Low findings of the PR #62 re-review: the WithTimeout test's comment claimed cancellation
travels on the context but only the language was asserted (the WithoutCancel mutant survived),
and PlanFor('') and PlanFor(unknown) were untested.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
PlanFor(lang) on the Ensemble and a LanguagePlanner the worker prefers: the page lists
the other engine (and the fallback if configured), not the base engine, the comparison and
the decisions as coming. Found by the PR #62 review.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The wrapper sits in front of both local engines; a fresh context there would send an auto
upload as language=sv and un-cancel the job, and no test noticed. Found by the PR #62 review.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Form field (Swedish preselected, Other = auto) read before the file part and refused
before anything is spooled if outside the closed set; stored on the row (migration 021,
CHECK constraint, default sv for history); carried on the Job and the worker's context;
shown on the status page. Feature scenarios and the coverage map are updated, and
ADR-033/034 amended.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The other engine transcribes alone, then the cloud fallback if it fails and the job is
still live. A cancelled job never reaches the fallback (mutation-checked).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The request used to hard-code language=sv for every upload, so a recording in another
language was decoded as Swedish. The language now comes from the context; no language in
the context is still sv.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Only the 8-byte override was exercised; >= vs >, a huge default and a non-positive option
all stayed green. Also says why the error-body read ignores its error. Found by the PR #58
review.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A 900-stream file cost ~570 MB RSS in each tool; with the cap the same file fails
in ~18 MB and is reported as unreadable.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Upload a recording (audio, or video with the camera off), get a private Swedish transcript, and chat about it.
- ADR-033: RLS-scoped private transcript store; ffprobe/ffmpeg intake; gateway transcription (iguana kb-whisper, berget fallback); in-process single-replica worker; chat over the transcript (local models only).
- ADR-034: optional two-engine mode (pianissimo + kb-whisper) reconciled by a LOCAL model; off unless TAPIR_RECONCILE_BASE_MODEL is set.
- ADR-035: real progress (phases, running/quiet/lost), phase logging, every judged spot stored, merged segments.
- Hardening from four adversarial reviews: errors logged and stored as kinds never text; per-engine deadlines; converted-length check; bounded queue; client hang-ups are not errors; strict local-alias check.
- Dockerfile: static ffmpeg/ffprobe 7.0.2 pinned by sha256 (#34). Built and inspected by hand; CI does not build the image on a PR, so its first CI build is this merge.
Uploads stay OFF until TAPIR_UPLOADS_ENABLED is set in the deployment.
Closes#33, #34, #42, #43, #44.
Co-authored-by: Mathias <mthbqv@gmail.com>
The up/down migration tests stepped a hard-coded number of Steps(-N)/Steps(+N)
down from HEAD and back. The counts assumed a specific latest migration, so
adding one shifted every count by one and unrelated tests (010/011/014) went
red with confusing off-by-one symptoms — a papercut on every new migration.
Drive the schema to an exact version with m.Migrate(version) via two helpers
(headVersion, migrateTo). Each test now steps to just below its target by
version, asserts the down effect, steps up to the target, asserts the up
effect, then restores to the captured HEAD. A migration added on top changes
HEAD but shifts no count, so no test needs editing.
Verified by adding a throwaway migration 017 on top: all four tests stayed
green with zero edits.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015QbdxXWxLefS5AwLN5eyze
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>
WithUsageHook callback fires with model + prompt/completion tokens parsed from the
response usage block. Keeps the copied stdlib-only llm package decoupled from
metrics (ADR-004) — the caller wires it to internal/metrics. TDD covered.
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.
Newest-first but quality-aware: excludes a video when its duration is KNOWN and
outside [minSeconds, maxSeconds], dropping Shorts and multi-hour livestream VODs
that waste a scarce caption fetch on a poor first impression. NULL/unknown
duration is kept (degrade-open) but ranked after known-good rows. 0/0 bounds
disable the filter (pure newest-first, the reversibility lever). RLS-scoped.
NewestUnsummarizedVideoIDs is left in place for callers that want pure-newest.
