TAPIR_E2E_RECONCILE=1 runs pianissimo + kb-whisper + gemma4-26b on the audio and
asserts the reconciliation is recorded, every contributor is named in the endpoint, and
the page says what was done. Run 2026-10-01 on 82 s of synthetic Swedish: 3 disputes, all
3 changed; against the known script 2 were right (virket, a missing 'procent') and 1 was
WRONG (Maria -> Vi replaced a correct base reading). A smoke test, not a measurement.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Default is the single-engine chain exactly as before. With TAPIR_RECONCILE_BASE_MODEL set,
the base engine and the primary run on the same WAV and a LOCAL model (config has already
refused any non-local alias) reconciles them, at temperature 0 with a short reply -- the
conditions its accuracy was measured under. The cloud fallback is unchanged and runs
only when both local engines fail.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Run 2026-10-01 with 82 s of synthetic Swedish muxed with black video (H.264 +
AAC .mp4, built with the real ffmpeg): accepted, had_video=true, transcribed by
iguana/kb-whisper, the page said only its audio was used, and chat answered.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
cmdServe assigned Uploads and UploadLimits but not UploadChatMaxChars, so chat
over a long recording would have used the 18,000-char caption default. It now
calls wireUploads, which is tested.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
I wired Uploads and UploadLimits in cmdServe but not UploadChatMaxChars: chat
over a long recording would have silently used the 18,000-char caption
default. A helper that main and the e2e test both call makes that a tested fact.
Red: wireUploads is undefined.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Real ffprobe/ffmpeg, the live gateway and a real chat model, a private
embedded Postgres. Skipped unless TAPIR_E2E_AUDIO is set. Run 2026-10-01 with
82 s of synthetic Swedish as AAC .m4a:
- primary: iguana/kb-whisper, 82.6 s audio, 1,186 chars; asked who owns the
timber order, the answer named Anders and the 15 November deadline from the
transcript; the spool directory was empty afterwards.
- primary forced to fail (TAPIR_TRANSCRIBE_MODEL=iguana/does-not-exist): the
fallback served it and the row recorded endpoint=berget/kb-whisper.
Synthetic audio only; this repo is public.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
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>
Returns nils when disabled (never a typed-nil Intake). When enabled: private
0700 spool dir, delete audio left by a crash, fail rows a restart left queued
or transcribing (the audio is not retained, so they cannot resume), then build
the iguana->berget chain, worker and intake. An empty
TAPIR_TRANSCRIBE_FALLBACK_MODEL keeps private audio on the homelab. Enabling
with no gateway fails at start-up, not at the first upload.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Wire the metrics package into the live paths and serve it:
- summarizer: per-endpoint latency by model/outcome(success|error|parse_error)/fallback + slog.
- youtube.FetchTranscript: latency by outcome (captions|none|rate_limited) + slog.
- chat: answer latency by model + slog.
- llm usage hook → token counts (prompt|completion) per model, wired in buildSummarizer/buildChat.
- oidc callback: login counter.
- cmdServe: wrap Router in metrics.HTTPMiddleware (request count + latency by bounded
route pattern) and serve /metrics on TAPIR_METRICS_ADDR (default :9090), a SEPARATE
port — never on the public app mux.
BDD: observability.feature scenarios un-pended + mapped. TDD: summarizer wiring tested
black-box via the /metrics scrape; metrics-not-on-public-mux asserted.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The return-usage gate (ADR-016) counted distinct active weeks over ALL history,
so pre-launch noise — testing churn and the period the pilot sat blocked on zero
summaries — would inflate the signal. Add a baseline: ActiveWeeks(ctx, since)
filters login_events + summary_actions to seen_at/acted_at >= since. The report
command sets it to TAPIR_USAGE_GATE_START (YYYY-MM-DD, default 2026-06-11 — the
morning the pilot was unblocked) and prints the baseline. The gate now measures
whether users RETURN once it genuinely works.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire the onboarding burst to its quality-aware selection and a stronger model:
- main.go onboard uses OnboardBurstVideoIDs (junk-avoiding) instead of pure
newest-first, bounded by MinVideoSeconds / OnboardMaxVideoSeconds.
- buildBurstProcessor builds a burst-only summarizer chain led by the onboard
model (burstChainModels: onboard -> standard ADR-022 chain, deduped, NDA lever
intact), over the same store/cache/sink. Collapses onto the shared Processor
when the onboard model is empty/equal-to-primary or config is incomplete.
