded_transcripts_test.go

@@ -0,0 +1,188 @@
+package store_test
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/store"
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// These exercise the Store API over uploaded_transcripts. The test Store connects
+// as the superuser, which BYPASSES row-level security — so the cross-user cases
+// here prove the query's own user_id predicate (defence in depth), while the
+// database-enforced guarantee is proven separately, as the non-superuser app
+// role, in uploaded_transcripts_isolation_test.go. Text is synthetic; this repo is
+// public and no real transcript may appear in it (#33).
+
+func seedUsers(t *testing.T, ids ...string) {
+	t.Helper()
+	super := rawPool(t)
+	resetDB(t, super)
+	for _, id := range ids {
+		_, err := super.Exec(context.Background(), `INSERT INTO users (id) VALUES ($1)`, id)
+		require.NoError(t, err)
+	}
+}
+
+func TestCreateUpload_StartsQueuedAndIsOwnedByTheCreator(t *testing.T) {
+	s := newStore(t)
+	seedUsers(t, userA)
+	ctx := context.Background()
+
+	id, err := s.CreateUpload(ctx, userA, "Teams export")
+	require.NoError(t, err)
+
+	got, err := s.GetUpload(ctx, userA, id)
+	require.NoError(t, err)
+	assert.Equal(t, id, got.ID)
+	assert.Equal(t, userA, got.UserID)
+	assert.Equal(t, "Teams export", got.Title)
+	assert.Equal(t, domain.UploadQueued, got.Status)
+	assert.Empty(t, got.Content)
+	assert.True(t, got.FinishedAt.IsZero(), "an unfinished upload has no finish time")
+}
+
+func TestUploadLifecycle_CompletedKeepsTextSegmentsEndpointAndDuration(t *testing.T) {
+	s := newStore(t)
+	seedUsers(t, userA)
+	ctx := context.Background()
+	id, err := s.CreateUpload(ctx, userA, "m")
+	require.NoError(t, err)
+
+	require.NoError(t, s.StartTranscribing(ctx, userA, id))
+	mid, err := s.GetUpload(ctx, userA, id)
+	require.NoError(t, err)
+	assert.Equal(t, domain.UploadTranscribing, mid.Status)
+
+	want := domain.Transcription{
+		Text: "God morgon allihopa. Vi går igenom läget.",
+		Segments: []domain.Segment{
+			{Start: 0.0, End: 1.5, Text: "God morgon allihopa."},
+			{Start: 1.5, End: 4.25, Text: "Vi går igenom läget."},
+		},
+		Endpoint:     "iguana/kb-whisper",
+		AudioSeconds: 4.25,
+	}
+	require.NoError(t, s.CompleteUpload(ctx, userA, id, want))
+
+	got, err := s.GetUpload(ctx, userA, id)
+	require.NoError(t, err)
+	assert.Equal(t, domain.UploadDone, got.Status)
+	assert.Equal(t, want.Text, got.Content)
+	assert.Equal(t, want.Segments, got.Segments, "segment times must survive the JSONB round trip")
+	assert.Equal(t, "iguana/kb-whisper", got.Endpoint, "the serving endpoint is a quality fact and is recorded per row")
+	assert.InDelta(t, 4.25, got.AudioSeconds, 1e-9)
+	assert.False(t, got.FinishedAt.IsZero())
+	assert.Empty(t, got.ErrorKind)
+}
+
+func TestUploadLifecycle_FailedRecordsTypedKindAndNoContent(t *testing.T) {
+	s := newStore(t)
+	seedUsers(t, userA)
+	ctx := context.Background()
+	id, err := s.CreateUpload(ctx, userA, "m")
+	require.NoError(t, err)
+	require.NoError(t, s.StartTranscribing(ctx, userA, id))
+
+	require.NoError(t, s.FailUpload(ctx, userA, id, domain.UploadErrTranscriptionFailed))
+
+	got, err := s.GetUpload(ctx, userA, id)
+	require.NoError(t, err)
+	assert.Equal(t, domain.UploadFailed, got.Status)
+	assert.Equal(t, domain.UploadErrTranscriptionFailed, got.ErrorKind)
+	assert.Empty(t, got.Content)
+	assert.False(t, got.FinishedAt.IsZero())
+}
+
+func TestGetUpload_AnotherUsersIDIsNotFound(t *testing.T) {
+	s := newStore(t)
+	seedUsers(t, userA, userB)
+	ctx := context.Background()
+	id, err := s.CreateUpload(ctx, userA, "A's meeting")
+	require.NoError(t, err)
+
+	_, err = s.GetUpload(ctx, userB, id)
+
+	assert.ErrorIs(t, err, store.ErrUploadNotFound,
+		"a non-owner must get not-found, not forbidden: existence is not disclosed")
+}
+
+func TestUploadWrites_ByAnotherUserChangeNothing(t *testing.T) {
+	s := newStore(t)
+	seedUsers(t, userA, userB)
+	ctx := context.Background()
+	id, err := s.CreateUpload(ctx, userA, "A's meeting")
+	require.NoError(t, err)
+
+	assert.ErrorIs(t, s.StartTranscribing(ctx, userB, id), store.ErrUploadNotFound)
+	assert.ErrorIs(t, s.CompleteUpload(ctx, userB, id, domain.Transcription{Text: "planted"}), store.ErrUploadNotFound)
+	assert.ErrorIs(t, s.FailUpload(ctx, userB, id, domain.UploadErrTranscriptionFailed), store.ErrUploadNotFound)
+
+	got, err := s.GetUpload(ctx, userA, id)
+	require.NoError(t, err)
+	assert.Equal(t, domain.UploadQueued, got.Status, "B's writes must leave A's upload exactly as it was")
+	assert.Empty(t, got.Content)
+}
+
+func TestListUploads_ReturnsOnlyTheCallersNewestFirstWithoutTranscriptText(t *testing.T) {
+	s := newStore(t)
+	seedUsers(t, userA, userB)
+	ctx := context.Background()
+	first, err := s.CreateUpload(ctx, userA, "first")
+	require.NoError(t, err)
+	second, err := s.CreateUpload(ctx, userA, "second")
+	require.NoError(t, err)
+	_, err = s.CreateUpload(ctx, userB, "B's")
+	require.NoError(t, err)
+	require.NoError(t, s.StartTranscribing(ctx, userA, first))
+	require.NoError(t, s.CompleteUpload(ctx, userA, first, domain.Transcription{Text: "long text body", Endpoint: "e"}))
+
+	got, err := s.ListUploads(ctx, userA)
+
+	require.NoError(t, err)
+	require.Len(t, got, 2, "A must see only A's uploads")
+	assert.Equal(t, second, got[0].ID, "newest first")
+	assert.Equal(t, first, got[1].ID)
+	assert.Empty(t, got[1].Content, "a list row must not drag the whole transcript along")
+	assert.Equal(t, domain.UploadDone, got[1].Status)
+}
+
+func TestFailInterruptedUploads_FailsOnlyUnfinishedRowsForEveryUser(t *testing.T) {
+	s := newStore(t)
+	seedUsers(t, userA, userB)
+	super := rawPool(t)
+	ctx := context.Background()
+	for subject, uid := range map[string]string{"sub-a": userA, "sub-b": userB} {
+		_, err := super.Exec(ctx, `INSERT INTO user_identities (dex_subject, user_id) VALUES ($1, $2)`, subject, uid)
+		require.NoError(t, err)
+	}
+	queued, err := s.CreateUpload(ctx, userA, "queued")
+	require.NoError(t, err)
+	running, err := s.CreateUpload(ctx, userB, "running")
+	require.NoError(t, err)
+	require.NoError(t, s.StartTranscribing(ctx, userB, running))
+	done, err := s.CreateUpload(ctx, userA, "done")
+	require.NoError(t, err)
+	require.NoError(t, s.StartTranscribing(ctx, userA, done))
+	require.NoError(t, s.CompleteUpload(ctx, userA, done, domain.Transcription{Text: "kept", Endpoint: "e"}))
+
+	n, err := s.FailInterruptedUploads(ctx)
+
+	require.NoError(t, err)
+	assert.Equal(t, 2, n, "the queued and the transcribing row, across both users")
+	for _, c := range []struct{ user, id string }{{userA, queued}, {userB, running}} {
+		got, err := s.GetUpload(ctx, c.user, c.id)
+		require.NoError(t, err)
+		assert.Equal(t, domain.UploadFailed, got.Status)
+		assert.Equal(t, domain.UploadErrInterrupted, got.ErrorKind,
+			"audio is not retained, so a restart cannot resume: the honest state is failed/interrupted")
+	}
+	kept, err := s.GetUpload(ctx, userA, done)
+	require.NoError(t, err)
+	assert.Equal(t, domain.UploadDone, kept.Status, "a finished transcript must survive the sweep")
+	assert.Equal(t, "kept", kept.Content)
+}

internal/adapters/transcribe/chain.go

@@ -0,0 +1,53 @@
+package transcribe
+
+import (
+	"context"
+	"errors"
+	"fmt"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// Backend is one transcription endpoint. *Gateway satisfies it; Name is the value
+// recorded as the serving endpoint.
+type Backend interface {
+	Name() string
+	Transcribe(ctx context.Context, wavPath string) (domain.Transcription, error)
+}
+
+// Chain tries a primary (local) backend and, if it fails for ANY reason, a
+// fallback. The fallback exists because iguana can wedge permanently on a Metal
+// OOM (infra#273) — which is also why a fallback is a data-egress decision: the
+// berget endpoint is a cloud service, so the result records which one served it.
+type Chain struct {
+	primary  Backend
+	fallback Backend // nil disables the fallback
+}
+
+// NewChain builds a Chain. A nil fallback means a primary failure is the result.
+func NewChain(primary, fallback Backend) *Chain {
+	return &Chain{primary: primary, fallback: fallback}
+}
+
+// Transcribe returns the first backend's success. It falls back on any primary
+// error — transport, timeout, non-2xx, empty transcript — and never inspects a
+// response body to decide. A cancelled context stops the chain: a job the user or
+// shutdown cancelled must not ship its audio to a cloud endpoint as a last act.
+func (c *Chain) Transcribe(ctx context.Context, wavPath string) (domain.Transcription, error) {
+	t, err := c.primary.Transcribe(ctx, wavPath)
+	if err == nil {
+		t.Endpoint = c.primary.Name()
+		return t, nil
+	}
+	if c.fallback == nil || ctx.Err() != nil {
+		return domain.Transcription{}, fmt.Errorf("%s: %w", c.primary.Name(), err)
+	}
+	t, ferr := c.fallback.Transcribe(ctx, wavPath)
+	if ferr != nil {
+		return domain.Transcription{}, errors.Join(
+			fmt.Errorf("%s: %w", c.primary.Name(), err),
+			fmt.Errorf("%s: %w", c.fallback.Name(), ferr))
+	}
+	t.Endpoint = c.fallback.Name()
+	return t, nil
+}

internal/adapters/transcribe/chain_test.go

@@ -0,0 +1,99 @@
+package transcribe_test
+
+import (
+	"context"
+	"errors"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/transcribe"
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+type fakeBackend struct {
+	name  string
+	text  string
+	err   error
+	calls int
+}
+
+func (f *fakeBackend) Name() string { return f.name }
+func (f *fakeBackend) Transcribe(_ context.Context, _ string) (domain.Transcription, error) {
+	f.calls++
+	if f.err != nil {
+		return domain.Transcription{}, f.err
+	}
+	return domain.Transcription{Text: f.text, Endpoint: f.name}, nil
+}
+
+func TestChain_PrimarySuccessNeverTouchesTheFallback(t *testing.T) {
+	primary := &fakeBackend{name: "iguana/kb-whisper", text: "primary text"}
+	fallback := &fakeBackend{name: "berget/kb-whisper", text: "fallback text"}
+
+	got, err := transcribe.NewChain(primary, fallback).Transcribe(context.Background(), "/a.wav")
+
+	require.NoError(t, err)
+	assert.Equal(t, "primary text", got.Text)
+	assert.Equal(t, "iguana/kb-whisper", got.Endpoint)
+	assert.Zero(t, fallback.calls, "audio must not leave the homelab when the local endpoint worked")
+}
+
+func TestChain_AnyPrimaryFailureFallsBackAndRecordsTheFallbackEndpoint(t *testing.T) {
+	tests := []struct {
+		name string
+		err  error
+	}{
+		{"server error", &transcribe.StatusError{Code: 500}},
+		{"gateway timeout", &transcribe.StatusError{Code: 504}},
+		{"client error", &transcribe.StatusError{Code: 400}},
+		{"transport failure", errors.New("dial tcp: connection refused")},
+		{"empty transcript", transcribe.ErrEmptyTranscript},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			primary := &fakeBackend{name: "iguana/kb-whisper", err: tt.err}
+			fallback := &fakeBackend{name: "berget/kb-whisper", text: "fallback text"}
+
+			got, err := transcribe.NewChain(primary, fallback).Transcribe(context.Background(), "/a.wav")
+
+			require.NoError(t, err)
+			assert.Equal(t, "fallback text", got.Text)
+			assert.Equal(t, "berget/kb-whisper", got.Endpoint,
+				"the row must say berget served it: that means the audio left the homelab")
+		})
+	}
+}
+
+func TestChain_BothFailingReportsBothErrors(t *testing.T) {
+	primary := &fakeBackend{name: "iguana/kb-whisper", err: errors.New("iguana down")}
+	fallback := &fakeBackend{name: "berget/kb-whisper", err: errors.New("berget down")}
+
+	_, err := transcribe.NewChain(primary, fallback).Transcribe(context.Background(), "/a.wav")
+
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "iguana down")
+	assert.Contains(t, err.Error(), "berget down")
+}
+
+func TestChain_WithoutAFallbackAPrimaryFailureIsTheResult(t *testing.T) {
+	primary := &fakeBackend{name: "iguana/kb-whisper", err: errors.New("iguana down")}
+
+	_, err := transcribe.NewChain(primary, nil).Transcribe(context.Background(), "/a.wav")
+
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "iguana down")
+}
+
+func TestChain_ACancelledJobDoesNotFallBack(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	primary := &fakeBackend{name: "iguana/kb-whisper", err: context.Canceled}
+	fallback := &fakeBackend{name: "berget/kb-whisper", text: "x"}
+
+	_, err := transcribe.NewChain(primary, fallback).Transcribe(ctx, "/a.wav")
+
+	assert.ErrorIs(t, err, context.Canceled)
+	assert.Zero(t, fallback.calls, "a cancelled job must not ship the audio to a cloud endpoint")
+}

internal/adapters/transcribe/gateway.go

@@ -0,0 +1,147 @@
+// Package transcribe turns a WAV file into a timed Swedish transcript by calling
+// the LiteLLM gateway's OpenAI-compatible /audio/transcriptions endpoint
+// (ADR-033). A Gateway is one model alias; a Chain tries a local one and falls
+// back to a second.
+package transcribe
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"mime/multipart"
+	"net/http"
+	"os"
+	"strings"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// ErrEmptyTranscript is returned when the endpoint answered 200 but with no text.
+// An empty transcript is not a success: chat over it could only answer "nothing".
+var ErrEmptyTranscript = errors.New("transcribe: endpoint returned an empty transcript")
+
+// StatusError is a non-2xx answer from the gateway. Nothing may branch on the
+// body — {"error":"failed to process audio"} has had at least five distinct
+// causes — so only the status code is exposed for logic; Snippet is for logs.
+type StatusError struct {
+	Code    int
+	Snippet string
+}
+
+func (e *StatusError) Error() string {
+	return fmt.Sprintf("transcribe: gateway returned HTTP %d: %s", e.Code, e.Snippet)
+}
+
+// Gateway transcribes through one model alias on the gateway. baseURL is the
+// gateway's /v1 root — the in-cluster NodePort, not the public hostname, whose
+// ingress rejects bodies over ~8-10 MB with 413 (docs/research/audio-length-ceiling).
+type Gateway struct {
+	baseURL string
+	key     string
+	model   string
+	client  *http.Client
+}
+
+// NewGateway builds a Gateway. The client should carry no overall Timeout: a 90
+// minute recording takes minutes, so the deadline belongs to the caller's context.
+func NewGateway(baseURL, key, model string, client *http.Client) *Gateway {
+	return &Gateway{baseURL: strings.TrimRight(baseURL, "/"), key: key, model: model, client: client}
+}
+
+// Name is the model alias; it is what gets recorded as the serving endpoint.
+func (g *Gateway) Name() string { return g.model }
+
+// Transcribe uploads the WAV at path and returns the normalised transcription.
+// The body is streamed from disk with a COMPUTED Content-Length: a 90 minute WAV
+// is ~170 MB, so it is neither buffered in memory nor sent chunked.
+func (g *Gateway) Transcribe(ctx context.Context, path string) (domain.Transcription, error) {
+	f, err := os.Open(path)
+	if err != nil {
+		return domain.Transcription{}, fmt.Errorf("transcribe: open audio: %w", err)
+	}
+	defer f.Close()
+	info, err := f.Stat()
+	if err != nil {
+		return domain.Transcription{}, fmt.Errorf("transcribe: stat audio: %w", err)
+	}
+
+	var prefix bytes.Buffer
+	mw := multipart.NewWriter(&prefix)
+	for _, kv := range [][2]string{{"model", g.model}, {"response_format", "verbose_json"}, {"language", "sv"}} {
+		if err := mw.WriteField(kv[0], kv[1]); err != nil {
+			return domain.Transcription{}, fmt.Errorf("transcribe: build request: %w", err)
+		}
+	}
+	// The multipart field is `file`; `audio_file` is rejected as "Invalid request".
+	if _, err := mw.CreateFormFile("file", "audio.wav"); err != nil {
+		return domain.Transcription{}, fmt.Errorf("transcribe: build request: %w", err)
+	}
+	suffix := []byte("\r\n--" + mw.Boundary() + "--\r\n")
+
+	body := io.MultiReader(bytes.NewReader(prefix.Bytes()), f, bytes.NewReader(suffix))
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, g.baseURL+"/audio/transcriptions", body)
+	if err != nil {
+		return domain.Transcription{}, fmt.Errorf("transcribe: build request: %w", err)
+	}
+	req.ContentLength = int64(prefix.Len()) + info.Size() + int64(len(suffix))
+	req.Header.Set("Content-Type", mw.FormDataContentType())
+	req.Header.Set("Authorization", "Bearer "+g.key)
+	// No "Expect: 100-continue": it trips the proxy for any body over 1 KB.
+
+	resp, err := g.client.Do(req)
+	if err != nil {
+		return domain.Transcription{}, fmt.Errorf("transcribe: %s: %w", g.model, err)
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode < 200 || resp.StatusCode > 299 {
+		snip, _ := io.ReadAll(io.LimitReader(resp.Body, 200))
+		return domain.Transcription{}, &StatusError{Code: resp.StatusCode, Snippet: strings.TrimSpace(string(snip))}
+	}
+
+	t, err := parseVerbose(resp.Body)
+	if err != nil {
+		return domain.Transcription{}, fmt.Errorf("transcribe: %s: %w", g.model, err)
+	}
+	t.Endpoint = g.model
+	return t, nil
+}
+
+// verboseJSON is the subset of the verbose_json response used. Both backends
+// return it; they differ in segment granularity (iguana: whole seconds, berget:
+// sub-second) and in the language token, neither of which matters here.
+type verboseJSON struct {
+	Text     string  `json:"text"`
+	Duration float64 `json:"duration"`
+	Segments []struct {
+		Start float64 `json:"start"`
+		End   float64 `json:"end"`
+		Text  string  `json:"text"`
+	} `json:"segments"`
+}
+
+// parseVerbose normalises a verbose_json body. Whisper's final window can end past
+// the audio (9.0 s for 4.25 s of audio was observed), so segment ends are clamped
+// to the reported duration.
+func parseVerbose(r io.Reader) (domain.Transcription, error) {
+	var v verboseJSON
+	if err := json.NewDecoder(r).Decode(&v); err != nil {
+		return domain.Transcription{}, fmt.Errorf("decode response: %w", err)
+	}
+	text := strings.TrimSpace(v.Text)
+	if text == "" {
+		return domain.Transcription{}, ErrEmptyTranscript
+	}
+	t := domain.Transcription{Text: text, AudioSeconds: v.Duration}
+	for _, s := range v.Segments {
+		seg := domain.Segment{Start: s.Start, End: s.End, Text: strings.TrimSpace(s.Text)}
+		if v.Duration > 0 {
+			seg.End = min(seg.End, v.Duration)
+			seg.Start = min(seg.Start, seg.End)
+		}
+		t.Segments = append(t.Segments, seg)
+	}
+	return t, nil
+}