ADR-023's filterLowValue already fetches each candidate's duration via the
cheap videos.list quota call to drop Shorts/live, then threw it away — the
videos.duration_s column (migration 001) was never written. Carry it onto the
kept domain.Video and have UpsertVideo persist it, COALESCE-preserving a known
value so an unknown (0) re-upsert never clobbers it (the channel_title backfill
stance, migration 014). This is the enabling change for length-aware burst
selection. No new migration — the column already exists.
A read-only deeper-dive over a video's already-stored transcript (ADR-021).
Safe by construction: the Service has no VideoSource and no caption-fetch
dependency — only a Completer factory over the existing LiteLLM gateway — so it
cannot reach YouTube or the rate gate. Reuses the summarizer's truncation
discipline (TAPIR_MAX_TRANSCRIPT_CHARS), reporting the cut so the UI can be
honest about a bounded transcript. Model defaults to the summary's model and is
switchable among an offered, local-first list; an un-offered alias is forced
back to the default so chat can never call the gateway with an arbitrary model.
Ephemeral: history is carried per-request, nothing persisted.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A push to main and its version tag fire two CI runs for the same commit. Both ran
`go test ./...`, which starts embedded-postgres on a FIXED port (54329/54330) and
a shared data dir. -p 1 serialises packages WITHIN a run, not across two
concurrent runs — so when the two runs overlapped they collided on the port/data
dir and BOTH failed the Lint/Test job (no image built). Prior commits passed only
because their two runs happened not to overlap.
Derive the port and runtime/data dirs from the PID; share only CachePath so the
PG archive downloads once. Proven: two concurrent `go test` of the store package
now both pass. Unblocks the v0.21.0 (Pillar A) build.
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>
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 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>
Reshape the dead per-user transcripts table (PK videos.id, user_id,
RLS-FORCEd — never read or written by app code) into the shared public
caption store ADR-021 specifies: keyed by (provider, provider_video_id),
no user_id, NOT RLS-scoped. Migration 015 (reversible). Add
ports.TranscriptStore + Store.GetTranscript/SaveTranscript via the raw
pool (no withUser): public content, shared across users by construction.
SaveTranscript persists only terminal outcomes (captions/none) and
refuses SourceRateLimited so a transient 429 can never be stored as a
false permanent absence (ADR-014).
Flip the isolation proof: transcripts leaves the RLS-scoped set;
TestTranscriptsTableIsSharedNotRLS asserts it is the SINGLE non-RLS
surface (writable/readable with no user scope, no user_id column, RLS off
on it alone, still on every user-owned table) — the proof the
public-content classification was applied exactly here and leaked nowhere.
appPool made idempotent so two tests can build it. Adjust the 010/011/014
up-down migration tests for the new HEAD. account.go: user deletion no
longer strips shared transcripts. Reconcile data-model.md + CLAUDE.md.
Wiring the engine to read-stored-first is the next commit.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The free-text 'channel' filter was dead: it exact-matched SummaryRow.Channel,
which is just the provider ('youtube'), because videos never stored their source
channel. Now they do.
- migration 014: videos.channel_title (nullable; existing rows backfill on the
next discovery pass, pasted videos immediately).
- discovery (NewVideos) + paste (VideoByID) populate channel_title; UpsertVideo
persists it, preserving an existing title when an update arrives empty.
- store.DistinctChannels lists a user's channels (RLS-scoped); SummaryRow carries
ChannelTitle via the shared projection.
- Filter: single Channel -> Channels []string, matching on ChannelTitle; the feed
renders a multi-select of DistinctChannels (hidden until channels exist).
- migrate tests: 014 reversibility + fixed the relative-step counts in the 010/011
up/down tests (014 shifted the topology).
TDD throughout: channel persist + distinct, adapter channel wiring, multi-channel
filter match, handler channel filter, migration up/down.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
POST /paste: parse the video id, fetch metadata via the VideoFetcher port (Data
API, ungated), upsert a subscription-less row scoped to the user (idempotent =
dedup), and — unless already summarized — RequestSummarize + start immediate
processing through the SAME globalFetchGate as the Summarize button. Explicit
paste overrides the recency window; a captionless video degrades to the honest
'no transcript' terminal state via the engine (ADR-010). Invalid URL -> 400,
not-found -> 404, both add nothing. Route mounts only when a Fetcher is wired.