- summarizerEndpoint + newYouTubeSource extracted so standard and burst wiring
share one definition.
- Remove now-superseded NewestUnsummarizedVideoIDs: OnboardBurstVideoIDs(.,0,0)
is identical pure-newest behaviour and its test covers RLS + ordering.
A deeper-dive chat entered from the summary view: ask questions about a video
against its already-stored transcript (ADR-021), no caption fetch, ever.
Safety by construction — the load-bearing property. The chat handlers reach the
chat.Service only after reading the SHARED stored transcript via Store.GetTranscript
(a pure DB read); the service holds no VideoSource. So an enabled chat cannot
trigger a caption fetch, touch the rate gate, or reach YouTube. A video with no
stored transcript gets an honest "not available" — no fetch, no model call. The
key web test wires the summarize/fetch collaborators as tripwires that fail the
test if chat ever routes into them, and asserts the model answered from the
stored text.
Model defaults to the summary's own model and is switchable among the ADR-022
chain (phi4-mini → gemma4-26b → mistral-small); switching re-runs against the
same transcript — deliberate model-comparison instrumentation. The cloud model
is absent from the switcher when TAPIR_CLOUD_FALLBACK_MODEL="" (the local-first /
NDA lever), honoured the same way the summarizer honours it. Reuses the existing
LiteLLM gateway client (a chat is a different call, not a new integration) and
the TAPIR_MAX_TRANSCRIPT_CHARS truncation, surfacing an honest bounded-context
note when a long transcript is cut.
Ephemeral v1: the multi-turn conversation rides in hidden request fields; no
table, no migration, nothing persisted. Entry is RLS-scoped through
GetSummaryByVideo, so chat is reachable only from the user's own summary view.
Show-source verification and on-demand fetch are deferred (ADR-027).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A user waiting on a Summarize click shared the per-IP caption gate equally with
the background firehose, so on a busy IP the click was slow or 429'd. Add a
context-marked priority lane: the web path (engineProcessor.ProcessVideo) marks
its context foreground; the gate serves foreground immediately while background
fetches yield until no foreground is pending. Threaded via a context value (no
new signatures) + a process-wide foregroundPending counter. Clicks are rare, so
the background barely loses throughput; the waiting human gets the cleaner slot.
Drops the credentials probe: ADR-010 and captions.go already settle it — the
timedtext/InnerTube path rejects authenticated requests and the OAuth token does
not authenticate it anyway, so auth cannot help and can hurt. Documented in
ADR-026 rather than built.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The lead-user rotation used an in-memory pass counter reset to 0 on every pod
restart, so the first-listed user always re-took the lead after a restart — a
deploy-heavy session re-starved the last user (the pilot stalled at 2 summaries
because each deploy reset his every-other-pass lead before the 2h tick fired).
Derive the offset from wall-clock (floor(now/interval)) so it advances with real
time and is identical across restarts: rotation stays fair however often the pod
bounces.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Clicking Summarize polled /status, which knew only "spinning" or "done". The web
ProcessVideo recorded an outcome only on success, so a 429'd or caption-less
click left transcript_status unset and the poll silently reverted to the
Summarize button — the rate limit was invisible and the click felt broken.
- ProcessVideo now records rate_limited / none / fetched (mirrors the runner); a
rate-limited video keeps its requested flag so the background sweep retries it.
- /status is state-aware: summary card (done), working spinner (in-flight),
a calm "waiting on rate limit, will retry" card that keeps polling so the
summary lands on its own (no re-click), and a terminal "no captions" card.
- Charm status text (Claude-Code / Crush inspired): the spinner cycles playful
tapir-themed gerunds via CSS only (no JS), decorative + an sr-only stable
status line for a11y.
Pillar A (foreground fetch priority lane) is a separate follow-up.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
After transcript caching (ADR-021) and the Shorts filter (ADR-023), the remaining
caption waste is the first fetch on every new video of a channel that never has
English captions — each costs one rate-limited fetch to resolve to "none", and on
a throttled IP churns the backoff machinery first.
