fix(recording): wire PGM video and AAC audio into the recorder

[wagenet]

Opened by Claude (AI) on behalf of @wagenet.

A production with a recording output activates, but writes no recording. The recorder has two broken inputs:

• Video. The generator links PGM to <recorder>:video_in, but builtin.recorder numbers its inputs from zero: video_in_{i} / audio_in_{i} (Strom upstream/main 8079d581, backend/src/blocks/builtin/recorder.rs:491, :501; static definition :1665, :1672). POST /api/flows drops the link as stale (backend/src/api/flows.rs:436), so the recorder is built with video track 0 unconnected. This PR links to video_in_0.
• Audio. The recorder refuses raw audio (backend/src/blocks/builtin/refusal.rs:65: "Recorder needs AAC, MP3, AC-3, DTS or Opus audio, but its audio input is raw"), and the audio mixer's main_out is raw. This PR routes main_out through a builtin.audioenc (codec: aac) into the recorder's audio_in_0.

builtin.audioenc first shipped in Strom 0.6.9, and an older Strom refuses to start any flow that references it. Open Live doesn't pin a Strom version, so activation checks GET /api/blocks first. If the block is missing, it leaves the encoder out and the production goes on air recording picture only. It also saves a recording-no-audio warning on the production (activationWarnings) and logs it. Once the production is live, the warning goes out as a WS ERROR frame, and again to each controller that connects while it stays live. Studio shows that frame as a toast. docs/controller-websocket.md documents it. Eyevinn/open-live-studio#161 tracks a persistent display. If the block list can't be read, activation keeps the encoder.

With only the video fix, the recorder writes a full-length MP4 with no sound, and deactivate uploads it and lists it as a normal recording. Both fixes are needed for a usable file.

The existing test asserted the wrong video pad name, which is why CI passed. It now asserts video_in_0 and an AAC encoder between main_out and audio_in_0, or no encoder plus a warning when Strom lacks the block.

Property rename and comments

The recorder's property is filename_prefix (recorder.rs:546, definition :1605). Strom ignores the unknown prefix, so files were named recording_…. They are now <production id>_…. The output dir and upload key already carry the production id, so nothing downstream changes. The filename-template comments in config.ts, flow-generator.ts and recording-uploader.ts now match Strom's {filename_prefix}_{timestamp}_%05d.{ext} (recorder.rs:638).

Testing

• pnpm typecheck and pnpm test (709 tests) pass, rebased onto current main.
• Reverting any of the three recorder fixes makes recording-wiring.test.ts fail on the matching assertion. Removing any of these fails its new test: the block check, saving the warning, sending it when the production goes live and not after an aborted activation, the connect-time frame and its live-only gate, or clearing the warning on re-activation and on failed activation.
• Live, against Strom upstream/main 8079d581 (dev build, headless). I posted the flow the generator produces, then ran the deactivate sequence (splitNow, then uploadRecordings with the S3 PUT stubbed) against the running recorder. A 10 s run gives:

Flow	Uploaded file
before	none (video link pruned, audio refused)
video fix only	H.264 only, 8.03 s, no sound
this PR	H.264 + AAC, 8.09 s, no pipeline error

Against Strom 0.6.7 (eyevinntechnology/strom:latest, no builtin.audioenc), activation succeeds, returns the recording-no-audio warning, and records H.264 only (6.03 s). Deactivate uploads that file. Against a Strom with the block, there's no warning and the file is H.264 + AAC.

A probe recorder on the encoded PGM pad received the same bytes (25,000,052 at 20 s) with this PR's recorder as with no recorder, so the added branches do not stall program output.

Not tested: a real MinIO upload, a production with real participants, or the studio toast end to end.

Known issue, fixed in #359

Deactivate uploads everything in recordings/<production id>, which is shared across activations, then deletes each uploaded file from Strom (#378). A file whose delete fails, or is skipped because the production changed during the upload, is uploaded again at the next deactivate and inserts a second recording document. Now that recordings are written, this is reachable. #359 fixes it and is built on this branch, so it should merge after this PR.

Possible follow-ups

• No AAC encoder on a new Strom. On Strom 0.6.9 or later, a host with none of the AAC encoders Strom tries (fdkaacenc, faac, voaacenc) fails flow start with "No encoder available for Aac", so recording can still keep a production off air. The Strom Docker image and Homebrew GStreamer both have one. The fix would be to retry the start without the encoder, plus the same warning, when it fails for that reason.
• Sound on Strom older than 0.6.9. Plain GStreamer elements in place of builtin.audioenc (audioconvert → audioresample → voaacenc → aacparse from main_out to audio_in_0) started on Strom 0.6.7 and recorded H.264 plus audible AAC. That was one 6 s run: the audio came out at 44.1 kHz with no caps filter, A/V sync wasn't checked, and it needs voaacenc on the host. It could replace the picture-only fallback after proper testing.

How to review

1. src/lib/flow-generator.ts recording branch: the video_in_0 link, the audio encoder and its block check (stromHasAudioEncoder), and filename_prefix.
2. src/routes/productions.ts and src/ws/controller.ts: saving, sending and clearing activationWarnings.
3. src/__tests__/recording-wiring.test.ts: the assertions for each.
4. Other builtin.* links in the generator were checked against Strom's definitions. They match, except the findPgmFeedPad fallback (flow-generator.ts:98), which returns <Enc PGM>:video_out. builtin.videoenc exposes encoded_out. That fallback only runs when the template has no builtin.mpegtssrt_output, and the default template has one. Not changed here.

