[codspeed]

Summary

This PR sets up continuous performance tracking with CodSpeed. It adds a pytest-codspeed benchmark suite and a GitHub Actions workflow that runs it on every push to main and on every pull request.

What changed

• benchmarks/: a new pytest benchmark suite with 21 benchmarks. It lives outside tests/, so python -m unittest discover -s tests does not pick it up.
  • test_bench_media_shrinker.py covers the CPU-bound paths of the CLI:
    • parse_silencedetect_intervals on 100 and 5,000 silence logs
    • build_segments with silence-aligned splits and with hard splits only
    • _parse_probe_payload on a multi-stream ffprobe payload
    • build_audio_plan for FLAC, Opus, AAC and MP3 across every segment of a 12h recording
    • calculate_audio_bitrate
  • test_bench_transcripts.py covers transcript post-processing:
    • SRT and WebVTT rendering
    • TranscriptIndex build, search and tokenization
    • diarize.merge_with_transcript
    • chapters.detect_chapters and chapter serialization
    • summarize.summarize_segments
  • All inputs are synthetic, deterministic and built in memory. Nothing calls ffmpeg, the filesystem or the network, so results stay stable under CodSpeed's CPU simulation instrument.
• benchmarks/requirements-bench.txt: hash-pinned pytest and pytest-codspeed, compiled from requirements-bench.in. It follows the same pattern as fuzz/requirements-fuzz.txt. The benchmarked modules are stdlib-only, so the job does not need the full runtime lock.
• .github/workflows/codspeed.yml: runs the suite in simulation mode on ubuntu-latest and authenticates with OIDC (id-token: write), so no secret is needed. It follows the conventions of the existing workflows:
  • action pinned by SHA (CodSpeedHQ/action v5.2.1)
  • persist-credentials: false
  • the same concurrency group and paths-ignore rules
  • workflow_dispatch, so CodSpeed can backfill baseline data
• tests/test_workflow_security.py: now also checks the new workflow, so its checkout step keeps persist-credentials: false.
• README.md: adds the CodSpeed badge and a short "Benchmarks" section.

Running locally

pip install --require-hashes -r benchmarks/requirements-bench.txt
python -m pytest benchmarks --codspeed

All 21 benchmarks passed locally under codspeed run --mode simulation.

Next steps

• Once this is merged, the first run on main sets the baseline. After that, each PR gets a performance report comment.
• Optionally, turn on performance checks in the CodSpeed dashboard so a regression can block a merge.
• The Rust backend (rust-core) is mostly filesystem work (inventory, hashing, staging). If you want to track it too, the next step is a divan benchmark via codspeed-divan-compat, run in walltime mode on CodSpeed macro runners.

[coderabbitai]

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: b118e593-5794-48ee-9d0e-05ae2f1778ac

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[codspeed]

Congrats! CodSpeed is installed 🎉

🆕 21 new benchmarks were detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

Detected benchmarks

• test_build_audio_plan_for_segments[aac]: 5.9 ms
• test_build_audio_plan_for_segments[flac_auto]: 5.8 ms
• test_build_audio_plan_for_segments[mp3]: 6 ms
• test_build_audio_plan_for_segments[opus_auto]: 8.9 ms
• test_build_segments_hard_splits: 11.7 ms
• test_build_segments_with_silences[10min_windows]: 40.5 ms
• test_build_segments_with_silences[4h_windows]: 4.9 ms
• test_calculate_audio_bitrate: 3.9 ms
• test_parse_probe_payload: 12.2 ms
• test_parse_silencedetect_intervals[100_silences]: 2.3 ms
• test_parse_silencedetect_intervals[5000_silences]: 112.9 ms
• test_chapters_serialize: 39.5 ms
• test_detect_chapters: 50.7 ms
• test_diarize_merge_with_transcript: 2.9 s
• test_subtitles_render[srt]: 76.6 ms
• test_subtitles_render[vtt]: 75 ms
• test_summarize_segments: 79.4 ms
• test_tokenize: 52.5 ms
• test_transcript_index_build: 434.9 ms
• test_transcript_index_search[multi_term]: 4 ms
• ...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---