internal/adapters/transcribe/gateway_test.go

@@ -0,0 +1,174 @@
+package transcribe_test
+
+import (
+	"bytes"
+	"context"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/transcribe"
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// The httptest server stands in for the LiteLLM gateway: it is the process
+// boundary, so a fake here is legitimate. The JSON it returns mirrors what the
+// real gateway returned for iguana/kb-whisper and berget/kb-whisper on 2026-10-01
+// (docs/research/audio-length-ceiling-2026-10-01.md). All text is synthetic.
+
+const verboseJSON = `{
+  "text": " God morgon allihopa.\n Vi går igenom läget.\n",
+  "language": "swedish",
+  "duration": 4.25,
+  "segments": [
+    {"id": 0, "start": 0.0, "end": 1.0, "text": " God morgon allihopa."},
+    {"id": 1, "start": 1.0, "end": 9.0, "text": " Vi går igenom läget."}
+  ]
+}`
+
+func wavFile(t *testing.T, content string) string {
+	t.Helper()
+	p := filepath.Join(t.TempDir(), "a.wav")
+	require.NoError(t, os.WriteFile(p, []byte(content), 0o600))
+	return p
+}
+
+type seen struct {
+	method, path, auth, expect, contentType string
+	contentLength                           int64
+	transferEncoding                        []string
+	fields                                  map[string]string
+	file                                    string
+	fileField                               string
+	bodyLen                                 int
+}
+
+// serve starts a gateway fake that records the request it received and replies
+// with status and body.
+func serve(t *testing.T, status int, reply string, got *seen) *httptest.Server {
+	t.Helper()
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if got != nil {
+			got.method, got.path = r.Method, r.URL.Path
+			got.auth = r.Header.Get("Authorization")
+			got.expect = r.Header.Get("Expect")
+			got.contentType = r.Header.Get("Content-Type")
+			got.contentLength = r.ContentLength
+			got.transferEncoding = r.TransferEncoding
+			raw, _ := io.ReadAll(r.Body)
+			got.bodyLen = len(raw)
+			r.Body = io.NopCloser(bytes.NewReader(raw))
+			require.NoError(t, r.ParseMultipartForm(32<<20))
+			got.fields = map[string]string{}
+			for k, v := range r.MultipartForm.Value {
+				got.fields[k] = v[0]
+			}
+			for field, fhs := range r.MultipartForm.File {
+				got.fileField = field
+				f, _ := fhs[0].Open()
+				b, _ := io.ReadAll(f)
+				got.file = string(b)
+			}
+		}
+		w.WriteHeader(status)
+		_, _ = io.WriteString(w, reply)
+	}))
+	t.Cleanup(srv.Close)
+	return srv
+}
+
+func TestGateway_SendsTheRequestTheGatewayExpects(t *testing.T) {
+	var got seen
+	srv := serve(t, 200, verboseJSON, &got)
+	g := transcribe.NewGateway(srv.URL+"/v1", "secret-key", "iguana/kb-whisper", srv.Client())
+
+	_, err := g.Transcribe(context.Background(), wavFile(t, "RIFF-synthetic-pcm"))
+
+	require.NoError(t, err)
+	assert.Equal(t, http.MethodPost, got.method)
+	assert.Equal(t, "/v1/audio/transcriptions", got.path)
+	assert.Equal(t, "Bearer secret-key", got.auth)
+	assert.Equal(t, "file", got.fileField, "the multipart field is `file`; `audio_file` is rejected as 'Invalid request'")
+	assert.Equal(t, "RIFF-synthetic-pcm", got.file)
+	assert.Equal(t, "iguana/kb-whisper", got.fields["model"])
+	assert.Equal(t, "verbose_json", got.fields["response_format"])
+	assert.Equal(t, "sv", got.fields["language"])
+	assert.Empty(t, got.expect, "no Expect: 100-continue: it trips the proxy for any body over 1 KB")
+}
+
+func TestGateway_SendsAKnownContentLengthNotChunked(t *testing.T) {
+	var got seen
+	srv := serve(t, 200, verboseJSON, &got)
+	g := transcribe.NewGateway(srv.URL+"/v1", "k", "m", srv.Client())
+
+	_, err := g.Transcribe(context.Background(), wavFile(t, "0123456789"))
+
+	require.NoError(t, err)
+	assert.Empty(t, got.transferEncoding, "a 90-minute WAV is ~170 MB: stream it with a computed length, not chunked and not buffered")
+	assert.Equal(t, int64(got.bodyLen), got.contentLength)
+}
+
+func TestGateway_NormalisesTheVerboseJSONResponse(t *testing.T) {
+	srv := serve(t, 200, verboseJSON, nil)
+	g := transcribe.NewGateway(srv.URL+"/v1", "k", "iguana/kb-whisper", srv.Client())
+
+	got, err := g.Transcribe(context.Background(), wavFile(t, "x"))
+
+	require.NoError(t, err)
+	assert.Equal(t, "God morgon allihopa.\n Vi går igenom läget.", got.Text)
+	assert.Equal(t, "iguana/kb-whisper", got.Endpoint, "the serving endpoint is part of the result")
+	assert.InDelta(t, 4.25, got.AudioSeconds, 1e-9)
+	assert.Equal(t, []domain.Segment{
+		{Start: 0, End: 1, Text: "God morgon allihopa."},
+		// The final segment ended at 9.0 s for 4.25 s of audio: whisper's last
+		// window can run past the end. Clamped to the audio duration.
+		{Start: 1, End: 4.25, Text: "Vi går igenom läget."},
+	}, got.Segments)
+}
+
+func TestGateway_ErrorsDoNotDependOnTheBody(t *testing.T) {
+	// {"error":"failed to process audio"} has had at least five distinct causes, so
+	// no code may branch on it. Two different bodies must be treated identically.
+	for _, reply := range []string{`{"error":"failed to process audio"}`, `{"error":"something else entirely"}`} {
+		srv := serve(t, 500, reply, nil)
+		g := transcribe.NewGateway(srv.URL+"/v1", "k", "m", srv.Client())
+
+		_, err := g.Transcribe(context.Background(), wavFile(t, "x"))
+
+		var se *transcribe.StatusError
+		require.ErrorAs(t, err, &se)
+		assert.Equal(t, 500, se.Code)
+	}
+}
+
+func TestGateway_AnEmptyTranscriptIsAFailureNotASuccess(t *testing.T) {
+	srv := serve(t, 200, `{"text":"  ","language":"swedish","duration":3,"segments":[]}`, nil)
+	g := transcribe.NewGateway(srv.URL+"/v1", "k", "m", srv.Client())
+
+	_, err := g.Transcribe(context.Background(), wavFile(t, "x"))
+
+	assert.ErrorIs(t, err, transcribe.ErrEmptyTranscript)
+}
+
+func TestGateway_MalformedJSONIsAnError(t *testing.T) {
+	srv := serve(t, 200, `not json`, nil)
+	g := transcribe.NewGateway(srv.URL+"/v1", "k", "m", srv.Client())
+
+	_, err := g.Transcribe(context.Background(), wavFile(t, "x"))
+
+	assert.Error(t, err)
+}
+
+func TestGateway_MissingAudioFileIsAnError(t *testing.T) {
+	g := transcribe.NewGateway("http://127.0.0.1:1/v1", "k", "m", http.DefaultClient)
+
+	_, err := g.Transcribe(context.Background(), filepath.Join(t.TempDir(), "nope.wav"))
+
+	assert.Error(t, err)
+}

internal/adapters/transcribe/live_test.go

@@ -0,0 +1,49 @@
+package transcribe_test
+
+import (
+	"context"
+	"net/http"
+	"os"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/transcribe"
+)
+
+// TestGateway_LiveGateway is an opt-in check against the REAL gateway, because
+// the httptest fake above can only prove what we think the gateway accepts. It
+// is skipped unless TAPIR_LIVE_WAV names a 16 kHz mono WAV and the gateway
+// coordinates are set:
+//
+//	TAPIR_LIVE_WAV=/path/a.wav TAPIR_LIVE_GATEWAY_URL=http://koala:30401/v1 \
+//	TAPIR_LIVE_MODEL=iguana/kb-whisper op run --env-file ~/.op-env -- \
+//	  go test ./internal/adapters/transcribe -run LiveGateway -v
+//
+// Use SYNTHETIC audio only: this repo is public and the gateway may be a cloud
+// endpoint (#33).
+func TestGateway_LiveGateway(t *testing.T) {
+	wav := os.Getenv("TAPIR_LIVE_WAV")
+	url := os.Getenv("TAPIR_LIVE_GATEWAY_URL")
+	if wav == "" || url == "" {
+		t.Skip("set TAPIR_LIVE_WAV and TAPIR_LIVE_GATEWAY_URL to run against the real gateway")
+	}
+	model := os.Getenv("TAPIR_LIVE_MODEL")
+	if model == "" {
+		model = "iguana/kb-whisper"
+	}
+	key := os.Getenv("LITELLM_MASTER_KEY")
+	require.NotEmpty(t, key, "LITELLM_MASTER_KEY must be set (via op run)")
+
+	ctx, cancel := context.WithTimeout(context.Background(), 25*time.Minute)
+	defer cancel()
+	got, err := transcribe.NewGateway(url, key, model, &http.Client{}).Transcribe(ctx, wav)
+
+	require.NoError(t, err)
+	assert.NotEmpty(t, got.Text)
+	assert.NotEmpty(t, got.Segments)
+	assert.Equal(t, model, got.Endpoint)
+	t.Logf("endpoint=%s audio_s=%.1f segments=%d chars=%d", got.Endpoint, got.AudioSeconds, len(got.Segments), len(got.Text))
+}

internal/config/config.go

@@ -20,6 +20,9 @@ import (
 
 // Config is Tapir's fully-resolved runtime configuration.
 type Config struct {
+	// UploadConfig is the audio-upload and private-transcription settings (ADR-033).
+	UploadConfig
+
 	// UserID is the Tapir user the run operates as. Stage 0 has exactly one.
 	// It must be a UUID: it keys the UUID user_id/video_id columns (data-model).
 	UserID string
@@ -320,6 +323,10 @@ func Load() (Config, error) {
 	}
 	c.OnboardMaxVideoSeconds = onboardMax
 
+	if c.UploadConfig, err = loadUploadConfig(); err != nil {
+		return Config{}, err
+	}
+
 	return c, nil
 }
 

internal/config/uploads.go

@@ -0,0 +1,110 @@
+package config
+
+import (
+	"fmt"
+	"os"
+	"strconv"
+	"time"
+)
+
+// UploadConfig holds the settings for audio upload and private transcription
+// (ADR-033). It is embedded in Config, so its fields read as c.UploadsEnabled etc.
+type UploadConfig struct {
+	// UploadsEnabled mounts the /uploads routes and starts the worker. Off by
+	// default: the runtime image must contain ffmpeg and ffprobe first (ADR-033).
+	UploadsEnabled bool
+	// UploadMaxBytes caps one upload. UploadMaxDuration caps one recording; 90
+	// minutes is the longest length MEASURED, not a ceiling that was found, so
+	// raising it is a measurement, not an edit.
+	UploadMaxBytes    int64
+	UploadMaxDuration time.Duration
+	// TranscribeModel is the primary gateway alias; TranscribeFallbackModel is the
+	// fallback, tried on ANY primary failure. The fallback is a cloud endpoint
+	// (berget): set it EMPTY to keep private audio on the homelab.
+	TranscribeModel         string
+	TranscribeFallbackModel string
+	// TranscribeTimeout bounds one job (conversion + transcription); a 90 minute
+	// recording took ~7.5 minutes on iguana.
+	TranscribeTimeout time.Duration
+	// UploadTmpDir is where uploads are spooled and converted. Audio is deleted
+	// after transcription, so this holds at most the in-flight files.
+	UploadTmpDir string
+	FFmpegBin    string
+	FFprobeBin   string
+	// MaxUploadedTranscriptChars bounds the transcript sent to a chat model for an
+	// UPLOADED recording. The 18,000-char default for captions is ~22 minutes of
+	// speech; an hour of meeting is ~50,000 chars. A guess until measured against
+	// the models chat offers.
+	MaxUploadedTranscriptChars int
+}
+
+const (
+	defaultUploadMaxMB             = 200
+	defaultUploadMaxMinutes        = 90
+	defaultTranscribeModel         = "iguana/kb-whisper"
+	defaultTranscribeFallbackModel = "berget/kb-whisper"
+	defaultTranscribeTimeout       = 30 * time.Minute
+	defaultMaxUploadedChars        = 120000
+)
+
+func loadUploadConfig() (UploadConfig, error) {
+	u := UploadConfig{
+		TranscribeModel:         envOr("TAPIR_TRANSCRIBE_MODEL", defaultTranscribeModel),
+		TranscribeFallbackModel: lookupOr("TAPIR_TRANSCRIBE_FALLBACK_MODEL", defaultTranscribeFallbackModel),
+		UploadTmpDir:            envOr("TAPIR_UPLOAD_TMP_DIR", os.TempDir()),
+		FFmpegBin:               envOr("TAPIR_FFMPEG_BIN", "ffmpeg"),
+		FFprobeBin:              envOr("TAPIR_FFPROBE_BIN", "ffprobe"),
+	}
+
+	if v := os.Getenv("TAPIR_UPLOADS_ENABLED"); v != "" {
+		on, err := strconv.ParseBool(v)
+		if err != nil {
+			return UploadConfig{}, fmt.Errorf("config: TAPIR_UPLOADS_ENABLED=%q: %w", v, err)
+		}
+		u.UploadsEnabled = on
+	}
+
+	mb, err := positiveInt("TAPIR_UPLOAD_MAX_MB", defaultUploadMaxMB)
+	if err != nil {
+		return UploadConfig{}, err
+	}
+	u.UploadMaxBytes = int64(mb) << 20
+
+	mins, err := positiveInt("TAPIR_UPLOAD_MAX_MINUTES", defaultUploadMaxMinutes)
+	if err != nil {
+		return UploadConfig{}, err
+	}
+	u.UploadMaxDuration = time.Duration(mins) * time.Minute
+
+	timeout, err := durationOr("TAPIR_TRANSCRIBE_TIMEOUT", defaultTranscribeTimeout)
+	if err != nil {
+		return UploadConfig{}, err
+	}
+	if timeout <= 0 {
+		return UploadConfig{}, fmt.Errorf("config: TAPIR_TRANSCRIBE_TIMEOUT must be positive, got %v", timeout)
+	}
+	u.TranscribeTimeout = timeout
+
+	chars, err := intOr("TAPIR_MAX_UPLOADED_TRANSCRIPT_CHARS", defaultMaxUploadedChars)
+	if err != nil {
+		return UploadConfig{}, err
+	}
+	if chars < 0 {
+		chars = 0
+	}
+	u.MaxUploadedTranscriptChars = chars
+	return u, nil
+}
+
+// positiveInt is intOr that refuses zero and negatives: a zero upload limit would
+// reject every file, which is a misconfiguration to fail on, not to run with.
+func positiveInt(key string, fallback int) (int, error) {
+	n, err := intOr(key, fallback)
+	if err != nil {
+		return 0, err
+	}
+	if n <= 0 {
+		return 0, fmt.Errorf("config: %s must be positive, got %d", key, n)
+	}
+	return n, nil
+}

internal/config/uploads_test.go

@@ -0,0 +1,123 @@
+package config
+
+import (
+	"os"
+	"testing"
+	"time"
+)
+
+// clearUploadEnv unsets every upload knob so a test observes the default, not the
+// host environment.
+func clearUploadEnv(t *testing.T) {
+	t.Helper()
+	for _, k := range []string{
+		"TAPIR_UPLOADS_ENABLED", "TAPIR_UPLOAD_MAX_MB", "TAPIR_UPLOAD_MAX_MINUTES",
+		"TAPIR_TRANSCRIBE_MODEL", "TAPIR_TRANSCRIBE_FALLBACK_MODEL", "TAPIR_TRANSCRIBE_TIMEOUT",
+		"TAPIR_UPLOAD_TMP_DIR", "TAPIR_FFMPEG_BIN", "TAPIR_FFPROBE_BIN",
+		"TAPIR_MAX_UPLOADED_TRANSCRIPT_CHARS",
+	} {
+		unset(t, k)
+	}
+}
+
+func TestLoad_UploadDefaults(t *testing.T) {
+	clearUploadEnv(t)
+
+	c, err := Load()
+	if err != nil {
+		t.Fatalf("Load: %v", err)
+	}
+
+	if c.UploadsEnabled {
+		t.Error("uploads must be OFF by default: the runtime image needs ffmpeg first (ADR-033)")
+	}
+	if c.UploadMaxBytes != 200<<20 {
+		t.Errorf("UploadMaxBytes = %d, want 200 MB", c.UploadMaxBytes)
+	}
+	if c.UploadMaxDuration != 90*time.Minute {
+		t.Errorf("UploadMaxDuration = %v, want 90m (the longest length measured, not a found ceiling)", c.UploadMaxDuration)
+	}
+	if c.TranscribeModel != "iguana/kb-whisper" {
+		t.Errorf("TranscribeModel = %q", c.TranscribeModel)
+	}
+	if c.TranscribeFallbackModel != "berget/kb-whisper" {
+		t.Errorf("TranscribeFallbackModel = %q", c.TranscribeFallbackModel)
+	}
+	if c.TranscribeTimeout != 30*time.Minute {
+		t.Errorf("TranscribeTimeout = %v, want 30m (a 90 min recording took ~7.5 min)", c.TranscribeTimeout)
+	}
+	if c.UploadTmpDir != os.TempDir() {
+		t.Errorf("UploadTmpDir = %q, want os.TempDir()", c.UploadTmpDir)
+	}
+	if c.FFmpegBin != "ffmpeg" || c.FFprobeBin != "ffprobe" {
+		t.Errorf("bins = %q, %q", c.FFmpegBin, c.FFprobeBin)
+	}
+	if c.MaxUploadedTranscriptChars != 120000 {
+		t.Errorf("MaxUploadedTranscriptChars = %d, want 120000", c.MaxUploadedTranscriptChars)
+	}
+}
+
+func TestLoad_UploadValuesAreParsed(t *testing.T) {
+	clearUploadEnv(t)
+	setEnv(t, map[string]string{
+		"TAPIR_UPLOADS_ENABLED":               "true",
+		"TAPIR_UPLOAD_MAX_MB":                 "50",
+		"TAPIR_UPLOAD_MAX_MINUTES":            "30",
+		"TAPIR_TRANSCRIBE_MODEL":              "koala/whisper",
+		"TAPIR_TRANSCRIBE_TIMEOUT":            "10m",
+		"TAPIR_UPLOAD_TMP_DIR":                "/var/spool/tapir",
+		"TAPIR_FFMPEG_BIN":                    "/opt/ffmpeg",
+		"TAPIR_FFPROBE_BIN":                   "/opt/ffprobe",
+		"TAPIR_MAX_UPLOADED_TRANSCRIPT_CHARS": "60000",
+	})
+
+	c, err := Load()
+	if err != nil {
+		t.Fatalf("Load: %v", err)
+	}
+
+	if !c.UploadsEnabled || c.UploadMaxBytes != 50<<20 || c.UploadMaxDuration != 30*time.Minute ||
+		c.TranscribeModel != "koala/whisper" || c.TranscribeTimeout != 10*time.Minute ||
+		c.UploadTmpDir != "/var/spool/tapir" || c.FFmpegBin != "/opt/ffmpeg" ||
+		c.FFprobeBin != "/opt/ffprobe" || c.MaxUploadedTranscriptChars != 60000 {
+		t.Errorf("values not applied: %+v", c)
+	}
+}
+
+func TestLoad_EmptyTranscribeFallbackDisablesIt(t *testing.T) {
+	clearUploadEnv(t)
+	t.Setenv("TAPIR_TRANSCRIBE_FALLBACK_MODEL", "")
+
+	c, err := Load()
+	if err != nil {
+		t.Fatalf("Load: %v", err)
+	}
+
+	if c.TranscribeFallbackModel != "" {
+		t.Errorf("TranscribeFallbackModel = %q: an explicit empty value must disable the cloud fallback, "+
+			"so private audio can be kept on the homelab", c.TranscribeFallbackModel)
+	}
+}
+
+func TestLoad_UploadRejectsInvalidValues(t *testing.T) {
+	tests := []struct{ key, val string }{
+		{"TAPIR_UPLOADS_ENABLED", "maybe"},
+		{"TAPIR_UPLOAD_MAX_MB", "lots"},
+		{"TAPIR_UPLOAD_MAX_MB", "0"},
+		{"TAPIR_UPLOAD_MAX_MB", "-5"},
+		{"TAPIR_UPLOAD_MAX_MINUTES", "0"},
+		{"TAPIR_UPLOAD_MAX_MINUTES", "forever"},
+		{"TAPIR_TRANSCRIBE_TIMEOUT", "soon"},
+		{"TAPIR_TRANSCRIBE_TIMEOUT", "0s"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.key+"="+tt.val, func(t *testing.T) {
+			clearUploadEnv(t)
+			t.Setenv(tt.key, tt.val)
+
+			if _, err := Load(); err == nil {
+				t.Errorf("Load accepted %s=%q; a zero or negative limit would reject every upload", tt.key, tt.val)
+			}
+		})
+	}
+}

internal/domain/domain.go

@@ -111,3 +111,62 @@ type Summary struct {
 	FallbackUsed bool
 	CreatedAt    time.Time
 }
+
+// Segment is one timed span of a transcription. Times are seconds from the start
+// of the audio.
+type Segment struct {
+	Start float64 `json:"start"`
+	End   float64 `json:"end"`
+	Text  string  `json:"text"`
+}
+
+// Transcription is the result of transcribing user-supplied audio (ADR-033):
+// the full text, its timed segments, which backend served it, and how long the
+// audio was. Endpoint is recorded because the backends differ in timestamp
+// granularity — which one served a transcript is a quality fact.
+type Transcription struct {
+	Text         string
+	Segments     []Segment
+	Endpoint     string
+	AudioSeconds float64
+}
+
+// UploadStatus is the state of an uploaded-audio transcription job.
+type UploadStatus string
+
+const (
+	UploadQueued       UploadStatus = "queued"
+	UploadTranscribing UploadStatus = "transcribing"
+	UploadDone         UploadStatus = "done"
+	UploadFailed       UploadStatus = "failed"
+)
+
+// UploadErrorKind is the typed reason an upload failed. It is deliberately NOT a
+// provider's error string: the whisper server's "failed to process audio" has had
+// at least five distinct causes, so nothing may branch on it (ADR-033).
+type UploadErrorKind string
+
+const (
+	// UploadErrInterrupted: the process restarted mid-job. The audio is not
+	// retained, so the job cannot resume; the user re-uploads.
+	UploadErrInterrupted UploadErrorKind = "interrupted"
+	// UploadErrTranscriptionFailed: every configured transcription endpoint failed.
+	UploadErrTranscriptionFailed UploadErrorKind = "transcription_failed"
+)
+
+// UploadedTranscript is a user's PRIVATE transcript of audio they uploaded. It is
+// never stored in, or merged with, the shared public-caption Transcript (ADR-021).
+// Content and Segments are empty until Status is UploadDone.
+type UploadedTranscript struct {
+	ID           string
+	UserID       string
+	Title        string
+	Status       UploadStatus
+	Content      string
+	Segments     []Segment
+	Endpoint     string
+	AudioSeconds float64
+	ErrorKind    UploadErrorKind
+	CreatedAt    time.Time
+	FinishedAt   time.Time // zero until the job reaches done or failed
+}