🤖 Generated with Claude Code

[birme]

This PR looks like the base of a two-PR dependency stack:

• fix(recording): wire PGM video and AAC audio into the recorder #358 (this PR) — wire PGM video + AAC audio into the recorder
• fix(recording): upload each recording once, including failed and orphaned sessions #359 — upload each recording once (its own body notes it "is only reachable with fix(recording): wire PGM video and AAC audio into the recorder #358")

Holding off on automated review/merge here — this needs a human (or the relevant dev agent) to assess the whole stack together, not a per-PR pass, and both PRs are drafts. Nothing further is needed from the author; this comment exists so the dependency is recorded on the PR rather than in a run log.

[birme]

Heads up: main has moved (several recording/ws fixes merged), and this branch now reports a real
merge conflict with main (mergeable: CONFLICTING) — not mere staleness, so an automated
branch refresh won't resolve it. Please merge current main into this branch (or rebase) and
resolve the conflicting files, then push. This PR is otherwise still held for holistic review as
part of its dependency stack (see the earlier stack note); this comment only records the new
conflict so it isn't lost.

[wagenet]

@birme I've rebased this. Also, to be clear, this PR stands on its own and can be merged without #359. However, #359 does depend on this one.

[birme]

This PR is part of wagenet's multi-PR recording feature line on open-live. Holding off on automated review/merge — this needs a human (or the relevant dev agent) to assess the whole set together, not a per-PR pass, because these touch the same recorder / deactivate-upload code and are merge-order-sensitive (e.g. #358 and #359 conflict in one test hunk).

Members, in dependency order:

• fix(recording): wire PGM video and AAC audio into the recorder #358 — fix recorder video pad + AAC audio wiring (base fix)
• fix(recording): upload each recording once, including failed and orphaned sessions #359 — upload each recording once (base fix; merges independently of fix(recording): wire PGM video and AAC audio into the recorder #358 but conflicts with it in a test hunk)
• feat(recording): record inputs chosen per source assignment #411 — record inputs chosen per source (Closes Record arriving inputs, chosen per source #410; stacked on fix(recording): wire PGM video and AAC audio into the recorder #358 + fix(recording): upload each recording once, including failed and orphaned sessions #359; draft)
• feat(recording): record through Strom's Live Recorder #489 — record through Strom's Live Recorder (stacked on feat(recording): record inputs chosen per source assignment #411; draft)
• feat(recording): record to Strom's disk without object storage #404 — record to Strom's disk without object storage (independent recording fix, same upload area)
• fix(recording): date swept segments by their own session's start #407 — date swept segments by their own session's start (independent recording fix, same upload area; already APPROVED)

Nothing further is needed from the author; this comment records the dependency on the PRs themselves rather than in a run log.

[birme]

pr-reviewer: APPROVE

Prior findings

#	Source	Finding	Status
1	birme stack-flag	Part of wagenet recording stack (#358→#359, #411/#489 stacked); hold for holistic review	DROPPED — automation hold note, not a code finding. Author confirms #358 stands alone; #359/#411 depend on it, not the reverse.
2	birme branch-blocked	Branch reported CONFLICTING with main	FIXED — author rebased; now MERGEABLE (BEHIND only, no conflict).

No formal reviews/comments exist on the PR. No OPEN findings.
No Closes #N in the body.

Verification against real source (fabrication check — Strom 8079d581)

Grepped Eyevinn/strom at the cited commit — every pad/property/element the generator links is REAL:

• builtin.recorder numbers inputs from 0: recorder.rs:491 format!("video_in_{}", i), :501 format!("audio_in_{}", i). The PR's video_in_0 / audio_in_0 are correct; the old video_in dropped the link. (Note audio_in_0 was already present pre-PR — only video was wrong, an internal inconsistency that confirms the bug.)
• filename_prefix property: recorder.rs:545-546, :1605. Old prefix was silently ignored. Filename template comment {filename_prefix}_{timestamp}_%05d.{ext} matches recorder.rs:648 (the code path, authoritative over the slightly older header comment on :33).
• builtin.audioenc: block id audioenc.rs:336; codec enum property default aac :342-353; pads audio_in :421 and encoded_out :428. The PR's { codec: 'aac' } + audio_in/encoded_out wiring are correct.
• refusal.rs:65 confirms the recorder refuses raw audio and names builtin.audioenc as the fix — the raw main_out→audio_in_0 link was genuinely refused, so routing through the encoder is required.
• stromHasAudioEncoder uses blocks.list() → BlockListResponse.blocks: BlockDefinition[], and BlockDefinition.id (strom.ts) exists, so b.id === 'builtin.audioenc' is sound; catch → true degrades safely.

The other output branches (whep_output, rtmp_output, mpegtssrt_output) legitimately keep :video_in — different block types, not recorders — so the scoped change is correct and complete.

Tests

recording-wiring.test.ts asserts video_in_0, audio_in_0, the AAC encoder (codec: aac, audio_in→encoded_out), filename_prefix, and the picture-only + recording-no-audio warning path when Strom lacks the block. Warning plumbing (save on activationWarnings, WS ERROR frame on go-live gated by !signal.aborted, connect-time re-send gated by status==='active', clear on re-activation and on failed activation) is covered.

CI

build-and-test PASS (1m3s). No lint script in this repo (per code-quality.md).

No secrets. Co-author trailer present on commits.

Safe to merge standalone on current main: yes — MERGEABLE, only BEHIND (no conflict). #359 is built on top and should merge after.