Weekly upstream sync 2026-05-12 (2 commits) - #7
Open
loganbronstein wants to merge 3 commits into
Open
loganbronstein wants to merge 3 commits into
loganbronstein wants to merge 3 commits into
Conversation
…ision flake (grandamenium#385) createTask generated ids as `task_${Date.now()}_${randomDigits(3)}`. With only 3 random digits, two createTask calls landing in the same millisecond had a ~1/1000 chance of colliding. When they did, the new task's id matched its own declared blocker — detectCycleOrThrow then threw "X ultimately blocks itself via X" and the test failed nondeterministically. CI run 25618845172 (https://github.com/grandamenium/cortextos/actions/runs/25618845172) hit exactly this case in tests/unit/bus/task.test.ts:761 ("once the dependent completes, the blocker becomes eligible"), which calls createTask twice back-to-back. The branch under test (feat/voice-transcription) does not touch src/bus/task.ts or its tests — this is a pre-existing flake. Bumping to randomDigits(8) drops collision probability to ~1e-8 per same-ms call pair. The id-format assertion in task.test.ts:40 was tightened to match. The looser regex in codex-bus-roundtrip.test.ts (/[a-z0-9]+/) is unaffected. Co-authored-by: Boris <boris@cortextos.ai>
…ine (grandamenium#384) * feat(telegram): wire whisper-cli voice transcription into media pipeline Voice .ogg files are now transcribed locally via whisper-cli + ffmpeg the moment they're downloaded in src/telegram/media.ts, threaded through ProcessedMedia.transcript, and surfaced to agents via formatTelegramVoiceMessage. Previously the framework downloaded voice notes but never produced a transcript — agents received only the .ogg path and had to invoke whisper themselves (which most templates did not), so voice messages effectively reached agents as "[voice message]". Behaviour: - Transcription runs best-effort with graceful fallback. Any failure (missing whisper-cli, missing ffmpeg, missing model, timeout) returns null and the agent still receives the .ogg path — no regression vs. the previous behaviour. - Defaults to ggml-tiny.en.bin (~75MB) at ~/.cortextos/models/. Override with CTX_WHISPER_MODEL, CTX_WHISPER_BIN, CTX_FFMPEG_BIN. Disable entirely with CTX_TELEGRAM_NO_TRANSCRIBE=1. - scripts/install-whisper-model.sh downloads the GGML model idempotently; intentionally separate from npm install so the bundle stays slim. whisper-cli + ffmpeg come from `brew install whisper-cpp ffmpeg` on macOS. Tests: 7 new unit tests in tests/unit/telegram/transcribe.test.ts covering env disable, missing inputs, missing binaries, and timeout. Existing media tests updated to set CTX_TELEGRAM_NO_TRANSCRIBE=1 so they don't shell out. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(install): auto-install whisper-cli + ffmpeg + GGML model Extends `cortextos install` to provision the voice-transcription stack that PR 384 wires into the media pipeline. New agents now get voice transcripts out of the box on macOS without a separate manual step. Behaviour: - whisper-cli auto-install via `brew install whisper-cpp` (mac). Linux/Windows print build/install instructions; install.ts continues. - ffmpeg auto-install via brew (mac), apt-get (linux), or winget/choco (windows). - GGML model fetched via the existing scripts/install-whisper-model.sh (idempotent, skips if present). - All three steps are best-effort: any failure prints a hint and lets install.ts continue. Voice messages still flow as path-only when the stack is incomplete — transcribe.ts already returns null gracefully. This complements PR 384 by making "default voice transcription for all agents" a real default rather than something each operator wires up manually after the fact. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Boris <boris@cortextos.ai> Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
This was referenced Jul 5, 2026
This was referenced Aug 2, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Weekly upstream sync (2026-05-12)
Upstream
grandamenium/cortextosadded 2 commit(s) since the last run.Each one is summarized in plain English below.
What's new
New feature in telegram: wire whisper-cli voice transcription into media pipeline (feat(telegram): wire whisper-cli voice transcription into media pipeline grandamenium/cortextos#384)
4815617(grandamenium@4815617)Bug fix in task: bump random suffix from 3 to 8 digits to eliminate ID-collision flake (fix(task): eliminate task-ID collision flake (3→8 random digits) grandamenium/cortextos#385)
f8eec59(grandamenium@f8eec59)How to merge
merge allto take everything.merge 1 3 5(numbers from the list above) to cherry-pick those.skipto dismiss this whole PR.This PR was opened automatically by boss every Sunday at 6:30pm Chicago time.