Remember, per (user, channel), a streak of consecutive no-caption outcomes
(channel_caption_state, migration 016, RLS-scoped). Once it reaches
TAPIR_CHANNEL_CAPTIONLESS_THRESHOLD (default 5) the channel is suppressed — videos
discovered/listed but not caption-fetched — for TAPIR_CHANNEL_CAPTIONLESS_WINDOW
(default 14d), then one is re-probed (auto-recovery). A successful fetch resets
the streak; a 429 does not count; an explicit manual request bypasses suppression.
threshold=0 disables.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The scarce resource is the per-IP timedtext caption fetch (ADR-014); the pilot's
candidate set was mostly Shorts/clips/livestreams, each burning a fetch (a "none"
result is a completed fetch — it costs budget even when it yields nothing).
NewVideos now enriches candidates with one cheap Data API videos.list call
(contentDetails.duration + snippet.liveBroadcastContent — the quota API, a
DIFFERENT limit from the timedtext 429) and drops, before returning: videos
shorter than TAPIR_MIN_VIDEO_SECONDS (default 60) and any live/upcoming
broadcast. Dropped videos are never persisted, so the list declutters too.
Degrade-open: MinVideoSeconds=0 disables it (no quota call); a videos.list error
returns candidates unfiltered so discovery never breaks on a metadata hiccup. The
paste-a-URL path (VideoByID) is not filtered — an explicit request is honoured.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The lead-user rotation rotated the full ListAllUsers set, so a connectionless
orphan identity ate a rotation slot — collapsing onto the next real user and
skewing the lead share (two real users got 2/3 vs 1/3 instead of 50/50). Filter
to connected users BEFORE rotating so the rotation is over exactly the users that
consume the caption budget. A dead identity can no longer skew fairness.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Caption fetches share one per-egress-IP rate budget; whoever runs first each pass
spends the pre-throttle window before YouTube starts 429ing. ListAllUsers order
is unspecified and was stable, so the last-listed user was permanently starved —
a friendly-pilot user got 0 fetches in 12h (all rate_limited) while the
first-listed user got every successful fetch. rotateUsers left-rotates the user
order by pass index so each user leads 1/N passes and the lead slot is shared.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
buildUserRunner built the engine without engine.Transcripts = st, so the
scheduler — unlike the web "Summarize now" path — never read or wrote the shared
transcript cache. Every discovery pass re-fetched transcripts it had already
fetched, burning the scarce per-egress-IP caption budget (ADR-014) on redundant
work and starving other users' first-time fetches. The transcripts table was
empty despite summaries existing. Wire the cache on this path too.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first friendly-pilot live run produced zero summaries: koala/phi4-mini hit
three silent failure modes — 8k context overflow on long transcripts (HTTP 400),
intermittent malformed JSON (highlights as a bare string), and no fallback wired
at all (summarizer.New(primary, nil)).
Keep phi4-mini as the fast primary and add resilience around it:
- Ordered endpoint chain (summarizer.NewChain): phi4-mini → koala/phi4-14b
(local) → berget/mistral-small (worst-case external). All reached through the
one LiteLLM gateway by alias.
- A parse failure now advances the chain like a transport error — the old
Primary→Fallback shape returned the parse error without trying anyone else.
- Tolerant parse: highlights/takeaways coerce string→[]string, absorbing the
common small-model quirk without spending a fallback round-trip.
- Transcript truncation (TAPIR_MAX_TRANSCRIPT_CHARS=18000) prevents the overflow
rather than recovering from it; validated to fit phi4-mini's 8k window.
- Bounded completion budget (TAPIR_SUMMARY_MAX_TOKENS=1500) — the old 8192 budget
itself contributed to the overflow.
Local-first guarantee preserved by ordering: external endpoint is tried only
after every local one fails. TAPIR_CLOUD_FALLBACK_MODEL="" disables it entirely
for client/NDA deployments.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The engine now resolves transcripts store-first: a stored transcript —
including a stored SourceNone — is summarized without touching YouTube,
so re-analysis never re-fetches. On a miss it fetches through the source
(caption call still gated, ADR-014) and persists the terminal outcome for
the next analysis by any user. A transient SourceRateLimited is surfaced
to the runner for per-user backoff but never cached, so persistence can
never mask a 429 as a permanent "no transcript".