internal/uploads/intake.go

@@ -0,0 +1,115 @@
+package uploads
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// ErrUnreadable is what a Prober returns when the FILE is the problem (not media,
+// truncated, damaged), as opposed to the prober itself failing. Accept turns it
+// into a user-facing rejection; any other prober error is our fault and surfaces
+// as an internal error — a missing ffprobe must not read as "your file is bad".
+var ErrUnreadable = errors.New("uploads: file is not readable media")
+
+// Prober inspects a file on disk. The ffprobe adapter implements it.
+type Prober interface {
+	Probe(ctx context.Context, path string) (Probe, error)
+}
+
+// Store is the slice of the private transcript store the intake needs.
+type Store interface {
+	CreateUpload(ctx context.Context, userID, title string) (string, error)
+	FailUpload(ctx context.Context, userID, id string, kind domain.UploadErrorKind) error
+}
+
+// Job is one accepted upload awaiting transcription. Path is a temp file the
+// receiver owns: it must delete it when finished (the audio is never retained).
+type Job struct {
+	UserID   string
+	UploadID string
+	Path     string
+}
+
+// Submitter hands accepted jobs to the in-process worker (ADR-033).
+type Submitter interface {
+	Submit(Job) error
+}
+
+// Intake validates an uploaded recording and, if it passes, queues it.
+type Intake struct {
+	store  Store
+	prober Prober
+	sub    Submitter
+	limits Limits
+	tmpDir string
+}
+
+// NewIntake wires an Intake. tmpDir is where uploads are spooled; it must exist.
+func NewIntake(s Store, p Prober, sub Submitter, l Limits, tmpDir string) *Intake {
+	return &Intake{store: s, prober: p, sub: sub, limits: l, tmpDir: tmpDir}
+}
+
+// Accept spools src to disk (bounded by MaxBytes), probes it, validates it, records
+// a queued upload and submits the job. It returns the upload id. A file that fails
+// validation yields a *RejectError and creates NO row — rejection is synchronous
+// and leaves nothing behind. On success the temp file is handed to the job and the
+// worker deletes it; on every failure path Accept deletes it itself.
+func (in *Intake) Accept(ctx context.Context, userID, title string, src io.Reader) (string, error) {
+	f, err := os.CreateTemp(in.tmpDir, "upload-*")
+	if err != nil {
+		return "", fmt.Errorf("uploads: spool: %w", err)
+	}
+	path := f.Name()
+	handedOver := false
+	defer func() {
+		if !handedOver {
+			_ = os.Remove(path)
+		}
+	}()
+
+	// Read one byte past the cap so "exactly at the limit" and "over" are told apart.
+	n, err := io.Copy(f, io.LimitReader(src, in.limits.MaxBytes+1))
+	if cerr := f.Close(); err == nil {
+		err = cerr
+	}
+	if err != nil {
+		return "", fmt.Errorf("uploads: receive: %w", err)
+	}
+	if n > in.limits.MaxBytes {
+		return "", &RejectError{Reason: ReasonTooLarge, Message: fmt.Sprintf(
+			"This file is larger than the %d MB limit.", in.limits.MaxBytes>>20)}
+	}
+
+	probe, err := in.prober.Probe(ctx, path)
+	if errors.Is(err, ErrUnreadable) {
+		return "", &RejectError{
+			Reason:  ReasonUnreadable,
+			Message: "This file could not be read as audio; it may be damaged or not a recording.",
+		}
+	}
+	if err != nil {
+		return "", fmt.Errorf("uploads: probe: %w", err)
+	}
+	if err := Validate(probe, in.limits); err != nil {
+		return "", err
+	}
+
+	id, err := in.store.CreateUpload(ctx, userID, title)
+	if err != nil {
+		return "", fmt.Errorf("uploads: create: %w", err)
+	}
+	if err := in.sub.Submit(Job{UserID: userID, UploadID: id, Path: path}); err != nil {
+		// A row that will never run must not sit queued: that is the stuck spinner.
+		if ferr := in.store.FailUpload(ctx, userID, id, domain.UploadErrInterrupted); ferr != nil {
+			return "", fmt.Errorf("uploads: submit: %w (and failing the row: %v)", err, ferr)
+		}
+		return "", fmt.Errorf("uploads: submit: %w", err)
+	}
+	handedOver = true
+	return id, nil
+}

internal/uploads/intake_test.go

@@ -0,0 +1,212 @@
+package uploads_test
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"io"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"testing/iotest"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+type fakeStore struct {
+	created []string // titles
+	failed  map[string]domain.UploadErrorKind
+	nextID  string
+}
+
+func (f *fakeStore) CreateUpload(_ context.Context, _, title string) (string, error) {
+	f.created = append(f.created, title)
+	return f.nextID, nil
+}
+
+func (f *fakeStore) FailUpload(_ context.Context, _, id string, kind domain.UploadErrorKind) error {
+	if f.failed == nil {
+		f.failed = map[string]domain.UploadErrorKind{}
+	}
+	f.failed[id] = kind
+	return nil
+}
+
+type fakeProber struct {
+	probe uploads.Probe
+	err   error
+	paths []string
+	seen  []byte // content of the file at probe time
+}
+
+func (f *fakeProber) Probe(_ context.Context, path string) (uploads.Probe, error) {
+	f.paths = append(f.paths, path)
+	f.seen, _ = os.ReadFile(path)
+	return f.probe, f.err
+}
+
+type fakeSubmitter struct {
+	jobs []uploads.Job
+	err  error
+}
+
+func (f *fakeSubmitter) Submit(j uploads.Job) error {
+	f.jobs = append(f.jobs, j)
+	return f.err
+}
+
+type rig struct {
+	in  *uploads.Intake
+	st  *fakeStore
+	pr  *fakeProber
+	sub *fakeSubmitter
+	tmp string
+}
+
+func newRig(t *testing.T) *rig {
+	t.Helper()
+	r := &rig{
+		st:  &fakeStore{nextID: "up-1"},
+		pr:  &fakeProber{probe: uploads.Probe{Duration: 10 * time.Minute, HasAudio: true}},
+		sub: &fakeSubmitter{},
+		tmp: t.TempDir(),
+	}
+	r.in = uploads.NewIntake(r.st, r.pr, r.sub, uploads.Limits{MaxBytes: 1024, MaxDuration: 90 * time.Minute}, r.tmp)
+	return r
+}
+
+func (r *rig) tmpFiles(t *testing.T) []string {
+	t.Helper()
+	ents, err := os.ReadDir(r.tmp)
+	require.NoError(t, err)
+	var names []string
+	for _, e := range ents {
+		names = append(names, e.Name())
+	}
+	return names
+}
+
+func TestAccept_ValidAudioIsStoredProbedQueuedAndSubmitted(t *testing.T) {
+	r := newRig(t)
+	audio := bytes.Repeat([]byte("a"), 500)
+
+	id, err := r.in.Accept(context.Background(), "user-1", "Teams export", bytes.NewReader(audio))
+
+	require.NoError(t, err)
+	assert.Equal(t, "up-1", id)
+	assert.Equal(t, []string{"Teams export"}, r.st.created)
+	assert.Equal(t, audio, r.pr.seen, "the prober must see the complete uploaded bytes, on disk")
+	require.Len(t, r.sub.jobs, 1)
+	job := r.sub.jobs[0]
+	assert.Equal(t, uploads.Job{UserID: "user-1", UploadID: "up-1", Path: job.Path}, job)
+	assert.Equal(t, filepath.Dir(job.Path), r.tmp, "the temp file lives in the configured directory")
+	got, err := os.ReadFile(job.Path)
+	require.NoError(t, err)
+	assert.Equal(t, audio, got, "the handed-over file still exists: the worker, not Accept, deletes it")
+}
+
+func TestAccept_OversizeBodyIsRejectedAndLeavesNothingBehind(t *testing.T) {
+	r := newRig(t)
+
+	_, err := r.in.Accept(context.Background(), "user-1", "big", strings.NewReader(strings.Repeat("a", 1025)))
+
+	var rej *uploads.RejectError
+	require.True(t, errors.As(err, &rej))
+	assert.Equal(t, uploads.ReasonTooLarge, rej.Reason)
+	assert.Contains(t, rej.Message, "limit")
+	assert.Empty(t, r.pr.paths, "a file already over the size cap must not be probed")
+	assert.Empty(t, r.st.created, "a rejected upload creates no row")
+	assert.Empty(t, r.sub.jobs)
+	assert.Empty(t, r.tmpFiles(t), "the partial temp file must be removed")
+}
+
+func TestAccept_BodyExactlyAtTheLimitIsAccepted(t *testing.T) {
+	r := newRig(t)
+
+	_, err := r.in.Accept(context.Background(), "user-1", "edge", strings.NewReader(strings.Repeat("a", 1024)))
+
+	require.NoError(t, err)
+}
+
+func TestAccept_ProbeVerdictRejectsWithoutRowAndRemovesTheFile(t *testing.T) {
+	tests := []struct {
+		name   string
+		probe  uploads.Probe
+		reason uploads.Reason
+	}{
+		{"video", uploads.Probe{Duration: time.Minute, HasAudio: true, HasVideo: true}, uploads.ReasonVideo},
+		{"too long", uploads.Probe{Duration: 2 * time.Hour, HasAudio: true}, uploads.ReasonTooLong},
+		{"no audio", uploads.Probe{Duration: time.Minute}, uploads.ReasonNoAudio},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			r := newRig(t)
+			r.pr.probe = tt.probe
+
+			_, err := r.in.Accept(context.Background(), "user-1", "t", strings.NewReader("data"))
+
+			var rej *uploads.RejectError
+			require.True(t, errors.As(err, &rej))
+			assert.Equal(t, tt.reason, rej.Reason)
+			assert.Empty(t, r.st.created)
+			assert.Empty(t, r.sub.jobs)
+			assert.Empty(t, r.tmpFiles(t))
+		})
+	}
+}
+
+func TestAccept_UnreadableFileIsARejectionButAProberFaultIsAnError(t *testing.T) {
+	t.Run("ffprobe says the file is not media", func(t *testing.T) {
+		r := newRig(t)
+		r.pr.err = uploads.ErrUnreadable
+
+		_, err := r.in.Accept(context.Background(), "user-1", "t", strings.NewReader("not media"))
+
+		var rej *uploads.RejectError
+		require.True(t, errors.As(err, &rej), "the user's file is the problem: tell them")
+		assert.Equal(t, uploads.ReasonUnreadable, rej.Reason)
+		assert.Empty(t, r.tmpFiles(t))
+	})
+	t.Run("the prober itself broke", func(t *testing.T) {
+		r := newRig(t)
+		r.pr.err = errors.New("exec: ffprobe: executable file not found")
+
+		_, err := r.in.Accept(context.Background(), "user-1", "t", strings.NewReader("data"))
+
+		require.Error(t, err)
+		var rej *uploads.RejectError
+		assert.False(t, errors.As(err, &rej), "our fault must not be reported as the user's bad file")
+		assert.Empty(t, r.st.created)
+		assert.Empty(t, r.tmpFiles(t))
+	})
+}
+
+func TestAccept_ClientDisconnectMidUploadLeavesNothingBehind(t *testing.T) {
+	r := newRig(t)
+	body := io.MultiReader(strings.NewReader("partial"), iotest.ErrReader(errors.New("client went away")))
+
+	_, err := r.in.Accept(context.Background(), "user-1", "t", body)
+
+	require.Error(t, err)
+	assert.Empty(t, r.pr.paths)
+	assert.Empty(t, r.st.created)
+	assert.Empty(t, r.tmpFiles(t))
+}
+
+func TestAccept_SubmitFailureFailsTheRowHonestlyAndRemovesTheFile(t *testing.T) {
+	r := newRig(t)
+	r.sub.err = errors.New("worker stopped")
+
+	_, err := r.in.Accept(context.Background(), "user-1", "t", strings.NewReader("data"))
+
+	require.Error(t, err)
+	assert.Equal(t, domain.UploadErrInterrupted, r.st.failed["up-1"],
+		"a row that will never run must not sit queued: that is the stuck spinner ADR-025 forbids")
+	assert.Empty(t, r.tmpFiles(t))
+}

internal/uploads/validate.go

@@ -0,0 +1,82 @@
+// Package uploads is the use case behind "upload an audio recording" (ADR-033):
+// validate the file by PROBING it, keep the transcript private, never retain the
+// audio. It depends only on ports (a Prober, a store, a job Submitter) — the
+// concrete ffprobe adapter, the web handler and the transcription worker are wired
+// from outside.
+package uploads
+
+import (
+	"fmt"
+	"time"
+)
+
+// Limits bound what an upload may be. Both are rejected synchronously, with a
+// specific message, rather than by a timeout (ADR-033).
+type Limits struct {
+	MaxBytes    int64
+	MaxDuration time.Duration
+}
+
+// Probe is what inspecting an uploaded file revealed. It is the input to Validate
+// because a name or MIME type cannot say whether a container carries video: a
+// Teams voice-only export is typically an .mp4 with no video stream.
+type Probe struct {
+	Duration time.Duration
+	HasAudio bool
+	HasVideo bool
+}
+
+// Reason is the machine-checkable class of a rejection.
+type Reason string
+
+const (
+	ReasonVideo      Reason = "video"
+	ReasonNoAudio    Reason = "no_audio"
+	ReasonTooLong    Reason = "too_long"
+	ReasonTooLarge   Reason = "too_large"
+	ReasonUnreadable Reason = "unreadable"
+)
+
+// RejectError is a synchronous, user-presentable refusal. Message is written for
+// the person who uploaded the file; Reason is for code and tests.
+type RejectError struct {
+	Reason  Reason
+	Message string
+}
+
+func (e *RejectError) Error() string { return string(e.Reason) + ": " + e.Message }
+
+// Validate decides whether a probed file may be transcribed. Video is checked
+// first so a video file is always refused as video, whatever its length.
+func Validate(p Probe, l Limits) error {
+	switch {
+	case p.HasVideo:
+		return &RejectError{Reason: ReasonVideo, Message: "This file contains a video track. " +
+			"Only audio recordings are supported for now; video upload is tracked in issue #28."}
+	case !p.HasAudio:
+		return &RejectError{Reason: ReasonNoAudio, Message: "This file has no audio track to transcribe."}
+	case p.Duration <= 0:
+		return &RejectError{Reason: ReasonUnreadable, Message: "Could not read a length from this file; it may be damaged."}
+	case p.Duration > l.MaxDuration:
+		return &RejectError{Reason: ReasonTooLong, Message: fmt.Sprintf(
+			"This recording is %s long; the limit is %s.", FormatDuration(p.Duration), FormatDuration(l.MaxDuration))}
+	}
+	return nil
+}
+
+// FormatDuration renders a length for people: "1 h 42 min", "45 min", "30 s".
+func FormatDuration(d time.Duration) string {
+	d = d.Round(time.Second)
+	h := int(d / time.Hour)
+	m := int(d % time.Hour / time.Minute)
+	switch {
+	case h > 0 && m > 0:
+		return fmt.Sprintf("%d h %d min", h, m)
+	case h > 0:
+		return fmt.Sprintf("%d h", h)
+	case m > 0:
+		return fmt.Sprintf("%d min", m)
+	default:
+		return fmt.Sprintf("%d s", int(d/time.Second))
+	}
+}

internal/uploads/validate_test.go

@@ -0,0 +1,95 @@
+package uploads_test
+
+import (
+	"errors"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+var testLimits = uploads.Limits{MaxBytes: 200 << 20, MaxDuration: 90 * time.Minute}
+
+func TestValidate(t *testing.T) {
+	tests := []struct {
+		name       string
+		probe      uploads.Probe
+		wantReason uploads.Reason // "" = accepted
+		wantInMsg  []string
+	}{
+		{
+			name:  "audio-only file within the duration limit is accepted",
+			probe: uploads.Probe{Duration: 45 * time.Minute, HasAudio: true},
+		},
+		{
+			name:  "a recording exactly at the limit is accepted",
+			probe: uploads.Probe{Duration: 90 * time.Minute, HasAudio: true},
+		},
+		{
+			name:       "a file with a video stream is rejected and names issue 28",
+			probe:      uploads.Probe{Duration: 4 * time.Minute, HasAudio: true, HasVideo: true},
+			wantReason: uploads.ReasonVideo,
+			wantInMsg:  []string{"video", "#28"},
+		},
+		{
+			name:       "video with no audio is still rejected as video, not as missing audio",
+			probe:      uploads.Probe{Duration: 4 * time.Minute, HasVideo: true},
+			wantReason: uploads.ReasonVideo,
+			wantInMsg:  []string{"#28"},
+		},
+		{
+			name:       "a file with no audio stream is rejected",
+			probe:      uploads.Probe{Duration: 4 * time.Minute},
+			wantReason: uploads.ReasonNoAudio,
+			wantInMsg:  []string{"no audio"},
+		},
+		{
+			name:       "a recording over the limit is rejected, stating both lengths",
+			probe:      uploads.Probe{Duration: 102 * time.Minute, HasAudio: true},
+			wantReason: uploads.ReasonTooLong,
+			wantInMsg:  []string{"1 h 42 min", "1 h 30 min"},
+		},
+		{
+			name:       "zero duration is unreadable, not accepted as an empty recording",
+			probe:      uploads.Probe{HasAudio: true},
+			wantReason: uploads.ReasonUnreadable,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := uploads.Validate(tt.probe, testLimits)
+
+			if tt.wantReason == "" {
+				require.NoError(t, err)
+				return
+			}
+			var rej *uploads.RejectError
+			require.True(t, errors.As(err, &rej), "want a *RejectError, got %v", err)
+			assert.Equal(t, tt.wantReason, rej.Reason)
+			for _, s := range tt.wantInMsg {
+				assert.Contains(t, rej.Message, s)
+			}
+		})
+	}
+}
+
+func TestFormatDuration(t *testing.T) {
+	tests := []struct {
+		d    time.Duration
+		want string
+	}{
+		{90 * time.Minute, "1 h 30 min"},
+		{102 * time.Minute, "1 h 42 min"},
+		{45 * time.Minute, "45 min"},
+		{2 * time.Hour, "2 h"},
+		{30 * time.Second, "30 s"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.want, func(t *testing.T) {
+			assert.Equal(t, tt.want, uploads.FormatDuration(tt.d))
+		})
+	}
+}

