Bound non-realtime RTSP frame buffering

[digaobarbosa]

Description

Non-realtime RTSP processing can accumulate decoded frames faster than inference consumes them; bound its decoder queue to the existing 60-frame file-source capacity and pause decoding when full, preserving frame order and timestamps.

Type of change

• Bug fix (non-breaking)

How has this change been tested?

• Four focused tests pass: backpressure, ordered delivery with real PyAV decoding, cancellation, and sanitized failures.
• Black and isort checks pass.
• RTSP transport validation in staging pending.

Any specific deployment considerations

• Inference server and WebRTC Modal worker.
• The queue bound does not impose a new resolution policy; no general RTSP pixel limit exists to reuse.

[github-actions]

👋 Thanks for the pull request! Here is how automated Claude review works here, so you spend credits (and reviewer time) wisely.

🚦 This PR is marked Ready for review, so automated Claude review will run — and every pass spends real credits.

Warning

💸 The Claude reviewer bills in credits, not vibes

Automated review spins up a real agent that reads real code and spends real credits on every pass. It is glad to help — but it is not a rubber duck, a linter you poke in a loop, or a substitute for reading the contributing guide. Treat it like an expensive senior reviewer whose time you booked, and show up prepared.

Draft when unsure, Ready when you mean it:

• 🌱 Not sure the PR is in good shape yet? Keep it (or set it back) as a draft — drafts pause review, so you can push and iterate without burning credits on a moving target.
• 💪 Feel strong about the contents? Mark it Ready for review and the reviewer will take a look.

However you get there, arrive prepared:

• 🧱 Bring a SOLID, thorough PR. Point your local agent at our skills/ to tune it to our guidelines first — or, if you are one of those fabled carbon-based contributors, read them yourself. A half-baked diff costs exactly the same to review as a finished one.
• ✅ Resolve every comment before you re-request review. Re-requesting with threads still open means paying twice for the same conversation.
• 🔁 Do not use CI review as an inner loop for a local agent. The reviewer is not a step-by-step debugger — do the unfolding locally and arrive with the answer, not the search.
• 🙋 If something looks off, ask a human. One question to a maintainer is cheaper and faster than three rounds of agent re-review chasing a misread.

Reviews are not free. A draft costs nothing to review; a Ready PR is a promise that it is worth reviewing.

• Prefer to skip automated review entirely? Add the skip-claude-review label.

[github-actions]

🤖 Claude review started at commit a110046568f472b3e402e58811b5d172fc728b3d.

New commits are not auto-reviewed. Add the claude-review label to request a re-review — the label is consumed when the review starts, so just add it again next time.

[github-actions]

Review summary

Skills: review-topic-concurrency-and-resource-safety, review-topic-test-hygiene (also considered review-topic-input-boundary-security for RTSP-URL credential handling).

No dedicated surface skill covers inference/core/interfaces/webrtc_worker/** — generic review plus topic skills only.

Scope: the new ThreadedRTSPTrack backpressure path is opt-in — webrtc_realtime_processing defaults to True (inference/core/env.py:841), so only requests that explicitly set it False with an rtsp_url reach the new code; realtime RTSP keeps the unchanged MediaPlayer.

Checked and clean:

• Bounded 60-frame queue with a stop-flag-gated daemon decode thread; the with av.open(...) container is closed on every exit path (matches the sibling ThreadedVideoFileTrack convention).
• Track stop() is reliably reached via video_processor.track.stop() in teardown and on connection failed/closed (webrtc.py:1385-1392); the player.video.stop() branch is realtime-only and correctly bypassed.
• FFmpeg failures are collapsed to a generic "Failed to decode RTSP stream", so source-URL credentials cannot leak (mirrors _open_media_player).
• PTS normalized to start at 0 with time_base preserved.
• Tests exercise the real changed path (backpressure bound, ordered lossless delivery, cancellation unblocking a full queue, sanitized failures), including one with real PyAV decoding.

No medium+ blocking findings; no open clarification questions. Not a versioned subsystem (not inference-models or Execution Engine, no public HTTP/SDK/workflow-block contract), so no changelog entry is required.

Reviewed at HEAD: a110046

[github-actions]

😎 PR passes the vibe-check and trust-me-bro verification.