The TranscriptStore is optional (nil → fetch every time), keeping the
pure-core and scaffold wiring valid. cmd/tapir wires the store as both
summary sink and transcript cache, so `tapir run` and the web summarize
path (incl. paste + onboarding) all share the dedup.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
5c: app.Fetcher = a per-user YouTube videoFetcher, so POST /paste mounts and
resolves arbitrary-video metadata (Feature 2 goes live).
6: the connect trigger now runs an onboarding burst after discovery — summarize
up to TAPIR_ONBOARD_SUMMARIZE_COUNT of the user's newest unsummarized videos via
the gated Processor (Feature 1). Hard cap; explicit so it bypasses recency; every
fetch still through globalFetchGate. No-op when count=0 or queue-only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A newly connected account showed no videos until the next 2h scheduled pass —
the gap that made onboarding look broken (a second user connected, saw nothing,
read as failure). The connect callback now fires an out-of-band discovery pass
for the connecting user, so videos appear promptly.
Concurrency: scheduled and connect-triggered passes share one lock (serialize),
preserving the single-fetcher invariant (ADR-018). A trigger interleaves between
the scheduler's per-user passes rather than fetching concurrently or waiting for
a whole pass. The trigger runs on the server ctx (survives the redirect) and is
non-blocking for the request goroutine.
Scope: connect-trigger only. The optional login-refresh / "Discover now" button
from #6 are intentionally not built — an unconditional login hook risks 429
storms (per the ticket's own recommendation); defer until wanted.
TDD: TestCallbackTriggersDiscovery, TestSerializeRunsOneAtATime,
TestDiscoveryTriggerEnqueueRunsUser; new BDD scenario mapped.
Refs #6
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The scheduler enumerates every user_identities row (ListAllUsers) and ran a
discovery pass for each — including users who never connected a video source.
Their per-user runner then tried to resolve a YouTube refresh token that was
never minted, logging a spurious "secrets: ref not found:
youtube/<uid>/refresh_token" every tick (e.g. stale Dex-era orphan identities
left by the Authentik migration).
Skip users whose ConnectionsForUser is empty before running their pass. Removes
the recurring noise — which actively misled a debug session into thinking a
healthy onboarded user was broken — with no change to connected users.
Refs #7
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stop the un-summarized back-catalogue from burying the readable summaries
(UX review B3/B4). One feed, with a noise-collapse — not sections:
- Summarized + recent un-summarized videos lead inline as cards.
- Un-summarized videos older than the recency window collapse into a single
"Show N older videos — summarize on demand" disclosure (they will not
auto-fill; they are manual-only). Window comes from App.RecencyWindow
(= cfg.AutoSummarizeWindow); 0 disables the collapse (all inline).
- Caption-less videos collapse into one honest line ("N videos have no
captions and can't be summarized") instead of N dead terminal cards.
bucketRows is a pure classifier (cutoff-driven; undated rows never age out);
App gains RecencyWindow + an injectable clock for the cutoff.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
In automatic mode the scheduler now only summarizes videos published within
TAPIR_AUTO_SUMMARIZE_WINDOW (default ~7d). Older videos are still discovered
and listed — they keep the manual "Summarize" affordance — but are not
auto-processed, so a large back-catalogue (the maintainer's ~256-deep queue)
stops self-inflicting 429s against the per-IP caption gate each cycle (UX
review B1, recency design).
- runner.WithAutoWindow + Stats.SkippedTooOld; tooOld() treats a zero window
as disabled and an undated video as never-aged-out (processed, not stranded).
- An explicit manual request bypasses the bound even in auto mode (requested
videos are loaded in auto mode when a window is active).
- Wired through cmdRun, the scheduler's per-user runner, sumStats, and pass
logging. config: TAPIR_AUTO_SUMMARIZE_WINDOW (default 168h), .env.example.
- Account copy (A7) updated to match: "Automatic summarizes new videos from
about the last week; older videos stay browsable — summarize on demand."
The rate gate is untouched; the manual path still serialises through it. This
bounds auto LOAD, it does not fetch harder.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Authentik owns invites now (infra ADR-0001). Delete adapters/dex, the
/invite set-password UI, the tapir invite CLI, the InvitationStore/
DexPasswordCreator ports + App wiring, the invite Templ pages, and the
invite Taskfile target. New users are invited via Authentik, log in via
OIDC, and hit the existing /register gate. invitations table (mig 009)
left in place (append-only; harmless). task check green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