internal/uploads/worker.go

@@ -0,0 +1,191 @@
+package uploads
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"log/slog"
+	"os"
+	"sync"
+	"time"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+)
+
+// ErrStopped is returned by Submit once the worker has stopped.
+var ErrStopped = errors.New("uploads: worker stopped")
+
+// Converter turns the uploaded file into the 16 kHz mono PCM WAV whisper wants.
+type Converter interface {
+	ToWAV(ctx context.Context, src, dst string) error
+}
+
+// Transcriber turns a WAV into a timed transcription. The gateway chain
+// (iguana primary, berget fallback) implements it.
+type Transcriber interface {
+	Transcribe(ctx context.Context, wavPath string) (domain.Transcription, error)
+}
+
+// WorkerStore is the slice of the private transcript store the worker drives.
+type WorkerStore interface {
+	StartTranscribing(ctx context.Context, userID, id string) error
+	CompleteUpload(ctx context.Context, userID, id string, t domain.Transcription) error
+	FailUpload(ctx context.Context, userID, id string, kind domain.UploadErrorKind) error
+}
+
+// Worker transcribes accepted uploads in-process, ONE AT A TIME (ADR-033). It is
+// not a second deployable: it shares the web process's lifetime, which is why it
+// inherits ADR-018's single-replica assumption — the queue is process-local. The
+// concurrency of one is deliberate: two simultaneous 90-minute jobs on iguana were
+// never measured, and a Metal OOM there latches the whisper backend (infra#273).
+type Worker struct {
+	store   WorkerStore
+	conv    Converter
+	tr      Transcriber
+	tmpDir  string
+	timeout time.Duration
+
+	mu      sync.Mutex
+	queue   []Job
+	stopped bool
+	wake    chan struct{}
+}
+
+// NewWorker wires a Worker. timeout bounds each job (conversion + transcription);
+// a job that outlives it is failed rather than left spinning.
+func NewWorker(s WorkerStore, c Converter, t Transcriber, tmpDir string, timeout time.Duration) *Worker {
+	return &Worker{store: s, conv: c, tr: t, tmpDir: tmpDir, timeout: timeout, wake: make(chan struct{}, 1)}
+}
+
+// Submit queues a job. It never blocks. It satisfies Submitter.
+func (w *Worker) Submit(j Job) error {
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	if w.stopped {
+		return ErrStopped
+	}
+	w.queue = append(w.queue, j)
+	select {
+	case w.wake <- struct{}{}:
+	default:
+	}
+	return nil
+}
+
+// Run processes queued jobs until ctx ends, then deletes the audio of anything
+// still queued and refuses new jobs. The rows of those jobs stay queued or
+// transcribing and are failed as interrupted by the start-up sweep
+// (store.FailInterruptedUploads): the audio is not retained, so they cannot resume.
+func (w *Worker) Run(ctx context.Context) {
+	defer w.stop()
+	for {
+		job, ok := w.next(ctx)
+		if !ok {
+			return
+		}
+		w.Process(ctx, job)
+	}
+}
+
+// next blocks for the next job. Cancellation wins over a non-empty queue: a job
+// still queued at shutdown must not be STARTED (it would begin a long conversion
+// with an already-cancelled context); stop() deletes its audio instead.
+func (w *Worker) next(ctx context.Context) (Job, bool) {
+	for {
+		if ctx.Err() != nil {
+			return Job{}, false
+		}
+		w.mu.Lock()
+		if len(w.queue) > 0 {
+			j := w.queue[0]
+			w.queue = w.queue[1:]
+			w.mu.Unlock()
+			return j, true
+		}
+		w.mu.Unlock()
+		select {
+		case <-w.wake:
+		case <-ctx.Done():
+			return Job{}, false
+		}
+	}
+}
+
+func (w *Worker) stop() {
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	w.stopped = true
+	for _, j := range w.queue {
+		_ = os.Remove(j.Path)
+	}
+	w.queue = nil
+}
+
+// Process runs one job: row -> transcribing, convert, transcribe, store. The
+// uploaded file and the converted WAV are deleted on every path — the audio is
+// never retained (ADR-033). A failure is recorded with a TYPED kind; a shutdown
+// mid-job is not a failure and leaves the row for the start-up sweep.
+func (w *Worker) Process(ctx context.Context, job Job) {
+	defer func() { _ = os.Remove(job.Path) }()
+	log := slog.Default().With("upload", job.UploadID)
+	start := time.Now()
+
+	if err := w.store.StartTranscribing(ctx, job.UserID, job.UploadID); err != nil {
+		log.Error("upload: start transcribing", "err", err)
+		w.fail(ctx, job, log)
+		return
+	}
+
+	wav, err := w.tempWAV()
+	if err != nil {
+		log.Error("upload: temp wav", "err", err)
+		w.fail(ctx, job, log)
+		return
+	}
+	defer func() { _ = os.Remove(wav) }()
+
+	jobCtx, cancel := context.WithTimeout(ctx, w.timeout)
+	defer cancel()
+	if err := w.conv.ToWAV(jobCtx, job.Path, wav); err != nil {
+		log.Error("upload: convert", "err", err)
+		w.fail(ctx, job, log)
+		return
+	}
+	t, err := w.tr.Transcribe(jobCtx, wav)
+	if err != nil {
+		log.Error("upload: transcribe", "err", err, "elapsed", time.Since(start).Round(time.Second))
+		w.fail(ctx, job, log)
+		return
+	}
+	if err := w.store.CompleteUpload(ctx, job.UserID, job.UploadID, t); err != nil {
+		log.Error("upload: store transcript", "err", err)
+		w.fail(ctx, job, log)
+		return
+	}
+	log.Info("upload transcribed", "endpoint", t.Endpoint, "audio_s", t.AudioSeconds,
+		"segments", len(t.Segments), "elapsed", time.Since(start).Round(time.Second))
+}
+
+// fail records a transcription failure — unless the PARENT context ended, which
+// is a shutdown, not a failure: the row is left for the start-up sweep.
+func (w *Worker) fail(ctx context.Context, job Job, log *slog.Logger) {
+	if ctx.Err() != nil {
+		return
+	}
+	if err := w.store.FailUpload(ctx, job.UserID, job.UploadID, domain.UploadErrTranscriptionFailed); err != nil {
+		log.Error("upload: record failure", "err", err)
+	}
+}
+
+func (w *Worker) tempWAV() (string, error) {
+	f, err := os.CreateTemp(w.tmpDir, "wav-*")
+	if err != nil {
+		return "", fmt.Errorf("uploads: temp wav: %w", err)
+	}
+	name := f.Name()
+	if err := f.Close(); err != nil {
+		_ = os.Remove(name)
+		return "", fmt.Errorf("uploads: temp wav: %w", err)
+	}
+	return name, nil
+}

internal/uploads/worker_test.go

@@ -0,0 +1,282 @@
+package uploads_test
+
+import (
+	"context"
+	"errors"
+	"os"
+	"path/filepath"
+	"sync"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+type wStore struct {
+	mu        sync.Mutex
+	started   []string
+	completed map[string]domain.Transcription
+	failed    map[string]domain.UploadErrorKind
+	startErr  error
+}
+
+func newWStore() *wStore {
+	return &wStore{completed: map[string]domain.Transcription{}, failed: map[string]domain.UploadErrorKind{}}
+}
+
+func (s *wStore) StartTranscribing(_ context.Context, _, id string) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.started = append(s.started, id)
+	return s.startErr
+}
+
+func (s *wStore) CompleteUpload(_ context.Context, _, id string, t domain.Transcription) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.completed[id] = t
+	return nil
+}
+
+func (s *wStore) FailUpload(_ context.Context, _, id string, k domain.UploadErrorKind) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.failed[id] = k
+	return nil
+}
+
+func (s *wStore) snapshot() (map[string]domain.Transcription, map[string]domain.UploadErrorKind) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	c := map[string]domain.Transcription{}
+	f := map[string]domain.UploadErrorKind{}
+	for k, v := range s.completed {
+		c[k] = v
+	}
+	for k, v := range s.failed {
+		f[k] = v
+	}
+	return c, f
+}
+
+type fakeConverter struct {
+	err  error
+	srcs []string
+	dsts []string
+}
+
+func (c *fakeConverter) ToWAV(_ context.Context, src, dst string) error {
+	c.srcs = append(c.srcs, src)
+	c.dsts = append(c.dsts, dst)
+	if c.err != nil {
+		return c.err
+	}
+	return os.WriteFile(dst, []byte("wav"), 0o600)
+}
+
+type fakeTranscriber struct {
+	result domain.Transcription
+	err    error
+	paths  []string
+	// block, when non-nil, holds every call until it is closed or ctx ends.
+	block   chan struct{}
+	running atomic.Int32
+	maxSeen atomic.Int32
+}
+
+func (f *fakeTranscriber) Transcribe(ctx context.Context, wav string) (domain.Transcription, error) {
+	f.paths = append(f.paths, wav)
+	n := f.running.Add(1)
+	defer f.running.Add(-1)
+	for {
+		m := f.maxSeen.Load()
+		if n <= m || f.maxSeen.CompareAndSwap(m, n) {
+			break
+		}
+	}
+	if f.block != nil {
+		select {
+		case <-f.block:
+		case <-ctx.Done():
+			return domain.Transcription{}, ctx.Err()
+		}
+	}
+	return f.result, f.err
+}
+
+type wrig struct {
+	w    *uploads.Worker
+	st   *wStore
+	conv *fakeConverter
+	tr   *fakeTranscriber
+	tmp  string
+}
+
+func newWRig(t *testing.T, timeout time.Duration) *wrig {
+	t.Helper()
+	r := &wrig{
+		st:   newWStore(),
+		conv: &fakeConverter{},
+		tr:   &fakeTranscriber{result: domain.Transcription{Text: "synthetic text", Endpoint: "iguana/kb-whisper", AudioSeconds: 3}},
+		tmp:  t.TempDir(),
+	}
+	r.w = uploads.NewWorker(r.st, r.conv, r.tr, r.tmp, timeout)
+	return r
+}
+
+// spool writes a fake uploaded file (what Intake would hand over) and returns a Job.
+func (r *wrig) spool(t *testing.T, id string) uploads.Job {
+	t.Helper()
+	p := filepath.Join(r.tmp, "upload-"+id)
+	require.NoError(t, os.WriteFile(p, []byte("audio"), 0o600))
+	return uploads.Job{UserID: "u1", UploadID: id, Path: p}
+}
+
+func (r *wrig) leftovers(t *testing.T) []string {
+	t.Helper()
+	ents, err := os.ReadDir(r.tmp)
+	require.NoError(t, err)
+	var n []string
+	for _, e := range ents {
+		n = append(n, e.Name())
+	}
+	return n
+}
+
+func TestWorkerProcess_SuccessConvertsTranscribesStoresAndDeletesTheAudio(t *testing.T) {
+	r := newWRig(t, time.Minute)
+	job := r.spool(t, "a")
+
+	r.w.Process(context.Background(), job)
+
+	c, f := r.st.snapshot()
+	assert.Equal(t, []string{"a"}, r.st.started)
+	assert.Equal(t, "synthetic text", c["a"].Text)
+	assert.Equal(t, "iguana/kb-whisper", c["a"].Endpoint)
+	assert.Empty(t, f)
+	require.Len(t, r.conv.srcs, 1)
+	assert.Equal(t, job.Path, r.conv.srcs[0], "the converter reads the uploaded file")
+	assert.Equal(t, r.conv.dsts, r.tr.paths, "whisper is sent the converted WAV, not the original")
+	assert.Equal(t, r.tmp, filepath.Dir(r.conv.dsts[0]))
+	assert.Empty(t, r.leftovers(t), "the audio is never retained: both the upload and the WAV are deleted")
+}
+
+func TestWorkerProcess_FailuresAreTypedAndStillDeleteTheAudio(t *testing.T) {
+	tests := []struct {
+		name  string
+		setup func(*wrig)
+	}{
+		{"conversion fails", func(r *wrig) { r.conv.err = errors.New("moov atom not found") }},
+		{"every transcription endpoint fails", func(r *wrig) { r.tr.err = errors.New("iguana down; berget down") }},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			r := newWRig(t, time.Minute)
+			tt.setup(r)
+			job := r.spool(t, "a")
+
+			r.w.Process(context.Background(), job)
+
+			c, f := r.st.snapshot()
+			assert.Empty(t, c)
+			assert.Equal(t, domain.UploadErrTranscriptionFailed, f["a"])
+			assert.Empty(t, r.leftovers(t))
+		})
+	}
+	t.Run("conversion failure never reaches whisper", func(t *testing.T) {
+		r := newWRig(t, time.Minute)
+		r.conv.err = errors.New("bad")
+
+		r.w.Process(context.Background(), r.spool(t, "a"))
+
+		assert.Empty(t, r.tr.paths)
+	})
+}
+
+func TestWorkerProcess_ShutdownMidJobLeavesTheRowForTheStartupSweep(t *testing.T) {
+	r := newWRig(t, time.Minute)
+	r.tr.block = make(chan struct{})
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan struct{})
+	job := r.spool(t, "a")
+
+	go func() { r.w.Process(ctx, job); close(done) }()
+	require.Eventually(t, func() bool { return r.tr.running.Load() == 1 }, 2*time.Second, 5*time.Millisecond)
+	cancel()
+	<-done
+
+	_, f := r.st.snapshot()
+	assert.Empty(t, f, "a shutdown is not a transcription failure: the row stays for FailInterruptedUploads to mark interrupted")
+	assert.Empty(t, r.leftovers(t), "but the audio is still deleted")
+}
+
+func TestWorkerProcess_AJobThatOutlivesItsTimeoutIsFailedNotStuck(t *testing.T) {
+	r := newWRig(t, 50*time.Millisecond)
+	r.tr.block = make(chan struct{}) // never released
+	done := make(chan struct{})
+
+	go func() { r.w.Process(context.Background(), r.spool(t, "a")); close(done) }()
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("Process did not return: a hung endpoint was not bounded by the job timeout")
+	}
+
+	_, f := r.st.snapshot()
+	assert.Equal(t, domain.UploadErrTranscriptionFailed, f["a"], "a hung endpoint must end in failed, not an eternal spinner (ADR-025)")
+	assert.Empty(t, r.leftovers(t))
+}
+
+func TestWorkerProcess_StartFailureDropsTheJobAndDeletesTheAudio(t *testing.T) {
+	r := newWRig(t, time.Minute)
+	r.st.startErr = errors.New("db down")
+
+	r.w.Process(context.Background(), r.spool(t, "a"))
+
+	assert.Empty(t, r.tr.paths, "never transcribe a job whose row could not be moved to transcribing")
+	_, f := r.st.snapshot()
+	assert.Contains(t, f, "a")
+	assert.Empty(t, r.leftovers(t))
+}
+
+func TestWorkerRun_ProcessesOneJobAtATime(t *testing.T) {
+	r := newWRig(t, time.Minute)
+	r.tr.block = make(chan struct{})
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	go r.w.Run(ctx)
+
+	for _, id := range []string{"a", "b", "c"} {
+		require.NoError(t, r.w.Submit(r.spool(t, id)))
+	}
+	require.Eventually(t, func() bool { return r.tr.running.Load() == 1 }, 2*time.Second, 5*time.Millisecond)
+	time.Sleep(50 * time.Millisecond) // give a (wrong) second job time to start
+	assert.Equal(t, int32(1), r.tr.running.Load(), "two simultaneous long jobs on iguana were never measured: concurrency is 1")
+
+	close(r.tr.block)
+	require.Eventually(t, func() bool { c, _ := r.st.snapshot(); return len(c) == 3 }, 3*time.Second, 5*time.Millisecond)
+	assert.Equal(t, int32(1), r.tr.maxSeen.Load())
+}
+
+func TestWorkerRun_StoppingDeletesQueuedAudioAndRefusesNewJobs(t *testing.T) {
+	r := newWRig(t, time.Minute)
+	r.tr.block = make(chan struct{})
+	ctx, cancel := context.WithCancel(context.Background())
+	stopped := make(chan struct{})
+	go func() { r.w.Run(ctx); close(stopped) }()
+	require.NoError(t, r.w.Submit(r.spool(t, "a")))
+	require.Eventually(t, func() bool { return r.tr.running.Load() == 1 }, 2*time.Second, 5*time.Millisecond)
+	require.NoError(t, r.w.Submit(r.spool(t, "b"))) // queued behind the running job
+
+	cancel()
+	<-stopped
+
+	assert.Len(t, r.conv.srcs, 1, "a job still queued at shutdown must never be started: only the running job was converted")
+	assert.Empty(t, r.leftovers(t), "audio of queued jobs is deleted on stop: the rows are swept to interrupted at next start")
+	assert.ErrorIs(t, r.w.Submit(r.spool(t, "c")), uploads.ErrStopped)
+}

internal/web/handlers.go

@@ -13,6 +13,7 @@ import (
 
 	"git.d-ma.be/mathias/tapir/internal/adapters/store"
 	"git.d-ma.be/mathias/tapir/internal/domain"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
 )
 
 // Store is the read/write surface the web handlers depend on — a narrow port over
@@ -53,6 +54,12 @@ type Store interface {
 	// the most recent discovery pass, shown on the account page as a warning.
 	ListChannelErrors(ctx context.Context, userID string) ([]store.ChannelError, error)
 
+	// Uploaded-audio transcripts (ADR-033): the caller's own PRIVATE uploads,
+	// stored apart from the shared public captions. Both are user-scoped; a
+	// non-owner's id is ErrUploadNotFound.
+	GetUpload(ctx context.Context, userID, id string) (*domain.UploadedTranscript, error)
+	ListUploads(ctx context.Context, userID string) ([]domain.UploadedTranscript, error)
+
 	// SetTranscriptStatus clears or updates a video's transcript backoff state.
 	// Used by handleRetryNow to clear rate_limited_at before immediate processing.
 	SetTranscriptStatus(ctx context.Context, userID, videoID, status string) error
@@ -100,6 +107,12 @@ type App struct {
 	// summary view shows no "dig deeper" affordance. It holds no caption-fetch
 	// dependency, so an enabled chat cannot reach YouTube or the rate gate.
 	Chat Chatter
+	// Uploads, when non-nil, accepts audio recordings for private transcription
+	// (ADR-033). Nil = the /uploads routes are not mounted. UploadLimits are the
+	// same limits the Uploader enforces, shown on the form so they are stated up
+	// front rather than discovered by failing.
+	Uploads      Uploader
+	UploadLimits uploads.Limits
 	// Processing tracks in-flight immediate summarizations so the status endpoint
 	// shows the animation until the summary lands. The zero value is ready to use.
 	Processing ProcessingSet
@@ -166,6 +179,14 @@ func (a *App) Router() http.Handler {
 		app.HandleFunc("GET /v/{videoId}/chat", a.handleChat)
 		app.HandleFunc("POST /v/{videoId}/chat", a.handleChatMessage)
 	}
+	// Audio upload → private transcript (ADR-033). Mounted only when an Uploader is
+	// wired; every handler is user-scoped, so another user's upload is a 404.
+	if a.Uploads != nil {
+		app.HandleFunc("GET /uploads", a.handleUploadsPage)
+		app.HandleFunc("POST /uploads", a.handleUploadPost)
+		app.HandleFunc("GET /uploads/{id}", a.handleUploadDetail)
+		app.HandleFunc("GET /uploads/{id}/status", a.handleUploadStatus)
+	}
 	app.HandleFunc("GET /register", a.handleRegisterForm)
 	app.HandleFunc("POST /register", a.handleRegister)
 

internal/web/paste_test.go

@@ -3,9 +3,10 @@ package web
 import (
 	"bytes"
 	"context"
-	"git.d-ma.be/mathias/tapir/internal/adapters/store"
 	"strings"
 	"testing"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/store"
 )
 
 func TestParseYouTubeVideoID(t *testing.T) {

internal/web/uploads.go

@@ -0,0 +1,191 @@
+package web
+
+import (
+	"context"
+	"errors"
+	"io"
+	"net/http"
+	"path/filepath"
+	"strings"
+	"unicode/utf8"
+
+	"git.d-ma.be/mathias/tapir/internal/adapters/store"
+	"git.d-ma.be/mathias/tapir/internal/domain"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+// Uploader accepts an uploaded recording for transcription (ADR-033).
+// *uploads.Intake satisfies it. It is the whole of what the handlers know about
+// validation, spooling and queueing.
+type Uploader interface {
+	Accept(ctx context.Context, userID, title string, src io.Reader) (id string, err error)
+}
+
+const (
+	// uploadOverhead is headroom above the file-size cap for the multipart framing
+	// and the title field; the cap itself is enforced on the file part by Intake.
+	uploadOverhead = 1 << 20
+	maxTitleRunes  = 200
+	// uploadPollEvery is how often an unfinished upload re-asks for its status.
+	uploadPollEvery = "5s"
+)
+
+// uploadsView is what the /uploads page renders.
+type uploadsView struct {
+	Limits uploads.Limits
+	Rows   []domain.UploadedTranscript
+	Error  string // a synchronous rejection message, empty when none
+}
+
+// handleUploadsPage renders the upload form and the caller's own uploads.
+func (a *App) handleUploadsPage(w http.ResponseWriter, r *http.Request) {
+	userID, ok := a.currentUserID(w, r)
+	if !ok {
+		return
+	}
+	a.renderUploadsPage(w, r, userID, http.StatusOK, "")
+}
+
+func (a *App) renderUploadsPage(w http.ResponseWriter, r *http.Request, userID string, status int, msg string) {
+	rows, err := a.Store.ListUploads(r.Context(), userID)
+	if err != nil {
+		a.serverError(w, r, "list uploads", err)
+		return
+	}
+	a.renderStatus(w, r, status, UploadsPage(uploadsView{Limits: a.UploadLimits, Rows: rows, Error: msg}))
+}
+
+// handleUploadPost streams the multipart body: the optional title field, then the
+// file part, which goes straight to Intake without being buffered in memory. A
+// rejection is synchronous — a status code and a message on the upload page, no
+// row, no timeout (ADR-033).
+func (a *App) handleUploadPost(w http.ResponseWriter, r *http.Request) {
+	userID, ok := a.currentUserID(w, r)
+	if !ok {
+		return
+	}
+	r.Body = http.MaxBytesReader(w, r.Body, a.UploadLimits.MaxBytes+uploadOverhead)
+	mr, err := r.MultipartReader()
+	if err != nil {
+		a.renderUploadsPage(w, r, userID, http.StatusBadRequest, "Choose an audio file to upload.")
+		return
+	}
+	title := ""
+	for {
+		part, err := mr.NextPart()
+		if errors.Is(err, io.EOF) {
+			break
+		}
+		if err != nil {
+			a.uploadFailure(w, r, userID, err)
+			return
+		}
+		switch part.FormName() {
+		case "title":
+			b, _ := io.ReadAll(io.LimitReader(part, maxTitleRunes*utf8.UTFMax))
+			title = clipRunes(strings.TrimSpace(string(b)), maxTitleRunes)
+		case "file":
+			if title == "" {
+				title = titleFromFilename(part.FileName())
+			}
+			id, err := a.Uploads.Accept(r.Context(), userID, title, part)
+			if err != nil {
+				a.uploadFailure(w, r, userID, err)
+				return
+			}
+			http.Redirect(w, r, "/uploads/"+id, http.StatusSeeOther)
+			return
+		}
+	}
+	a.renderUploadsPage(w, r, userID, http.StatusBadRequest, "Choose an audio file to upload.")
+}
+
+// uploadFailure maps an Accept (or multipart) error to a response: a validation
+// rejection is the user's to fix (422, or 413 for size) and carries its own
+// message; anything else is ours.
+func (a *App) uploadFailure(w http.ResponseWriter, r *http.Request, userID string, err error) {
+	var rej *uploads.RejectError
+	var tooBig *http.MaxBytesError
+	switch {
+	case errors.As(err, &rej):
+		status := http.StatusUnprocessableEntity
+		if rej.Reason == uploads.ReasonTooLarge {
+			status = http.StatusRequestEntityTooLarge
+		}
+		a.renderUploadsPage(w, r, userID, status, rej.Message)
+	case errors.As(err, &tooBig):
+		a.renderUploadsPage(w, r, userID, http.StatusRequestEntityTooLarge,
+			"This file is larger than the upload limit.")
+	default:
+		a.serverError(w, r, "upload", err)
+	}
+}
+
+// handleUploadDetail renders one upload's page (GET /uploads/{id}). The row is
+// read through the user-scoped store, so another user's id is a 404 — existence is
+// not disclosed, and the title never appears in the response.
+func (a *App) handleUploadDetail(w http.ResponseWriter, r *http.Request) {
+	u, ok := a.loadOwnedUpload(w, r)
+	if !ok {
+		return
+	}
+	a.render(w, r, UploadPage(*u))
+}
+
+// handleUploadStatus renders the status fragment HTMX polls. Unfinished states
+// keep polling; terminal states render without hx-trigger so the page stops
+// asking (ADR-025: never a stuck spinner).
+func (a *App) handleUploadStatus(w http.ResponseWriter, r *http.Request) {
+	u, ok := a.loadOwnedUpload(w, r)
+	if !ok {
+		return
+	}
+	a.render(w, r, uploadStatus(*u))
+}
+
+func (a *App) loadOwnedUpload(w http.ResponseWriter, r *http.Request) (*domain.UploadedTranscript, bool) {
+	userID, ok := a.currentUserID(w, r)
+	if !ok {
+		return nil, false
+	}
+	u, err := a.Store.GetUpload(r.Context(), userID, r.PathValue("id"))
+	if errors.Is(err, store.ErrUploadNotFound) {
+		http.NotFound(w, r)
+		return nil, false
+	}
+	if err != nil {
+		a.serverError(w, r, "get upload", err)
+		return nil, false
+	}
+	return u, true
+}
+
+// titleFromFilename is the default title: the file's name without its extension.
+func titleFromFilename(name string) string {
+	base := filepath.Base(name)
+	t := strings.TrimSuffix(base, filepath.Ext(base))
+	if t == "" || t == "." {
+		return "Untitled recording"
+	}
+	return clipRunes(t, maxTitleRunes)
+}
+
+func clipRunes(s string, n int) string {
+	if utf8.RuneCountInString(s) <= n {
+		return s
+	}
+	return string([]rune(s)[:n])
+}
+
+// uploadFailureMessage turns a typed failure kind into words for the user. It is
+// keyed on OUR kinds, never on a provider's error string (ADR-033).
+func uploadFailureMessage(k domain.UploadErrorKind) string {
+	switch k {
+	case domain.UploadErrInterrupted:
+		return "The server restarted before this recording was transcribed. Please upload it again."
+	case domain.UploadErrTranscriptionFailed:
+		return "Transcription failed on every available service. Please upload it again later."
+	default:
+		return "This upload failed. Please upload it again."
+	}
+}