Moves the video-not-found sentinel to domain (shared by adapter + web, no
cross-adapter coupling). Tests: valid add+queue, invalid, not-found, dedup.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
videos.list (part=snippet) for a single id, including channels the user does
not follow. Data API call (1 quota unit), NOT the rate-limited caption path —
ungated metadata; only the later transcript fetch hits globalFetchGate. Returns
a subscription-less domain.Video scoped to the user, or ErrVideoNotFound for a
deleted/private/typo'd id. Foundation for Feature 2.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Returns up to limit of a user's newest videos (published_at DESC, NULLS LAST)
that have no summary yet. RLS-scoped via withUser — the test proves a second
user's newer video never leaks. Drives the connect-time onboarding burst
(Feature 1); the caller routes each through the shared rate gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ListVideos now sorts summarized-first, then published_at DESC with undated
videos last, then seen_at DESC as a tiebreak (was seen_at only). Aligns the
list with the recency framing — newest content surfaces first — so the
recency-bounded feed reads coherently (UX review B2).
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>
Tapir used human-readable substitutions ('@' -> '-at-', '.' -> '-dot-') when
deriving the Password CR name from an email. Dex's internal passwordID() maps
every non-[a-z0-9-] character to plain '-'. This caused a name mismatch:
Tapir wrote the CR as 'mathias-at-d-ma-dot-be', Dex looked it up as
'mathias-d-ma-be', got not-found, and returned 'Invalid credentials' on every
invite login — while static configmap passwords (a different code path) worked
fine. Diagnosed by adding the email to staticPasswords and confirming login
succeeded, proving the kubernetes CR lookup was the failure point.
The original NOTE claimed Dex's kubernetes storage types Hash as []byte,
requiring the bcrypt string to be base64-encoded before storage. This was
wrong: Dex v2.41 stores and compares the hash field as a plain string. The
base64-encoding caused every invite login to fail with 'Invalid credentials'
because Dex passed the base64 bytes (starting with 'J' not '$') directly to
bcrypt. Static passwords in the configmap always used raw bcrypt strings and
worked fine — confirming the dynamic CR encoding was the bug.
Three UX improvements for the pending-transcript state:
1. Summarized videos sort to top (ORDER BY (s.id IS NOT NULL) DESC, seen_at DESC)
so completed summaries are always immediately visible without filtering.
ListVideos default limit raised from 50 to 500 to show the full backlog.
2. Pipeline stats bar above the video list: '2 summarized · 256 fetching soon · 12
no captions' — computed from the unfiltered row set, hidden when everything is
summarized.
3. 'Try now' button on rate-limited cards replaces the passive 'Retrying later'
chip. POST /v/{id}/retry-now clears rate_limited_at then calls ProcessVideo
through the shared globalFetchGate — same rate limiting as the scheduler, safe
under concurrent use.
YouTube channels that 404 on playlist discovery (deleted/private) are now:
1. Wrapped in domain.ErrChannelUnavailable by the YouTube adapter (instead of
a generic error), so the runner can identify them without string-matching.
2. Stored per-user in channel_errors (migration 013, RLS-guarded) via runner's
new UpsertChannelError path — removed from the generic Errors counter,
counted separately as ChannelUnavailable.
3. Shown on the account page under "Unavailable channels" with name, chip-warn
badge, and first-seen date, so users know why some subscribed channels
produce no videos.
Migration 011's UPDATE ran without tapir.current_user_id set, so FORCE RLS
blocked all rows and 0 users were updated (skipped_manual=607 in scheduler).
Migration 012 temporarily drops FORCE so the table owner can run the UPDATE,
then restores it.
Migration 011 flips the auto_summarize column default to TRUE and brings
existing rows (maintainer + current registrations) along. Onboarded friends
now get zero-friction discovery: scheduled discovery (ADR-018) both discovers
AND summarizes new videos, so a user's list fills and summarizes itself
instead of presenting an empty list of manual Summarize buttons.
Safe only because the process-wide caption-fetch rate gate (ADR-014 item 2,
prior commit) now exists — auto + scheduled + multi-user would otherwise
self-inflict 429s every cycle. The down migration reverts the default but
intentionally leaves existing rows as-is (no surprise manual regression on
rollback). RegisterUser already lets the column default drive the value, so
no app change is needed; the account-page manual toggle still works.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
mathiasandClaude Sonnet 5.5