internal/web/uploads.templ

@@ -0,0 +1,91 @@
+package web
+
+import (
+	"fmt"
+	"time"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+// UploadsPage is the upload form plus the caller's own uploads (ADR-033).
+templ UploadsPage(v uploadsView) {
+	@Layout("Tapir — Upload a recording") {
+		<article class="account">
+			<h1>Upload a recording</h1>
+			<p class="muted">
+				Audio only. Up to { fmt.Sprintf("%d MB", v.Limits.MaxBytes>>20) } and
+				{ uploads.FormatDuration(v.Limits.MaxDuration) }. The transcript is private to you;
+				the audio is deleted once it has been transcribed.
+			</p>
+			if v.Error != "" {
+				<div class="flash flash-error" role="alert">{ v.Error }</div>
+			}
+			<form method="post" action="/uploads" enctype="multipart/form-data">
+				<p><label>Title (optional) <input type="text" name="title" maxlength="200"/></label></p>
+				<p><input type="file" name="file" accept="audio/*,.m4a,.mp3,.wav,.opus,.ogg,.mp4" required/></p>
+				<p><button type="submit">Upload and transcribe</button></p>
+			</form>
+			if len(v.Rows) > 0 {
+				<section>
+					<h2>Your recordings</h2>
+					<ul>
+						for _, u := range v.Rows {
+							<li>
+								<a href={ templ.SafeURL("/uploads/" + u.ID) }>{ u.Title }</a>
+								<span class="chip">{ string(u.Status) }</span>
+							</li>
+						}
+					</ul>
+				</section>
+			}
+		</article>
+	}
+}
+
+// UploadPage is one upload's page: its title and the polling status region.
+templ UploadPage(u domain.UploadedTranscript) {
+	@Layout("Tapir — " + u.Title) {
+		<article class="account">
+			<h1>{ u.Title }</h1>
+			@uploadStatus(u)
+			<p><a href="/uploads">All recordings</a></p>
+		</article>
+	}
+}
+
+// uploadStatus is the region HTMX polls. Unfinished states carry hx-trigger so the
+// page re-asks; done and failed render without it, so nothing polls forever.
+templ uploadStatus(u domain.UploadedTranscript) {
+	switch u.Status {
+		case domain.UploadDone:
+			<div class="upload-status upload-done" role="status">
+				<p>
+					Transcribed { uploads.FormatDuration(time.Duration(u.AudioSeconds * float64(time.Second))) }
+					of audio by <span class="chip">{ u.Endpoint }</span>.
+				</p>
+				<details open>
+					<summary>Transcript</summary>
+					<pre class="transcript">{ u.Content }</pre>
+				</details>
+			</div>
+		case domain.UploadFailed:
+			<div class="upload-status upload-failed flash flash-error" role="alert">
+				<p>{ uploadFailureMessage(u.ErrorKind) }</p>
+			</div>
+		default:
+			<div
+				class="upload-status upload-working"
+				role="status"
+				hx-get={ "/uploads/" + u.ID + "/status" }
+				hx-trigger={ "every " + uploadPollEvery }
+				hx-swap="outerHTML"
+			>
+				if u.Status == domain.UploadTranscribing {
+					<p>Transcribing… a long recording takes a few minutes.</p>
+				} else {
+					<p>Queued — waiting for its turn to be transcribed.</p>
+				}
+			</div>
+	}
+}

internal/web/uploads_handler_test.go

@@ -0,0 +1,277 @@
+package web_test
+
+import (
+	"bytes"
+	"context"
+	"mime/multipart"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+	"git.d-ma.be/mathias/tapir/internal/web"
+)
+
+// Upload route tests drive the real Router, the real Intake and the real Postgres
+// store; only the process boundary (ffprobe) and the transcription worker are
+// faked. All audio bytes and transcript text are synthetic — this repo is public
+// and no real recording or transcript may appear in it (#33).
+
+const otherUserID = "22222222-2222-2222-2222-222222222222"
+
+type uploadProber struct{ probe uploads.Probe }
+
+func (p uploadProber) Probe(context.Context, string) (uploads.Probe, error) { return p.probe, nil }
+
+type uploadSubmitter struct{ jobs []uploads.Job }
+
+func (s *uploadSubmitter) Submit(j uploads.Job) error { s.jobs = append(s.jobs, j); return nil }
+
+func newUploadApp(t *testing.T, probe uploads.Probe, maxBytes int64) (*web.App, *uploadSubmitter) {
+	t.Helper()
+	app := newApp(t)
+	resetDB(t, rawPool(t))
+	sub := &uploadSubmitter{}
+	limits := uploads.Limits{MaxBytes: maxBytes, MaxDuration: 90 * time.Minute}
+	s := newStore(t)
+	app.Uploads = uploads.NewIntake(s, uploadProber{probe: probe}, sub, limits, t.TempDir())
+	app.UploadLimits = limits
+	return app, sub
+}
+
+var audioOK = uploads.Probe{Duration: 45 * time.Minute, HasAudio: true}
+
+func uploadReq(t *testing.T, title, filename string, content []byte) *http.Request {
+	t.Helper()
+	var buf bytes.Buffer
+	mw := multipart.NewWriter(&buf)
+	if title != "" {
+		require.NoError(t, mw.WriteField("title", title))
+	}
+	if filename != "" {
+		fw, err := mw.CreateFormFile("file", filename)
+		require.NoError(t, err)
+		_, err = fw.Write(content)
+		require.NoError(t, err)
+	}
+	require.NoError(t, mw.Close())
+	req := httptest.NewRequest(http.MethodPost, "/uploads", &buf)
+	req.Header.Set("Content-Type", mw.FormDataContentType())
+	return req
+}
+
+func seedUpload(t *testing.T, owner, title, status, content string) string {
+	t.Helper()
+	p := rawPool(t)
+	_, err := p.Exec(context.Background(), `INSERT INTO users (id) VALUES ($1) ON CONFLICT DO NOTHING`, owner)
+	require.NoError(t, err)
+	var id string
+	switch status {
+	case "done":
+		require.NoError(t, p.QueryRow(context.Background(),
+			`INSERT INTO uploaded_transcripts (user_id, title, status, content, endpoint, audio_seconds, finished_at)
+			 VALUES ($1, $2, 'done', $3, 'berget/kb-whisper', 2700, NOW()) RETURNING id`,
+			owner, title, content).Scan(&id))
+	case "failed":
+		require.NoError(t, p.QueryRow(context.Background(),
+			`INSERT INTO uploaded_transcripts (user_id, title, status, error_kind, finished_at)
+			 VALUES ($1, $2, 'failed', $3, NOW()) RETURNING id`,
+			owner, title, content).Scan(&id)) // content carries the error_kind here
+	default:
+		require.NoError(t, p.QueryRow(context.Background(),
+			`INSERT INTO uploaded_transcripts (user_id, title, status) VALUES ($1, $2, $3) RETURNING id`,
+			owner, title, status).Scan(&id))
+	}
+	return id
+}
+
+func get(path string) *http.Request { return httptest.NewRequest(http.MethodGet, path, nil) }
+
+func TestUploadsPage_ShowsAMultipartFormWithAFileField(t *testing.T) {
+	app, _ := newUploadApp(t, audioOK, 200<<20)
+
+	rec := do(t, app, get("/uploads"))
+
+	require.Equal(t, http.StatusOK, rec.Code)
+	b := body(t, rec)
+	assert.Contains(t, b, `enctype="multipart/form-data"`)
+	assert.Contains(t, b, `name="file"`, "the multipart field is `file`")
+	assert.Contains(t, b, "200 MB", "the size limit is stated up front, not discovered by failing")
+	assert.Contains(t, b, "1 h 30 min", "the length limit is stated up front")
+}
+
+func TestUploadsPage_ListsOnlyTheCallersUploads(t *testing.T) {
+	app, _ := newUploadApp(t, audioOK, 200<<20)
+	seedUpload(t, userID, "My weekly sync", "queued", "")
+	seedUpload(t, otherUserID, "Someone else's board meeting", "queued", "")
+
+	b := body(t, do(t, app, get("/uploads")))
+
+	assert.Contains(t, b, "My weekly sync")
+	assert.NotContains(t, b, "Someone else's board meeting")
+}
+
+func TestUploadPost_AcceptedAudioQueuesARowAndRedirectsToIt(t *testing.T) {
+	app, sub := newUploadApp(t, audioOK, 200<<20)
+
+	rec := do(t, app, uploadReq(t, "Sprint review", "recording.m4a", []byte("synthetic audio bytes")))
+
+	require.Equal(t, http.StatusSeeOther, rec.Code)
+	var id, status, title string
+	require.NoError(t, rawPool(t).QueryRow(context.Background(),
+		`SELECT id, status, title FROM uploaded_transcripts WHERE user_id = $1`, userID).Scan(&id, &status, &title))
+	assert.Equal(t, "queued", status)
+	assert.Equal(t, "Sprint review", title)
+	assert.Equal(t, "/uploads/"+id, rec.Header().Get("Location"))
+	require.Len(t, sub.jobs, 1)
+	assert.Equal(t, uploads.Job{UserID: userID, UploadID: id, Path: sub.jobs[0].Path}, sub.jobs[0])
+	spooled, err := os.ReadFile(sub.jobs[0].Path)
+	require.NoError(t, err)
+	assert.Equal(t, "synthetic audio bytes", string(spooled))
+}
+
+func TestUploadPost_WithoutATitleUsesTheFilenameMinusItsExtension(t *testing.T) {
+	app, _ := newUploadApp(t, audioOK, 200<<20)
+
+	rec := do(t, app, uploadReq(t, "", "Teams meeting 2026-09-30.mp4", []byte("x")))
+
+	require.Equal(t, http.StatusSeeOther, rec.Code)
+	var title string
+	require.NoError(t, rawPool(t).QueryRow(context.Background(),
+		`SELECT title FROM uploaded_transcripts WHERE user_id = $1`, userID).Scan(&title))
+	assert.Equal(t, "Teams meeting 2026-09-30", title)
+}
+
+func TestUploadPost_RejectionsAreSynchronousWithAMessageAndNoRow(t *testing.T) {
+	tests := []struct {
+		name       string
+		probe      uploads.Probe
+		maxBytes   int64
+		content    string
+		wantStatus int
+		wantInBody []string
+	}{
+		{
+			name:       "video names issue 28",
+			probe:      uploads.Probe{Duration: time.Minute, HasAudio: true, HasVideo: true},
+			maxBytes:   1 << 20,
+			content:    "x",
+			wantStatus: http.StatusUnprocessableEntity,
+			wantInBody: []string{"video", "#28"},
+		},
+		{
+			name:       "too long states both lengths",
+			probe:      uploads.Probe{Duration: 2 * time.Hour, HasAudio: true},
+			maxBytes:   1 << 20,
+			content:    "x",
+			wantStatus: http.StatusUnprocessableEntity,
+			wantInBody: []string{"2 h", "1 h 30 min"},
+		},
+		{
+			name:       "too large is 413",
+			probe:      audioOK,
+			maxBytes:   4,
+			content:    "more than four bytes",
+			wantStatus: http.StatusRequestEntityTooLarge,
+			wantInBody: []string{"limit"},
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			app, sub := newUploadApp(t, tt.probe, tt.maxBytes)
+
+			rec := do(t, app, uploadReq(t, "t", "r.m4a", []byte(tt.content)))
+
+			assert.Equal(t, tt.wantStatus, rec.Code)
+			b := body(t, rec)
+			for _, s := range tt.wantInBody {
+				assert.Contains(t, b, s)
+			}
+			var n int
+			require.NoError(t, rawPool(t).QueryRow(context.Background(),
+				`SELECT count(*) FROM uploaded_transcripts`).Scan(&n))
+			assert.Zero(t, n, "a rejected upload leaves no row")
+			assert.Empty(t, sub.jobs)
+		})
+	}
+}
+
+func TestUploadPost_WithNoFilePartIs400(t *testing.T) {
+	app, _ := newUploadApp(t, audioOK, 200<<20)
+
+	rec := do(t, app, uploadReq(t, "only a title", "", nil))
+
+	assert.Equal(t, http.StatusBadRequest, rec.Code)
+	assert.Contains(t, body(t, rec), "Choose an audio file")
+}
+
+func TestUploadRoutes_AreNotMountedWithoutAnUploader(t *testing.T) {
+	app := newApp(t)
+	resetDB(t, rawPool(t))
+
+	assert.Equal(t, http.StatusNotFound, do(t, app, get("/uploads")).Code)
+}
+
+func TestUploadPage_OwnerSeesTitleStatusAndNonOwnerGetsNotFound(t *testing.T) {
+	app, _ := newUploadApp(t, audioOK, 200<<20)
+	mine := seedUpload(t, userID, "Roadmap sync", "queued", "")
+	theirs := seedUpload(t, otherUserID, "Their private meeting", "queued", "")
+
+	own := do(t, app, get("/uploads/"+mine))
+	require.Equal(t, http.StatusOK, own.Code)
+	assert.Contains(t, body(t, own), "Roadmap sync")
+
+	for _, path := range []string{"/uploads/" + theirs, "/uploads/" + theirs + "/status"} {
+		rec := do(t, app, get(path))
+		assert.Equal(t, http.StatusNotFound, rec.Code, path)
+		assert.NotContains(t, body(t, rec), "Their private meeting", "a 404 must not leak the title")
+	}
+}
+
+func TestUploadStatus_UnfinishedStatesKeepPollingTerminalStatesStop(t *testing.T) {
+	app, _ := newUploadApp(t, audioOK, 200<<20)
+	tests := []struct {
+		name, status, content string
+		polls                 bool
+		wantInBody            string
+	}{
+		{"queued", "queued", "", true, "Queued"},
+		{"transcribing", "transcribing", "", true, "ranscrib"},
+		{"done", "done", "Synthetic transcript text.", false, "Synthetic transcript text."},
+		{"failed interrupted", "failed", "interrupted", false, "upload it again"},
+		{"failed transcription", "failed", "transcription_failed", false, "Transcription failed"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			id := seedUpload(t, userID, "m-"+tt.name, tt.status, tt.content)
+
+			rec := do(t, app, get("/uploads/"+id+"/status"))
+
+			require.Equal(t, http.StatusOK, rec.Code)
+			b := body(t, rec)
+			assert.Contains(t, b, tt.wantInBody)
+			if tt.polls {
+				assert.Contains(t, b, `hx-trigger="every`, "an unfinished upload must keep polling")
+			} else {
+				assert.NotContains(t, b, "hx-trigger", "a finished upload must not poll forever: no stuck spinner")
+			}
+		})
+	}
+}
+
+func TestUploadStatus_DoneNamesTheEndpointThatServedIt(t *testing.T) {
+	app, _ := newUploadApp(t, audioOK, 200<<20)
+	id := seedUpload(t, userID, "m", "done", "Synthetic transcript text.")
+
+	b := body(t, do(t, app, get("/uploads/"+id+"/status")))
+
+	assert.Contains(t, b, "berget/kb-whisper",
+		"which backend served a transcript is a quality fact, and berget means the audio left the homelab")
+	assert.True(t, strings.Contains(b, "45 min"), "the audio length is shown")
+}

internal/web/uploads_templ.go

@@ -0,0 +1,381 @@
+// Code generated by templ - DO NOT EDIT.
+
+// templ: version: v0.3.1020
+package web
+
+//lint:file-ignore SA4006 This context is only used if a nested component is present.
+
+import "github.com/a-h/templ"
+import templruntime "github.com/a-h/templ/runtime"
+
+import (
+	"fmt"
+	"time"
+
+	"git.d-ma.be/mathias/tapir/internal/domain"
+	"git.d-ma.be/mathias/tapir/internal/uploads"
+)
+
+// UploadsPage is the upload form plus the caller's own uploads (ADR-033).
+func UploadsPage(v uploadsView) templ.Component {
+	return templruntime.GeneratedTemplate(func(templ_7745c5c3_Input templruntime.GeneratedComponentInput) (templ_7745c5c3_Err error) {
+		templ_7745c5c3_W, ctx := templ_7745c5c3_Input.Writer, templ_7745c5c3_Input.Context
+		if templ_7745c5c3_CtxErr := ctx.Err(); templ_7745c5c3_CtxErr != nil {
+			return templ_7745c5c3_CtxErr
+		}
+		templ_7745c5c3_Buffer, templ_7745c5c3_IsBuffer := templruntime.GetBuffer(templ_7745c5c3_W)
+		if !templ_7745c5c3_IsBuffer {
+			defer func() {
+				templ_7745c5c3_BufErr := templruntime.ReleaseBuffer(templ_7745c5c3_Buffer)
+				if templ_7745c5c3_Err == nil {
+					templ_7745c5c3_Err = templ_7745c5c3_BufErr
+				}
+			}()
+		}
+		ctx = templ.InitializeContext(ctx)
+		templ_7745c5c3_Var1 := templ.GetChildren(ctx)
+		if templ_7745c5c3_Var1 == nil {
+			templ_7745c5c3_Var1 = templ.NopComponent
+		}
+		ctx = templ.ClearChildren(ctx)
+		templ_7745c5c3_Var2 := templruntime.GeneratedTemplate(func(templ_7745c5c3_Input templruntime.GeneratedComponentInput) (templ_7745c5c3_Err error) {
+			templ_7745c5c3_W, ctx := templ_7745c5c3_Input.Writer, templ_7745c5c3_Input.Context
+			templ_7745c5c3_Buffer, templ_7745c5c3_IsBuffer := templruntime.GetBuffer(templ_7745c5c3_W)
+			if !templ_7745c5c3_IsBuffer {
+				defer func() {
+					templ_7745c5c3_BufErr := templruntime.ReleaseBuffer(templ_7745c5c3_Buffer)
+					if templ_7745c5c3_Err == nil {
+						templ_7745c5c3_Err = templ_7745c5c3_BufErr
+					}
+				}()
+			}
+			ctx = templ.InitializeContext(ctx)
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 1, "<article class=\"account\"><h1>Upload a recording</h1><p class=\"muted\">Audio only. Up to ")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var3 string
+			templ_7745c5c3_Var3, templ_7745c5c3_Err = templ.JoinStringErrs(fmt.Sprintf("%d MB", v.Limits.MaxBytes>>20))
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 17, Col: 67}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var3))
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 2, " and ")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var4 string
+			templ_7745c5c3_Var4, templ_7745c5c3_Err = templ.JoinStringErrs(uploads.FormatDuration(v.Limits.MaxDuration))
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 18, Col: 50}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var4))
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 3, ". The transcript is private to you; the audio is deleted once it has been transcribed.</p>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			if v.Error != "" {
+				templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 4, "<div class=\"flash flash-error\" role=\"alert\">")
+				if templ_7745c5c3_Err != nil {
+					return templ_7745c5c3_Err
+				}
+				var templ_7745c5c3_Var5 string
+				templ_7745c5c3_Var5, templ_7745c5c3_Err = templ.JoinStringErrs(v.Error)
+				if templ_7745c5c3_Err != nil {
+					return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 22, Col: 57}
+				}
+				_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var5))
+				if templ_7745c5c3_Err != nil {
+					return templ_7745c5c3_Err
+				}
+				templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 5, "</div>")
+				if templ_7745c5c3_Err != nil {
+					return templ_7745c5c3_Err
+				}
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 6, "<form method=\"post\" action=\"/uploads\" enctype=\"multipart/form-data\"><p><label>Title (optional) <input type=\"text\" name=\"title\" maxlength=\"200\"></label></p><p><input type=\"file\" name=\"file\" accept=\"audio/*,.m4a,.mp3,.wav,.opus,.ogg,.mp4\" required></p><p><button type=\"submit\">Upload and transcribe</button></p></form>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			if len(v.Rows) > 0 {
+				templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 7, "<section><h2>Your recordings</h2><ul>")
+				if templ_7745c5c3_Err != nil {
+					return templ_7745c5c3_Err
+				}
+				for _, u := range v.Rows {
+					templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 8, "<li><a href=\"")
+					if templ_7745c5c3_Err != nil {
+						return templ_7745c5c3_Err
+					}
+					var templ_7745c5c3_Var6 templ.SafeURL
+					templ_7745c5c3_Var6, templ_7745c5c3_Err = templ.JoinURLErrs(templ.SafeURL("/uploads/" + u.ID))
+					if templ_7745c5c3_Err != nil {
+						return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 35, Col: 51}
+					}
+					_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var6))
+					if templ_7745c5c3_Err != nil {
+						return templ_7745c5c3_Err
+					}
+					templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 9, "\">")
+					if templ_7745c5c3_Err != nil {
+						return templ_7745c5c3_Err
+					}
+					var templ_7745c5c3_Var7 string
+					templ_7745c5c3_Var7, templ_7745c5c3_Err = templ.JoinStringErrs(u.Title)
+					if templ_7745c5c3_Err != nil {
+						return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 35, Col: 63}
+					}
+					_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var7))
+					if templ_7745c5c3_Err != nil {
+						return templ_7745c5c3_Err
+					}
+					templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 10, "</a> <span class=\"chip\">")
+					if templ_7745c5c3_Err != nil {
+						return templ_7745c5c3_Err
+					}
+					var templ_7745c5c3_Var8 string
+					templ_7745c5c3_Var8, templ_7745c5c3_Err = templ.JoinStringErrs(string(u.Status))
+					if templ_7745c5c3_Err != nil {
+						return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 36, Col: 45}
+					}
+					_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var8))
+					if templ_7745c5c3_Err != nil {
+						return templ_7745c5c3_Err
+					}
+					templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 11, "</span></li>")
+					if templ_7745c5c3_Err != nil {
+						return templ_7745c5c3_Err
+					}
+				}
+				templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 12, "</ul></section>")
+				if templ_7745c5c3_Err != nil {
+					return templ_7745c5c3_Err
+				}
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 13, "</article>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			return nil
+		})
+		templ_7745c5c3_Err = Layout("Tapir — Upload a recording").Render(templ.WithChildren(ctx, templ_7745c5c3_Var2), templ_7745c5c3_Buffer)
+		if templ_7745c5c3_Err != nil {
+			return templ_7745c5c3_Err
+		}
+		return nil
+	})
+}
+
+// UploadPage is one upload's page: its title and the polling status region.
+func UploadPage(u domain.UploadedTranscript) templ.Component {
+	return templruntime.GeneratedTemplate(func(templ_7745c5c3_Input templruntime.GeneratedComponentInput) (templ_7745c5c3_Err error) {
+		templ_7745c5c3_W, ctx := templ_7745c5c3_Input.Writer, templ_7745c5c3_Input.Context
+		if templ_7745c5c3_CtxErr := ctx.Err(); templ_7745c5c3_CtxErr != nil {
+			return templ_7745c5c3_CtxErr
+		}
+		templ_7745c5c3_Buffer, templ_7745c5c3_IsBuffer := templruntime.GetBuffer(templ_7745c5c3_W)
+		if !templ_7745c5c3_IsBuffer {
+			defer func() {
+				templ_7745c5c3_BufErr := templruntime.ReleaseBuffer(templ_7745c5c3_Buffer)
+				if templ_7745c5c3_Err == nil {
+					templ_7745c5c3_Err = templ_7745c5c3_BufErr
+				}
+			}()
+		}
+		ctx = templ.InitializeContext(ctx)
+		templ_7745c5c3_Var9 := templ.GetChildren(ctx)
+		if templ_7745c5c3_Var9 == nil {
+			templ_7745c5c3_Var9 = templ.NopComponent
+		}
+		ctx = templ.ClearChildren(ctx)
+		templ_7745c5c3_Var10 := templruntime.GeneratedTemplate(func(templ_7745c5c3_Input templruntime.GeneratedComponentInput) (templ_7745c5c3_Err error) {
+			templ_7745c5c3_W, ctx := templ_7745c5c3_Input.Writer, templ_7745c5c3_Input.Context
+			templ_7745c5c3_Buffer, templ_7745c5c3_IsBuffer := templruntime.GetBuffer(templ_7745c5c3_W)
+			if !templ_7745c5c3_IsBuffer {
+				defer func() {
+					templ_7745c5c3_BufErr := templruntime.ReleaseBuffer(templ_7745c5c3_Buffer)
+					if templ_7745c5c3_Err == nil {
+						templ_7745c5c3_Err = templ_7745c5c3_BufErr
+					}
+				}()
+			}
+			ctx = templ.InitializeContext(ctx)
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 14, "<article class=\"account\"><h1>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var11 string
+			templ_7745c5c3_Var11, templ_7745c5c3_Err = templ.JoinStringErrs(u.Title)
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 50, Col: 16}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var11))
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 15, "</h1>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = uploadStatus(u).Render(ctx, templ_7745c5c3_Buffer)
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 16, "<p><a href=\"/uploads\">All recordings</a></p></article>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			return nil
+		})
+		templ_7745c5c3_Err = Layout("Tapir — "+u.Title).Render(templ.WithChildren(ctx, templ_7745c5c3_Var10), templ_7745c5c3_Buffer)
+		if templ_7745c5c3_Err != nil {
+			return templ_7745c5c3_Err
+		}
+		return nil
+	})
+}
+
+// uploadStatus is the region HTMX polls. Unfinished states carry hx-trigger so the
+// page re-asks; done and failed render without it, so nothing polls forever.
+func uploadStatus(u domain.UploadedTranscript) templ.Component {
+	return templruntime.GeneratedTemplate(func(templ_7745c5c3_Input templruntime.GeneratedComponentInput) (templ_7745c5c3_Err error) {
+		templ_7745c5c3_W, ctx := templ_7745c5c3_Input.Writer, templ_7745c5c3_Input.Context
+		if templ_7745c5c3_CtxErr := ctx.Err(); templ_7745c5c3_CtxErr != nil {
+			return templ_7745c5c3_CtxErr
+		}
+		templ_7745c5c3_Buffer, templ_7745c5c3_IsBuffer := templruntime.GetBuffer(templ_7745c5c3_W)
+		if !templ_7745c5c3_IsBuffer {
+			defer func() {
+				templ_7745c5c3_BufErr := templruntime.ReleaseBuffer(templ_7745c5c3_Buffer)
+				if templ_7745c5c3_Err == nil {
+					templ_7745c5c3_Err = templ_7745c5c3_BufErr
+				}
+			}()
+		}
+		ctx = templ.InitializeContext(ctx)
+		templ_7745c5c3_Var12 := templ.GetChildren(ctx)
+		if templ_7745c5c3_Var12 == nil {
+			templ_7745c5c3_Var12 = templ.NopComponent
+		}
+		ctx = templ.ClearChildren(ctx)
+		switch u.Status {
+		case domain.UploadDone:
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 17, "<div class=\"upload-status upload-done\" role=\"status\"><p>Transcribed ")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var13 string
+			templ_7745c5c3_Var13, templ_7745c5c3_Err = templ.JoinStringErrs(uploads.FormatDuration(time.Duration(u.AudioSeconds * float64(time.Second))))
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 64, Col: 95}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var13))
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 18, " of audio by <span class=\"chip\">")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var14 string
+			templ_7745c5c3_Var14, templ_7745c5c3_Err = templ.JoinStringErrs(u.Endpoint)
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 65, Col: 48}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var14))
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 19, "</span>.</p><details open><summary>Transcript</summary><pre class=\"transcript\">")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var15 string
+			templ_7745c5c3_Var15, templ_7745c5c3_Err = templ.JoinStringErrs(u.Content)
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 69, Col: 40}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var15))
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 20, "</pre></details></div>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+		case domain.UploadFailed:
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 21, "<div class=\"upload-status upload-failed flash flash-error\" role=\"alert\"><p>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var16 string
+			templ_7745c5c3_Var16, templ_7745c5c3_Err = templ.JoinStringErrs(uploadFailureMessage(u.ErrorKind))
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 74, Col: 42}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ.EscapeString(templ_7745c5c3_Var16))
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 22, "</p></div>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+		default:
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 23, "<div class=\"upload-status upload-working\" role=\"status\" hx-get=\"")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var17 string
+			templ_7745c5c3_Var17, templ_7745c5c3_Err = templ.ResolveAttributeValue("/uploads/" + u.ID + "/status")
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 80, Col: 43}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ_7745c5c3_Var17)
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 24, "\" hx-trigger=\"")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			var templ_7745c5c3_Var18 string
+			templ_7745c5c3_Var18, templ_7745c5c3_Err = templ.ResolveAttributeValue("every " + uploadPollEvery)
+			if templ_7745c5c3_Err != nil {
+				return templ.Error{Err: templ_7745c5c3_Err, FileName: `internal/web/uploads.templ`, Line: 81, Col: 43}
+			}
+			_, templ_7745c5c3_Err = templ_7745c5c3_Buffer.WriteString(templ_7745c5c3_Var18)
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 25, "\" hx-swap=\"outerHTML\">")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+			if u.Status == domain.UploadTranscribing {
+				templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 26, "<p>Transcribing… a long recording takes a few minutes.</p>")
+				if templ_7745c5c3_Err != nil {
+					return templ_7745c5c3_Err
+				}
+			} else {
+				templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 27, "<p>Queued — waiting for its turn to be transcribed.</p>")
+				if templ_7745c5c3_Err != nil {
+					return templ_7745c5c3_Err
+				}
+			}
+			templ_7745c5c3_Err = templruntime.WriteString(templ_7745c5c3_Buffer, 28, "</div>")
+			if templ_7745c5c3_Err != nil {
+				return templ_7745c5c3_Err
+			}
+		}
+		return nil
+	})
+}
+
+var _ = templruntime.GeneratedTemplate

internal/web/views.templ

@@ -4,7 +4,7 @@ import (
 	"fmt"
 	"strings"
 
-	"gitea.d-ma.be/mathias/tapir/internal/adapters/store"
+	"git.d-ma.be/mathias/tapir/internal/adapters/store"
 )
 
 // Layout is the shared HTML shell. HTMX drives the progressive interactions

internal/web/visual_theme_test.go

@@ -14,7 +14,8 @@ import (
 // page renders any GET path's full HTML (helper for theme/structure assertions).
 func page(t *testing.T, app interface {
 	Router() http.Handler
-}, path string) string {
+}, path string,
+) string {
 	t.Helper()
 	rec := httptest.NewRecorder()
 	app.Router().ServeHTTP(rec, httptest.NewRequest(http.MethodGet, path, nil))

scripts/spike-media/README.md

@@ -0,0 +1,72 @@
+# spike-media — the `/bygge` toolchain, recovered
+
+Throwaway spike tooling for #28 (upload → transcript → takeaways). Committed
+because it was found in `tmpfs` one reboot from gone, not because it is finished.
+
+Nothing here is wired into tapir. No database, no HTTP handler, no ADR.
+
+## What produced `/bygge`
+
+The prototype at `tapir.d-ma.be/bygge` was **not** hand-built, which the epic and
+the first round of spike issues both got wrong. It came from these three scripts:
+
+```
+IMG_1233.mov --transcode.yaml--> IMG_1233.web.mp4   (k3s Job, h264, +faststart)
+IMG_1233.wav --whisper---------> IMG_1233.sv.srt    (interactive — the one manual step)
+IMG_1233.sv.srt --analyze_srt.py--> analys.json     (berget/mistral-medium, temperature=0)
+analys.json + video --build_page.py--> page HTML
+```
+
+The reference analysis was produced by **`berget/mistral-medium`** — a mid-tier
+cloud model, not a frontier one. That matters when judging what a local model
+has to clear: the bar is mistral-medium, and it should be re-run as the control
+arm rather than excluded.
+
+## The validator is the valuable part
+
+`analyze_srt.py` grades its own output on four checks. Two of them catch classes
+of error that timestamp- and citation-checking cannot:
+
+| Check | Catches |
+|---|---|
+| citation exact / **partial** / absent | Paraphrase presented as a quote. Three tiers, because `alternativ` citations legitimately stitch non-contiguous passages — a strict substring test fails the reference output. |
+| `TALET … FINNS INTE I CITATET` | A wrong number carried by a *verbatim* citation and a *real* timestamp. Converts Swedish number-words to digits first. It caught `45×230` where the transcript said `220`. |
+| `TACKNINGSLUCKA` | Spans of the clip no section covers. **The recall term** — without it an empty analysis scores perfectly on every other check. |
+| timestamp exists in the SRT | Fabricated seek targets. Near-solved: 74/75 on the reference, because the model copies the cue starts it was handed. |
+
+## Known defects — read before running
+
+- **`max_tokens: 6000` is too low.** The 4-minute reference analysis is ~4,700
+  output tokens, i.e. ~78% of budget. A 10-minute clip truncates for every model,
+  and truncation reads as a JSON-discipline failure if you are not counting
+  tokens. Record `finish_reason` and separate `truncated` from `malformed`.
+- **`build_page.py` hardcodes the title** (`"Jonas – genomgång av tillbyggnad"`).
+- **The schema carries fields nothing renders** — `projekt.sammanfattning`,
+  `projekt.skede`, `avsnitt.bildtid`, `avsnitt.bildmotiv`, `material[]`,
+  `oklart[]` — left over from an earlier still-frame/markdown output that
+  `build_page.py` replaced. They are paid for in output tokens on every run.
+- **Speaker attribution is inferred, not diarized.** The output attributes
+  statements by name from an SRT with no speaker labels. All four checks above
+  are blind to a misattribution.
+- **The web encode does not downscale.** The source is 720×1280; `-preset
+  veryfast -crf 24` on handheld motion is what makes it ~5 Mbps, not the
+  resolution. The lever is the preset.
+
+## Fixtures
+
+`fixtures/fake.sv.srt` + `fixtures/fake.analys.json` — a synthetic 7-cue Swedish
+construction walkthrough with planted numbers. No real people, no real project.
+It exercises the parser and all four validator checks **without a model call**,
+which is what makes it usable in a test.
+
+## Why the real artifacts are not here
+
+**This repo is public.** The prototype's transcript and analysis are a verbatim
+recording of a named person discussing a client's building project — private
+third-party content, and exactly the class of data #28 calls out as the reason
+uploaded transcripts cannot live in the shared `transcripts` table.
+
+So `IMG_1233.sv.srt`, `payload.json` and the model-comparison runs stay out of
+git here. That leaves `/bygge` without a versioned source of truth, which is a
+real gap and a decision for #28: a private repo, an encrypted blob, or accepting
+that the reference artifacts live outside version control.

scripts/spike-media/analyze_srt.py

@@ -0,0 +1,500 @@
+#!/usr/bin/env python3
+"""Analysera en svensk SRT-transkription av ett byggprojekt.
+
+Steg 1 av 2: producerar tidsstämplad, källbelagd JSON som steg 2
+(ffmpeg-bildutdrag + markdown-montering) konsumerar.
+
+Varje påstående måste bära sitt ordagranna citat och sin tidsstämpel.
+Det gör hallucination synlig i stället för trolig, och ger samtidigt
+bildtiderna gratis.
+
+Körning:
+    export DMABE_LLMAPI_KEY=...            # redan satt på koala
+    python3 analyze_srt.py IMG_1233.sv.srt > analys.json
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import sys
+import time
+import urllib.error
+import urllib.request
+from dataclasses import dataclass, asdict
+
+GATEWAY = os.environ.get("LITELLM_BASE", "http://koala:30401/v1")
+DEFAULT_MODEL = "berget/mistral-medium"
+
+# ---------------------------------------------------------------- SRT-parsning
+
+TS = re.compile(
+    r"(\d{2}):(\d{2}):(\d{2})[,.](\d{3})\s*-->\s*(\d{2}):(\d{2}):(\d{2})[,.](\d{3})"
+)
+
+
+@dataclass
+class Cue:
+    idx: int
+    start: float  # sekunder
+    end: float
+    text: str
+
+    @property
+    def start_hms(self) -> str:
+        return secs_to_hms(self.start)
+
+
+def secs_to_hms(s: float) -> str:
+    h, rem = divmod(int(s), 3600)
+    m, sec = divmod(rem, 60)
+    return f"{h:02d}:{m:02d}:{sec:02d}"
+
+
+def hms_to_secs(v: str) -> float:
+    """Tål '00:01:23', '01:23', '83' och '00:01:23,450'."""
+    v = v.strip().replace(",", ".")
+    parts = v.split(":")
+    try:
+        nums = [float(p) for p in parts]
+    except ValueError as e:
+        raise ValueError(f"ogiltig tidsstämpel: {v!r}") from e
+    if len(nums) == 3:
+        return nums[0] * 3600 + nums[1] * 60 + nums[2]
+    if len(nums) == 2:
+        return nums[0] * 60 + nums[1]
+    if len(nums) == 1:
+        return nums[0]
+    raise ValueError(f"ogiltig tidsstämpel: {v!r}")
+
+
+def parse_srt(raw: str) -> list[Cue]:
+    """Blockbaserad parsning. Tål saknade index och CRLF."""
+    raw = raw.replace("\r\n", "\n").replace("", "")
+    cues: list[Cue] = []
+    for block in re.split(r"\n\s*\n", raw.strip()):
+        lines = [ln for ln in block.split("\n") if ln.strip()]
+        if not lines:
+            continue
+        m = None
+        text_from = 0
+        for i, ln in enumerate(lines[:2]):
+            m = TS.search(ln)
+            if m:
+                text_from = i + 1
+                break
+        if not m:
+            continue
+        g = [int(x) for x in m.groups()]
+        start = g[0] * 3600 + g[1] * 60 + g[2] + g[3] / 1000
+        end = g[4] * 3600 + g[5] * 60 + g[6] + g[7] / 1000
+        text = " ".join(lines[text_from:]).strip()
+        if text:
+            cues.append(Cue(len(cues) + 1, start, end, text))
+    return cues
+
+
+def as_numbered_transcript(cues: list[Cue]) -> str:
+    """Tidsstämplad text som modellen kan citera exakt ur."""
+    return "\n".join(f"[{c.start_hms}] {c.text}" for c in cues)
+
+
+# ------------------------------------------------------------------ LLM-anrop
+
+SYSTEM = """Du är byggteknisk analytiker. Du analyserar en transkription från ett \
+platsbesök där Jonas beskriver ett PLANERAT byggprojekt för minst en annan person.
+
+Detta är ett SAMTAL, inte en monolog. Korta inpass som "Hela vägen?", "Vilken höjd?" \
+eller "Måste det bli det?" är motpartens frågor - de driver samtalet men är inte \
+Jonas påståenden. Attribuera inte en fråga som ett beslut.
+
+Det mesta som sägs är FÖRSLAG under diskussion, inte färdiga beslut. Signalord som \
+"vi tänker", "alternativt", "kanske", "förmodligen", "eventuellt", "om möjligt", \
+"troligast" och "beror på" markerar något ÖPPET. Blanda aldrig ihop ett övervägt \
+alternativ med ett fattat beslut - den skillnaden är hela poängen med analysen.
+
+ABSOLUTA KRAV:
+- Rapportera ENDAST det som faktiskt sägs. Dra inga egna slutsatser, lägg inte till \
+byggmoment, material eller mått som inte nämns explicit.
+- Varje påstående MÅSTE bära ett ordagrant citat ur transkriptionen samt dess tidsstämpel.
+- Mått revideras under samtalets gång ("nio meter" -> "9,40"). Återge alla varianter \
+som nämns, med citat, i stället för att välja en.
+- Är något oklart, ohörbart eller avbrutet: skriv det i fältet "oklart" i stället för \
+att gissa.
+- Svara ENDAST med giltig JSON. Ingen markdown, inga kodstaket, ingen förklarande text."""
+
+USER_TMPL = """Dela in transkriptionen i tematiska avsnitt (typiskt 4-8 stycken) som \
+följer hur Jonas rör sig genom projektet.
+
+TACKNINGSKRAV: avsnitten maste tillsammans tacka HELA klippet, fran 00:00:00 till \
+{slut}. Forsta avsnittet borjar 00:00:00, sista avsnittet slutar {slut}, och varje \
+avsnitts "start" ar lika med foregaende avsnitts "slut". Lamna inga luckor - ocksa \
+korta avslutande kommentarer (tidplan, kostnader, nasta steg) ska tillhora ett avsnitt.
+
+Svara med exakt denna JSON-struktur:
+
+{{
+  "projekt": {{
+    "sammanfattning": "2-3 meningar om vad projektet är",
+    "typ": "t.ex. nybyggnad villa / tillbyggnad / renovering - eller null om det inte framgår",
+    "skede": "vilket byggskede som visas - eller null"
+  }},
+  "avsnitt": [
+    {{
+      "rubrik": "kort beskrivande rubrik",
+      "start": "HH:MM:SS",
+      "slut": "HH:MM:SS",
+      "sammanfattning": "vad Jonas beskriver i detta avsnitt, 1-3 meningar",
+      "nyckelpunkter": [
+        {{"pastaende": "vad som sags", "citat": "ordagrant citat", "tid": "HH:MM:SS"}}
+      ],
+      "bildtid": "HH:MM:SS - ogonblicket dar det Jonas beskriver syns tydligast",
+      "bildmotiv": "vad som bor synas i bilden vid bildtid"
+    }}
+  ],
+  "beslut": [{{"vad": "nagot som uttalas som bestamt", "citat": "ordagrant citat", "tid": "HH:MM:SS"}}],
+  "alternativ": [{{
+    "fraga": "vad som vags mot varandra",
+    "optioner": ["alternativ A", "alternativ B"],
+    "lutar_at": "det alternativ som forespraakas - eller null om oavgjort",
+    "citat": "ordagrant citat", "tid": "HH:MM:SS"
+  }}],
+  "oppna_fragor": [{{
+    "fraga": "det som inte ar avgjort",
+    "beror_pa": "vad avgorandet hanger pa - eller null",
+    "citat": "ordagrant citat", "tid": "HH:MM:SS"
+  }}],
+  "matt": [{{"vad": "matt/dimension", "citat": "ordagrant citat", "tid": "HH:MM:SS"}}],
+  "material": [{{"vad": "material/produkt", "citat": "ordagrant citat", "tid": "HH:MM:SS"}}],
+  "oklart": ["saker som ar ohorbara, avbrutna eller tvetydiga i transkriptionen"]
+}}
+
+Valj "bildtid" omsorgsfullt. I detta samtal pekar Jonas standigt pa saker med ord som \
+"har", "hit", "den typ sa" - orden ar obegripliga utan bilden, sa bildtiden ska ligga \
+dar foremalet han syftar pa syns, inte dar han byter amne. Tomma listor ar helt i sin \
+ordning om inget relevant namns.
+
+TRANSKRIPTION:
+{transcript}"""
+
+
+def call_llm(
+    model: str, transcript: str, slut: str, timeout: int = 300, retries: int = 2
+) -> str:
+    key = os.environ.get("DMABE_LLMAPI_KEY") or os.environ.get("LITELLM_API_KEY")
+    if not key:
+        sys.exit("DMABE_LLMAPI_KEY (eller LITELLM_API_KEY) saknas i miljon.")
+
+    body = json.dumps(
+        {
+            "model": model,
+            "temperature": 0,
+            "max_tokens": 6000,
+            "messages": [
+                {"role": "system", "content": SYSTEM},
+                {"role": "user", "content": USER_TMPL.format(transcript=transcript, slut=slut)},
+            ],
+        }
+    ).encode()
+
+    # Gatewayen begransar burst (limit_req burst=5) och berget-anropet kan
+    # spika. TimeoutError arvs INTE av URLError, sa den maste fangas separat -
+    # annars blir felet en rasptrace efter flera minuters vantan.
+    last = ""
+    for attempt in range(retries + 1):
+        req = urllib.request.Request(
+            f"{GATEWAY}/chat/completions",
+            data=body,
+            headers={"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
+        )
+        try:
+            with urllib.request.urlopen(req, timeout=timeout) as r:
+                payload = json.load(r)
+            break
+        except urllib.error.HTTPError as e:
+            detail = e.read()[:400].decode(errors="replace")
+            if e.code == 429 and attempt < retries:
+                last = f"HTTP 429 (rate limit)"
+            elif e.code >= 500 and attempt < retries:
+                last = f"HTTP {e.code}"
+            else:
+                sys.exit(f"gateway HTTP {e.code}: {detail}")
+        except (TimeoutError, urllib.error.URLError, OSError) as e:
+            reason = getattr(e, "reason", e)
+            if attempt < retries:
+                last = f"{type(e).__name__}: {reason}"
+            else:
+                sys.exit(
+                    f"nadde inte gateway {GATEWAY} efter {retries + 1} forsok "
+                    f"({type(e).__name__}: {reason}). Kontrollera att koala:30401 svarar."
+                )
+        backoff = 3 * (attempt + 1)
+        print(f"[!] {last} - forsok {attempt + 2}/{retries + 1} om {backoff}s", file=sys.stderr)
+        time.sleep(backoff)
+    else:
+        sys.exit(f"gav upp efter {retries + 1} forsok: {last}")
+
+    choice = payload["choices"][0]
+    content = (choice.get("message") or {}).get("content") or ""
+    if not content.strip():
+        # Kant lage for thinking-modeller: resonemanget ater hela budgeten.
+        sys.exit(
+            f"tomt svar (finish_reason={choice.get('finish_reason')}). "
+            "Valj en icke-resonerande modell for strukturerad utdata."
+        )
+    return content
+
+
+def extract_json(s: str) -> dict:
+    s = s.strip()
+    if s.startswith("```"):
+        s = re.sub(r"^```(?:json)?\s*", "", s)
+        s = re.sub(r"\s*```$", "", s)
+    try:
+        return json.loads(s)
+    except json.JSONDecodeError:
+        # Sista utvag: forsta {...sista }
+        i, j = s.find("{"), s.rfind("}")
+        if i >= 0 and j > i:
+            return json.loads(s[i : j + 1])
+        raise
+
+
+# -------------------------------------------------------------- Verifiering
+
+def norm(s: str) -> str:
+    return re.sub(r"[^\wåäöÅÄÖ]+", " ", (s or "").lower()).strip()
+
+
+# --- svenska talord -> siffror -------------------------------------------
+# Byggmatt sags i klartext ("fyrtiofem ganger tvahundratjugo") men refereras
+# i siffror. Utan denna oversattning kan ett felaktigt matt bara ett giltigt
+# ordagrant citat och passera oupptackt - den farligaste feltypen i ett
+# byggdokument.
+
+_SW_UNITS = {
+    "noll": 0, "en": 1, "ett": 1, "två": 2, "tva": 2, "tre": 3, "fyra": 4,
+    "fem": 5, "sex": 6, "sju": 7, "åtta": 8, "atta": 8, "nio": 9, "tio": 10,
+    "elva": 11, "tolv": 12, "tretton": 13, "fjorton": 14, "femton": 15,
+    "sexton": 16, "sjutton": 17, "arton": 18, "nitton": 19,
+}
+_SW_TENS = {
+    "tjugo": 20, "trettio": 30, "fyrtio": 40, "femtio": 50, "sextio": 60,
+    "sjuttio": 70, "åttio": 80, "attio": 80, "nittio": 90,
+}
+_SW_ATOMS = {**_SW_UNITS, **_SW_TENS, "hundra": 100, "tusen": 1000}
+# Langsta forst sa att "sexton" vinner over "sex", "attio" over "atta".
+_SW_ORDER = sorted(_SW_ATOMS, key=len, reverse=True)
+
+
+def _sw_word_to_int(word: str) -> int | None:
+    """'tvåhundratjugo' -> 220. None om ordet inte ar ett rent talord."""
+    w = word.lower()
+    toks: list[str] = []
+    while w:
+        for atom in _SW_ORDER:
+            if w.startswith(atom):
+                toks.append(atom)
+                w = w[len(atom) :]
+                break
+        else:
+            return None
+    if not toks:
+        return None
+
+    total = current = 0
+    for t in toks:
+        v = _SW_ATOMS[t]
+        if t == "hundra":
+            current = (current or 1) * 100
+        elif t == "tusen":
+            total += (current or 1) * 1000
+            current = 0
+        else:
+            current += v
+    return total + current
+
+
+def numbers_in(text: str) -> set[float]:
+    """Alla tal i texten, som siffror och som svenska talord.
+
+    Svenska anvander decimalkomma ("9,40" = 9.4). Naiv \\d+-matchning skulle
+    lasa det som tva separata heltal 9 och 40 - och just har revideras matten
+    live i samtalet ("nio meter" -> "9,40"), sa den skillnaden ar betydelsebarande.
+    """
+    found: set[float] = set()
+    for m in re.finditer(r"\d+(?:[.,]\d+)?", text or ""):
+        found.add(float(m.group().replace(",", ".")))
+    for w in re.findall(r"[a-zåäöA-ZÅÄÖ]+", text or ""):
+        # "en"/"ett" ar obestamd artikel vida oftare an raknetalet 1
+        # ("ett sedumtak"), sa fristaende traffar ger falsklarm. Sammansatta
+        # former ("etthundra") fangas anda av den giriga tokeniseringen.
+        if len(w) < 3 or w.lower() in {"en", "ett"}:
+            continue
+        v = _sw_word_to_int(w)
+        if v is not None and v > 0:
+            found.add(float(v))
+    return found
+
+
+def fmt_num(v: float) -> str:
+    return str(int(v)) if v == int(v) else f"{v:g}"
+
+
+def unsupported(claimed: set[float], supported: set[float]) -> list[float]:
+    """Tal i pastaendet som inte styrks av citatet (med liten tolerans)."""
+    return sorted(
+        n for n in claimed if not any(abs(n - s) < 0.01 for s in supported)
+    )
+
+
+def verify(doc: dict, cues: list[Cue], duration: float) -> list[str]:
+    """Varje citat ska ga att aterfinna; varje tid ska ligga inom klippet.
+
+    Detta ar sjalva poangen med strukturen - ett pastaende utan verifierbar
+    kalla ar en hallucination, och den ska synas har och inte i dokumentet.
+    """
+    warnings: list[str] = []
+    haystack = norm(" ".join(c.text for c in cues))
+
+    def check_quote(where: str, quote: str) -> None:
+        n = norm(quote)
+        if not n:
+            warnings.append(f"{where}: tomt citat")
+            return
+        if n in haystack:
+            return
+        # Delvis traff: minst 60 % av orden i foljd nagonstans
+        words = n.split()
+        if len(words) >= 4:
+            for size in (len(words), max(4, int(len(words) * 0.6))):
+                for i in range(0, len(words) - size + 1):
+                    if " ".join(words[i : i + size]) in haystack:
+                        warnings.append(f"{where}: citat matchar bara delvis: {quote!r}")
+                        return
+        warnings.append(f"{where}: CITAT SAKNAS I TRANSKRIPTION: {quote!r}")
+
+    def check_time(where: str, t: str) -> None:
+        try:
+            v = hms_to_secs(t)
+        except ValueError:
+            warnings.append(f"{where}: ogiltig tid {t!r}")
+            return
+        if not (0 <= v <= duration + 1):
+            warnings.append(f"{where}: tid {t} utanfor klippet (0-{secs_to_hms(duration)})")
+
+    for i, sec in enumerate(doc.get("avsnitt") or [], 1):
+        tag = f"avsnitt {i} ({sec.get('rubrik','?')})"
+        for f in ("start", "slut", "bildtid"):
+            if sec.get(f):
+                check_time(f"{tag}.{f}", sec[f])
+        for j, kp in enumerate(sec.get("nyckelpunkter") or [], 1):
+            check_quote(f"{tag}.nyckelpunkt {j}", kp.get("citat", ""))
+            if kp.get("tid"):
+                check_time(f"{tag}.nyckelpunkt {j}.tid", kp["tid"])
+
+    def check_numbers(where: str, claim: str, quote: str) -> None:
+        """Varje tal i pastaendet maste finnas i dess eget citat."""
+        claimed = numbers_in(claim)
+        if not claimed:
+            return
+        supported = numbers_in(quote)
+        for n in unsupported(claimed, supported):
+            stod = ", ".join(fmt_num(s) for s in sorted(supported)) or "inga tal"
+            warnings.append(
+                f"{where}: TALET {fmt_num(n)} FINNS INTE I CITATET "
+                f"(citat stoder {stod}): {claim!r}"
+            )
+
+    for i, sec in enumerate(doc.get("avsnitt") or [], 1):
+        tag = f"avsnitt {i} ({sec.get('rubrik','?')})"
+        for j, kp in enumerate(sec.get("nyckelpunkter") or [], 1):
+            check_numbers(f"{tag}.nyckelpunkt {j}", kp.get("pastaende", ""), kp.get("citat", ""))
+
+    for key in ("material", "matt", "beslut"):
+        for j, it in enumerate(doc.get(key) or [], 1):
+            check_quote(f"{key}[{j}]", it.get("citat", ""))
+            if it.get("tid"):
+                check_time(f"{key}[{j}].tid", it["tid"])
+            check_numbers(f"{key}[{j}]", it.get("vad", ""), it.get("citat", ""))
+
+    # Tackningskontroll: avsnitten ska tacka hela klippet utan hal, annars
+    # tappas material tyst (och inget bildutdrag gors for den luckan).
+    spans: list[tuple[float, float]] = []
+    for sec in doc.get("avsnitt") or []:
+        try:
+            spans.append((hms_to_secs(sec["start"]), hms_to_secs(sec["slut"])))
+        except (KeyError, ValueError):
+            continue
+    spans.sort()
+    cursor = 0.0
+    for s, e in spans:
+        if s - cursor > 2.0:
+            gap_text = " ".join(
+                c.text for c in cues if c.start >= cursor - 0.5 and c.end <= s + 0.5
+            )
+            warnings.append(
+                f"TACKNINGSLUCKA {secs_to_hms(cursor)}-{secs_to_hms(s)} "
+                f"({s - cursor:.0f}s utan avsnitt): {gap_text[:90]!r}"
+            )
+        cursor = max(cursor, e)
+    if duration - cursor > 2.0:
+        warnings.append(
+            f"TACKNINGSLUCKA {secs_to_hms(cursor)}-{secs_to_hms(duration)} "
+            f"({duration - cursor:.0f}s i slutet utan avsnitt)"
+        )
+    return warnings
+
+
+# ------------------------------------------------------------------- main
+
+def main() -> None:
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument("srt")
+    ap.add_argument("--model", default=DEFAULT_MODEL)
+    ap.add_argument("--raw", action="store_true", help="skriv aven modellens rasvar till stderr")
+    args = ap.parse_args()
+
+    with open(args.srt, encoding="utf-8") as fh:
+        cues = parse_srt(fh.read())
+    if not cues:
+        sys.exit(f"inga cues hittades i {args.srt} - fel format?")
+    duration = cues[-1].end
+
+    print(
+        f"[i] {len(cues)} cues, langd {secs_to_hms(duration)}, modell {args.model}",
+        file=sys.stderr,
+    )
+
+    content = call_llm(args.model, as_numbered_transcript(cues), secs_to_hms(duration))
+    if args.raw:
+        print(content, file=sys.stderr)
+
+    try:
+        doc = extract_json(content)
+    except json.JSONDecodeError as e:
+        sys.exit(f"modellen returnerade ogiltig JSON: {e}\n---\n{content[:800]}")
+
+    warnings = verify(doc, cues, duration)
+    doc["_meta"] = {
+        "modell": args.model,
+        "kallfil": os.path.basename(args.srt),
+        "antal_cues": len(cues),
+        "langd": secs_to_hms(duration),
+        "varningar": warnings,
+    }
+
+    print(json.dumps(doc, ensure_ascii=False, indent=2))
+
+    if warnings:
+        print(f"\n[!] {len(warnings)} verifieringsvarningar:", file=sys.stderr)
+        for w in warnings:
+            print(f"    - {w}", file=sys.stderr)
+    else:
+        print("[ok] alla citat aterfunna, alla tider inom klippet", file=sys.stderr)
+
+
+if __name__ == "__main__":
+    main()

scripts/spike-media/build_page.py

@@ -0,0 +1,357 @@
+#!/usr/bin/env python3
+"""Bygg en fristaende HTML-sida: video + SRT + tidssynkad analys.
+
+Sidan ar sjalvbarande sa nar som pa videofilen - transkription och analys
+bakas in i HTML:en. Det gor att den fungerar aven over file:// (fetch mot
+en lokal JSON blockeras av CORS).
+
+    python3 build_page.py payload.json IMG_1233.web.mp4 > index.html
+"""
+
+from __future__ import annotations
+
+import html
+import json
+import sys
+
+TEMPLATE = """<!doctype html>
+<html lang="sv">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>__TITLE__</title>
+<style>
+  :root {
+    --bg:#f6f7f9; --panel:#fff; --ink:#16181d; --muted:#6b7280; --line:#e3e6ea;
+    --accent:#2563eb; --accent-soft:#eff4ff;
+    --beslut:#0f9d58; --alt:#d97706; --oppen:#dc2626; --matt:#7c3aed;
+  }
+  @media (prefers-color-scheme: dark) {
+    :root {
+      --bg:#0e1116; --panel:#161a21; --ink:#e6e9ef; --muted:#9aa3b2; --line:#262c36;
+      --accent:#60a5fa; --accent-soft:#17233b;
+      --beslut:#34d399; --alt:#fbbf24; --oppen:#f87171; --matt:#a78bfa;
+    }
+  }
+  :root[data-theme="dark"] {
+    --bg:#0e1116; --panel:#161a21; --ink:#e6e9ef; --muted:#9aa3b2; --line:#262c36;
+    --accent:#60a5fa; --accent-soft:#17233b;
+    --beslut:#34d399; --alt:#fbbf24; --oppen:#f87171; --matt:#a78bfa;
+  }
+  :root[data-theme="light"] {
+    --bg:#f6f7f9; --panel:#fff; --ink:#16181d; --muted:#6b7280; --line:#e3e6ea;
+    --accent:#2563eb; --accent-soft:#eff4ff;
+    --beslut:#0f9d58; --alt:#d97706; --oppen:#dc2626; --matt:#7c3aed;
+  }
+  * { box-sizing:border-box; }
+  body {
+    margin:0; background:var(--bg); color:var(--ink);
+    font:15px/1.55 -apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,sans-serif;
+  }
+  header {
+    padding:14px 20px; border-bottom:1px solid var(--line); background:var(--panel);
+    display:flex; align-items:baseline; gap:14px; flex-wrap:wrap;
+  }
+  header h1 { margin:0; font-size:17px; font-weight:650; }
+  header .sub { color:var(--muted); font-size:13px; }
+
+  .wrap {
+    display:grid; grid-template-columns: minmax(0,1.35fr) minmax(320px,1fr);
+    gap:20px; padding:20px; align-items:start; max-width:1600px; margin:0 auto;
+  }
+  @media (max-width: 950px) { .wrap { grid-template-columns:1fr; } }
+
+  /* Videon star stilla; analysen ror sig bredvid. */
+  .videocol { position:sticky; top:20px; }
+  video { width:100%; border-radius:10px; background:#000; display:block; }
+
+  .nowbar {
+    margin-top:10px; padding:12px 14px; background:var(--panel);
+    border:1px solid var(--line); border-radius:10px; min-height:64px;
+  }
+  .nowbar .label { font-size:11px; text-transform:uppercase; letter-spacing:.07em; color:var(--muted); }
+  .nowbar .text { margin-top:4px; font-size:16px; }
+
+  .tabs { display:flex; gap:6px; margin-bottom:12px; flex-wrap:wrap; }
+  .tab {
+    padding:6px 12px; border:1px solid var(--line); background:var(--panel);
+    border-radius:999px; cursor:pointer; font-size:13px; color:var(--ink);
+  }
+  .tab[aria-selected="true"] { background:var(--accent); border-color:var(--accent); color:#fff; }
+
+  .pane { display:none; }
+  .pane.active { display:block; }
+
+  .scroller {
+    max-height: calc(100vh - 190px); overflow-y:auto; padding-right:6px;
+    scroll-behavior:smooth;
+  }
+
+  .card {
+    background:var(--panel); border:1px solid var(--line); border-left:3px solid var(--line);
+    border-radius:8px; padding:11px 13px; margin-bottom:9px; cursor:pointer;
+    transition:border-color .15s, background .15s;
+  }
+  .card:hover { border-color:var(--accent); }
+  .card.on { background:var(--accent-soft); border-left-color:var(--accent); }
+  .card .k {
+    font-size:10px; text-transform:uppercase; letter-spacing:.07em;
+    font-weight:700; margin-bottom:5px;
+  }
+  .card .t { font-size:11px; color:var(--muted); font-variant-numeric:tabular-nums; }
+  .card .cit { display:block; margin-top:6px; color:var(--muted); font-size:13px; font-style:italic; }
+  .k.beslut{color:var(--beslut)} .k.alt{color:var(--alt)}
+  .k.oppen{color:var(--oppen)} .k.matt{color:var(--matt)}
+  .card.beslut{border-left-color:var(--beslut)} .card.alt{border-left-color:var(--alt)}
+  .card.oppen{border-left-color:var(--oppen)} .card.matt{border-left-color:var(--matt)}
+  .opts { margin:6px 0 0; padding-left:18px; font-size:13px; }
+  .lutar { font-size:12px; color:var(--muted); margin-top:5px; }
+
+  .cue {
+    padding:5px 9px; border-radius:6px; cursor:pointer; display:flex; gap:10px;
+    font-size:14px; border-left:2px solid transparent;
+  }
+  .cue:hover { background:var(--accent-soft); }
+  .cue.on { background:var(--accent-soft); border-left-color:var(--accent); font-weight:600; }
+  .cue .ts {
+    color:var(--muted); font-variant-numeric:tabular-nums; font-size:12px;
+    flex:0 0 46px; padding-top:2px;
+  }
+  .sec-head {
+    margin:16px 0 7px; font-size:12px; font-weight:700; text-transform:uppercase;
+    letter-spacing:.06em; color:var(--muted); border-bottom:1px solid var(--line);
+    padding-bottom:5px;
+  }
+  .sec-head:first-child { margin-top:0; }
+  .warn {
+    background:var(--panel); border:1px solid var(--line); border-left:3px solid var(--alt);
+    border-radius:8px; padding:11px 13px; font-size:13px; color:var(--muted);
+  }
+  .warn b { color:var(--ink); }
+  .warn ul { margin:6px 0 0; padding-left:18px; }
+</style>
+</head>
+<body>
+<header>
+  <h1>__TITLE__</h1>
+  <span class="sub">__SUB__</span>
+</header>
+
+<div class="wrap">
+  <div class="videocol">
+    <video id="v" controls preload="metadata" src="__VIDEO__"></video>
+    <div class="nowbar">
+      <div class="label">Sägs nu</div>
+      <div class="text" id="now">—</div>
+    </div>
+  </div>
+
+  <div>
+    <div class="tabs" role="tablist">
+      <button class="tab" role="tab" aria-selected="true"  data-pane="analys">Analys</button>
+      <button class="tab" role="tab" aria-selected="false" data-pane="text">Transkription</button>
+      <button class="tab" role="tab" aria-selected="false" data-pane="kvalitet">Kvalitet</button>
+    </div>
+    <div class="pane active" id="pane-analys"><div class="scroller" id="analys"></div></div>
+    <div class="pane" id="pane-text"><div class="scroller" id="text"></div></div>
+    <div class="pane" id="pane-kvalitet"><div class="scroller" id="kvalitet"></div></div>
+  </div>
+</div>
+
+<script>
+const DATA = __DATA__;
+const v = document.getElementById('v');
+
+const hms = s => {
+  s = Math.max(0, Math.floor(s));
+  return String(Math.floor(s/60)).padStart(2,'0') + ':' + String(s%60).padStart(2,'0');
+};
+const toSec = t => {
+  if (t == null) return null;
+  const p = String(t).trim().split(':').map(Number);
+  if (p.some(isNaN)) return null;
+  return p.length === 3 ? p[0]*3600+p[1]*60+p[2] : p.length === 2 ? p[0]*60+p[1] : p[0];
+};
+const esc = s => { const d = document.createElement('div'); d.textContent = s ?? ''; return d.innerHTML; };
+
+/* Varje kort bar sin tid; ett klick soker dit. Analysen ar tidsforankrad
+   hela vagen ned, sa synkningen ar bara en sortering pa den tiden. */
+const items = [];
+const A = DATA.analys;
+
+(A.avsnitt || []).forEach((s, i) => {
+  const t = toSec(s.start) ?? 0;
+  items.push({t, end: toSec(s.slut) ?? 1e9, kind:'avsnitt', html:
+    `<div class="k">Avsnitt ${i+1} · ${esc(s.start)}–${esc(s.slut)}</div>
+     <div style="font-weight:650;font-size:15px">${esc(s.rubrik)}</div>
+     <div style="margin-top:5px">${esc(s.sammanfattning)}</div>`});
+  (s.nyckelpunkter || []).forEach(k => {
+    const kt = toSec(k.tid); if (kt == null) return;
+    items.push({t:kt, end:kt+6, kind:'punkt', citat:k.citat, html:
+      `<div class="t">${esc(k.tid)}</div><div>${esc(k.pastaende)}</div>
+       <div class="cit">”${esc(k.citat)}”</div>`});
+  });
+});
+(A.beslut || []).forEach(b => {
+  const t = toSec(b.tid); if (t == null) return;
+  items.push({t, end:t+8, kind:'beslut', citat:b.citat, html:
+    `<div class="k beslut">Beslut</div><div class="t">${esc(b.tid)}</div>
+     <div style="font-weight:600">${esc(b.vad)}</div>
+     ${b.skal ? `<div class="lutar">Skäl: ${esc(b.skal)}</div>` : ''}
+     <div class="cit">”${esc(b.citat)}”</div>`});
+});
+(A.alternativ || []).forEach(a => {
+  const t = toSec(a.tid); if (t == null) return;
+  items.push({t, end:t+8, kind:'alt', citat:a.citat, html:
+    `<div class="k alt">Alternativ</div><div class="t">${esc(a.tid)}</div>
+     <div style="font-weight:600">${esc(a.fraga)}</div>
+     <ul class="opts">${(a.optioner||[]).map(o=>`<li>${esc(o)}</li>`).join('')}</ul>
+     <div class="lutar">${a.lutar_at ? 'Lutar åt: <b>'+esc(a.lutar_at)+'</b>' : 'Oavgjort'}</div>
+     <div class="cit">”${esc(a.citat)}”</div>`});
+});
+(A.oppna_fragor || []).forEach(o => {
+  const t = toSec(o.tid); if (t == null) return;
+  items.push({t, end:t+8, kind:'oppen', citat:o.citat, html:
+    `<div class="k oppen">Öppen fråga</div><div class="t">${esc(o.tid)}</div>
+     <div style="font-weight:600">${esc(o.fraga)}</div>
+     ${o.beror_pa ? `<div class="lutar">Beror på: ${esc(o.beror_pa)}</div>` : ''}
+     <div class="cit">”${esc(o.citat)}”</div>`});
+});
+(A.matt || []).forEach(m => {
+  const t = toSec(m.tid); if (t == null) return;
+  items.push({t, end:t+6, kind:'matt', citat:m.citat, html:
+    `<div class="k matt">Mått</div><div class="t">${esc(m.tid)}</div>
+     <div>${esc(m.vad)}</div><div class="cit">”${esc(m.citat)}”</div>`});
+});
+
+const RANK = {avsnitt:0, beslut:1, oppen:2, alt:3, matt:4, punkt:5};
+/* Samma replik klassas ofta bade som nyckelpunkt och som matt/beslut.
+   Visa den mest specifika varianten en gang i stallet for tva nastan
+   identiska kort - annars later flodet som ett eko. */
+const bestByQuote = new Map();
+items.forEach(it => {
+  const key = (it.citat||'').trim().toLowerCase() + '@' + it.t;
+  if (!it.citat) return;
+  const prev = bestByQuote.get(key);
+  if (prev === undefined || RANK[it.kind] < RANK[prev.kind]) bestByQuote.set(key, it);
+});
+const kept = items.filter(it => {
+  if (!it.citat) return true;
+  const key = (it.citat||'').trim().toLowerCase() + '@' + it.t;
+  return bestByQuote.get(key) === it;
+});
+items.length = 0; items.push(...kept);
+items.sort((a,b) => a.t - b.t || RANK[a.kind] - RANK[b.kind]);
+
+const analysEl = document.getElementById('analys');
+analysEl.innerHTML = items.map((it,i) =>
+  `<div class="card ${it.kind}" data-i="${i}" data-t="${it.t}">${it.html}</div>`).join('');
+
+const textEl = document.getElementById('text');
+textEl.innerHTML = DATA.cues.map((c,i) =>
+  `<div class="cue" data-ci="${i}" data-t="${c.t}">
+     <span class="ts">${hms(c.t)}</span><span>${esc(c.text)}</span></div>`).join('');
+
+const warns = (A._meta && A._meta.varningar) || [];
+document.getElementById('kvalitet').innerHTML = `
+  <div class="warn">
+    <b>Automatisk källkontroll</b><br>
+    Modell: ${esc((A._meta&&A._meta.modell)||'–')} · ${DATA.cues.length} repliker · längd ${esc((A._meta&&A._meta.langd)||'–')}<br><br>
+    Varje påstående kontrolleras mot sitt eget citat: citatet måste återfinnas
+    ordagrant i transkriptionen, tiden måste ligga inom klippet, och varje tal i
+    påståendet måste styrkas av citatet.
+    ${warns.length
+      ? `<br><br><b>${warns.length} varningar:</b><ul>${warns.map(w=>`<li>${esc(w)}</li>`).join('')}</ul>`
+      : '<br><br>Inga avvikelser.'}
+  </div>`;
+
+/* Klick söker i videon - åt båda håll: kort och replik. */
+document.addEventListener('click', e => {
+  const el = e.target.closest('.card, .cue');
+  if (!el) return;
+  v.currentTime = parseFloat(el.dataset.t) + 0.05;
+  v.play().catch(()=>{});
+});
+
+document.querySelectorAll('.tab').forEach(tab => {
+  tab.addEventListener('click', () => {
+    document.querySelectorAll('.tab').forEach(t => t.setAttribute('aria-selected','false'));
+    tab.setAttribute('aria-selected','true');
+    document.querySelectorAll('.pane').forEach(p => p.classList.remove('active'));
+    document.getElementById('pane-' + tab.dataset.pane).classList.add('active');
+  });
+});
+
+/* Auto-scroll bara nar anvandaren inte sjalv scrollar - annars slits vyn
+   ur handerna pa den som lasar. */
+let userScrolled = 0;
+document.querySelectorAll('.scroller').forEach(s =>
+  s.addEventListener('wheel', () => { userScrolled = Date.now(); }, {passive:true}));
+
+let lastCard = -1, lastCue = -1;
+function sync() {
+  const t = v.currentTime;
+
+  let ci = -1;
+  for (let i = 0; i < DATA.cues.length; i++) {
+    if (t >= DATA.cues[i].t - 0.15) ci = i; else break;
+  }
+  if (ci !== lastCue) {
+    document.querySelectorAll('.cue.on').forEach(e => e.classList.remove('on'));
+    if (ci >= 0) {
+      const el = textEl.querySelector(`[data-ci="${ci}"]`);
+      if (el) {
+        el.classList.add('on');
+        if (Date.now() - userScrolled > 3000) el.scrollIntoView({block:'center'});
+      }
+      document.getElementById('now').textContent = DATA.cues[ci].text;
+    }
+    lastCue = ci;
+  }
+
+  let ai = -1;
+  for (let i = 0; i < items.length; i++) {
+    if (t >= items[i].t - 0.15) ai = i; else break;
+  }
+  if (ai !== lastCard) {
+    document.querySelectorAll('.card.on').forEach(e => e.classList.remove('on'));
+    if (ai >= 0) {
+      const el = analysEl.querySelector(`[data-i="${ai}"]`);
+      if (el) {
+        el.classList.add('on');
+        if (Date.now() - userScrolled > 3000 &&
+            document.getElementById('pane-analys').classList.contains('active')) {
+          el.scrollIntoView({block:'center'});
+        }
+      }
+    }
+    lastCard = ai;
+  }
+  requestAnimationFrame(sync);
+}
+requestAnimationFrame(sync);
+</script>
+</body>
+</html>
+"""
+
+
+def main() -> None:
+    payload_path, video = sys.argv[1], sys.argv[2]
+    payload = json.load(open(payload_path, encoding="utf-8"))
+    a = payload["analys"]
+    proj = a.get("projekt") or {}
+    sub = " · ".join(
+        x for x in [proj.get("typ"), (a.get("_meta") or {}).get("langd"), video] if x
+    )
+    out = (
+        TEMPLATE.replace("__TITLE__", html.escape("Jonas – genomgång av tillbyggnad"))
+        .replace("__SUB__", html.escape(sub))
+        .replace("__VIDEO__", html.escape(video))
+        .replace("__DATA__", json.dumps(payload, ensure_ascii=False))
+    )
+    sys.stdout.write(out)
+
+
+if __name__ == "__main__":
+    main()

scripts/spike-media/fixtures/fake.analys.json

@@ -0,0 +1,173 @@
+{
+  "projekt": {
+    "sammanfattning": "Projektet är en nybyggnad med grundläggning, betongplatta och limträstomme. Taket kommer att vara ett sedumtak.",
+    "typ": "nybyggnad",
+    "skede": "stomme"
+  },
+  "avsnitt": [
+    {
+      "rubrik": "Introduktion och grundläggning",
+      "start": "00:00:00",
+      "slut": "00:00:06",
+      "sammanfattning": "Jonas introducerar platsen och meddelar att grundläggningen är klar.",
+      "nyckelpunkter": [
+        {
+          "pastaende": "Grundläggningen är klar",
+          "citat": "Vi har precis blivit klara med grundläggningen.",
+          "tid": "00:00:03"
+        }
+      ],
+      "bildtid": "00:00:03",
+      "bildmotiv": "tomten med färdig grundläggning"
+    },
+    {
+      "rubrik": "Dränering och grundvatten",
+      "start": "00:00:06",
+      "slut": "00:00:14",
+      "sammanfattning": "Jonas beskriver dräneringsarbetet som krävdes på grund av högt grundvatten.",
+      "nyckelpunkter": [
+        {
+          "pastaende": "Dränering utfördes på grund av högt grundvatten",
+          "citat": "Vi fick dränera ordentligt först, för grundvattnet stod väldigt högt här.",
+          "tid": "00:00:06"
+        }
+      ],
+      "bildtid": "00:00:10",
+      "bildmotiv": "dräneringssystem eller markarbete"
+    },
+    {
+      "rubrik": "Betongplatta och isolering",
+      "start": "00:00:14",
+      "slut": "00:00:22",
+      "sammanfattning": "Jonas beskriver betongplattan och isoleringen under den.",
+      "nyckelpunkter": [
+        {
+          "pastaende": "Betongplatta på mark med cellplast under",
+          "citat": "det är en betongplatta på mark med tvåhundra millimeter cellplast under.",
+          "tid": "00:00:16"
+        }
+      ],
+      "bildtid": "00:00:18",
+      "bildmotiv": "betongplattan med cellplast"
+    },
+    {
+      "rubrik": "Stomme och bjälklag",
+      "start": "00:00:22",
+      "slut": "00:00:31",
+      "sammanfattning": "Jonas beskriver stommen i limträ och bjälklagets dimensioner.",
+      "nyckelpunkter": [
+        {
+          "pastaende": "Stommen är i limträ",
+          "citat": "Här ser ni stommen, den är i limträ.",
+          "tid": "00:00:22"
+        },
+        {
+          "pastaende": "Pelarnas dimensioner",
+          "citat": "Pelarna är hundraåttio gånger hundraåttio.",
+          "tid": "00:00:25"
+        },
+        {
+          "pastaende": "Bjälklagets spännvidd och dimensioner",
+          "citat": "Bjälklaget spänner sex meter fritt, och vi har fyrtiofem gånger tvåhundratjugo som balkar.",
+          "tid": "00:00:31"
+        }
+      ],
+      "bildtid": "00:00:28",
+      "bildmotiv": "limträstommen med pelare och bjälklag"
+    },
+    {
+      "rubrik": "Tak och tidplan",
+      "start": "00:00:40",
+      "slut": "00:00:48",
+      "sammanfattning": "Jonas nämner taktypen och förklarar att projektet är försenat på grund av leveransproblem.",
+      "nyckelpunkter": [
+        {
+          "pastaende": "Taket kommer att vara sedumtak",
+          "citat": "Taket blir sedumtak, det var byggherrens önskemål.",
+          "tid": "00:00:40"
+        },
+        {
+          "pastaende": "Sedumtaket ställer krav på bärigheten",
+          "citat": "Det ställer krav på bärigheten.",
+          "tid": "00:00:44"
+        },
+        {
+          "pastaende": "Projektet är två veckor efter tidplanen",
+          "citat": "Vi ligger ungefär två veckor efter tidplanen på grund av leveransproblem på limträet.",
+          "tid": "00:00:48"
+        }
+      ],
+      "bildtid": "00:00:44",
+      "bildmotiv": "stommen med eventuella förberedelser för taket"
+    }
+  ],
+  "material": [
+    {
+      "vad": "cellplast",
+      "citat": "med tvåhundra millimeter cellplast under",
+      "tid": "00:00:18"
+    },
+    {
+      "vad": "betongplatta",
+      "citat": "det är en betongplatta på mark",
+      "tid": "00:00:16"
+    },
+    {
+      "vad": "limträ",
+      "citat": "den är i limträ",
+      "tid": "00:00:22"
+    },
+    {
+      "vad": "sedumtak",
+      "citat": "Taket blir sedumtak",
+      "tid": "00:00:40"
+    }
+  ],
+  "matt": [
+    {
+      "vad": "200 mm cellplast",
+      "citat": "tvåhundra millimeter cellplast",
+      "tid": "00:00:18"
+    },
+    {
+      "vad": "180x180 mm pelare",
+      "citat": "Pelarna är hundraåttio gånger hundraåttio",
+      "tid": "00:00:25"
+    },
+    {
+      "vad": "6 meter spännvidd för bjälklag",
+      "citat": "Bjälklaget spänner sex meter fritt",
+      "tid": "00:00:31"
+    },
+    {
+      "vad": "45x230 mm balkar",
+      "citat": "fyrtiofem gånger tvåhundratjugo som balkar",
+      "tid": "00:00:31"
+    }
+  ],
+  "problem": [
+    {
+      "vad": "högt grundvatten",
+      "citat": "grundvattnet stod väldigt högt här",
+      "tid": "00:00:10"
+    },
+    {
+      "vad": "leveransproblem med limträ",
+      "citat": "leveransproblem på limträet",
+      "tid": "00:00:48"
+    },
+    {
+      "vad": "försenat projekt",
+      "citat": "Vi ligger ungefär två veckor efter tidplanen",
+      "tid": "00:00:48"
+    }
+  ],
+  "oklart": [],
+  "_meta": {
+    "modell": "berget/mistral-medium",
+    "kallfil": "fake.sv.srt",
+    "antal_cues": 7,
+    "langd": "00:00:56",
+    "varningar": []
+  }
+}

scripts/spike-media/fixtures/fake.sv.srt

@@ -0,0 +1,27 @@
+1
+00:00:00,000 --> 00:00:06,500
+Ja, då står vi här på tomten. Vi har precis blivit klara med grundläggningen.
+
+2
+00:00:06,500 --> 00:00:14,000
+Vi fick dränera ordentligt först, för grundvattnet stod väldigt högt här.
+
+3
+00:00:14,000 --> 00:00:22,000
+Sen göt vi plattan, det är en betongplatta på mark med tvåhundra millimeter cellplast under.
+
+4
+00:00:22,000 --> 00:00:31,000
+Här ser ni stommen, den är i limträ. Pelarna är hundraåttio gånger hundraåttio.
+
+5
+00:00:31,000 --> 00:00:40,000
+Bjälklaget spänner sex meter fritt, och vi har fyrtiofem gånger tvåhundratjugo som balkar.
+
+6
+00:00:40,000 --> 00:00:48,000
+Taket blir sedumtak, det var byggherrens önskemål. Det ställer krav på bärigheten.
+
+7
+00:00:48,000 --> 00:00:56,000
+Vi ligger ungefär två veckor efter tidplanen på grund av leveransproblem på limträet.

scripts/spike-media/transcode.yaml

@@ -0,0 +1,51 @@
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: jonas-transcode
+  namespace: default
+spec:
+  ttlSecondsAfterFinished: 600
+  activeDeadlineSeconds: 1800
+  backoffLimit: 0
+  template:
+    spec:
+      restartPolicy: Never
+      nodeSelector:
+        kubernetes.io/hostname: koala
+      securityContext:
+        runAsUser: 1000
+        runAsGroup: 1000
+        fsGroup: 1000
+      containers:
+        - name: ffmpeg
+          image: mwader/static-ffmpeg:7.1
+          args:
+            - "-nostdin"
+            - "-i"
+            - "/work/IMG_1233.mov"
+            - "-c:v"
+            - "libx264"
+            - "-preset"
+            - "veryfast"
+            - "-crf"
+            - "24"
+            - "-vf"
+            - "scale='min(1280,iw)':-2"
+            - "-c:a"
+            - "aac"
+            - "-b:a"
+            - "128k"
+            - "-movflags"
+            - "+faststart"
+            - "-y"
+            - "/work/IMG_1233.web.mp4"
+          volumeMounts:
+            - name: work
+              mountPath: /work
+          resources:
+            limits: {cpu: "6", memory: "4Gi"}
+      volumes:
+        - name: work
+          hostPath:
+            path: /home/mathias/dev/AGENTS/agentsquad
+            type: Directory

test/acceptance/summarize_new_video_test.go

@@ -25,9 +25,11 @@ type fakeSource struct {
 func (f fakeSource) ListSubscriptions(ctx context.Context, userID string) ([]domain.Subscription, error) {
 	return nil, nil
 }
+
 func (f fakeSource) NewVideos(ctx context.Context, sub domain.Subscription) ([]domain.Video, error) {
 	return nil, nil
 }
+
 func (f fakeSource) FetchTranscript(ctx context.Context, v domain.Video) (domain.Transcript, error) {
 	return f.transcript, nil
 }
@@ -132,10 +134,12 @@ type watchSource struct {
 func (s *watchSource) ListSubscriptions(ctx context.Context, userID string) ([]domain.Subscription, error) {
 	return s.subs, nil
 }
+
 func (s *watchSource) NewVideos(ctx context.Context, sub domain.Subscription) ([]domain.Video, error) {
 	s.newVideoHits++
 	return s.videosBySub[sub.ID], nil
 }
+
 func (s *watchSource) FetchTranscript(ctx context.Context, v domain.Video) (domain.Transcript, error) {
 	return s.transcripts[v.ID], nil
 }