feat(tts): Phase 10C — TTS generation via OpenRouter /audio/speech#22
Merged
Conversation
- Add generate_tts LangGraph tool + node (features/audio/)
- service.py: POST to openrouter /audio/speech with gpt-4o-mini-tts,
upload mp3 to Supabase Storage, return 7-day signed URL
- tool.py: extract text arg from supervisor tool_call, call service,
return ToolMessage with markdown audio link
- schemas.py: AudioGenerateResult Pydantic model
- Wire generate_tts into supervisor + orchestrator
- supervisor.py: bind generate_tts tool, add TTS section to system prompt,
detect text-format tool calls (DeepSeek compat) and convert to structured
- orchestrator.py: add generate_tts node, conditional edge from supervisor
- Config: tts_model=openai/gpt-4o-mini-tts-2025-12-15, tts_voice, audio_bucket
- State: add pending_audio field for cross-turn save confirmation (Phase 10D)
- Emit Generating audio... SSE tool_use event from chat service
- Frontend: inline <audio> player for .mp3 links in MessageBubble
- Frontend: purple Volume2 icon in ChatWindow during TTS tool_use status
- Supabase migration: audio bucket with RLS policies scoped to {user_id}/*
|
Caution Review failedPull request was closed or merged during review WalkthroughSupervisor can request TTS via a new generate_tts tool; orchestrator routes to a generate_tts node that calls OpenRouter, uploads MP3 to Supabase, stores pending_audio in state, and frontend renders audio players and a TTS status icon. ChangesText-to-Speech Feature
Sequence DiagramsequenceDiagram
participant User
participant Frontend
participant Supervisor
participant Orchestrator
participant generate_tts_node
participant AudioService
participant SupabaseStorage
User->>Frontend: send message / request TTS
Frontend->>Supervisor: user input
Supervisor->>Orchestrator: supervisor emits tool_call generate_tts(text)
Orchestrator->>generate_tts_node: route to "generate_tts"
generate_tts_node->>AudioService: generate_tts(text, user_id) -> mp3 bytes + filename
AudioService->>SupabaseStorage: upload mp3 -> signed URL
SupabaseStorage-->>AudioService: signed URL
AudioService-->>generate_tts_node: (signed_url, filename)
generate_tts_node-->>Orchestrator: ToolMessage + pending_audio payload
Orchestrator-->>Supervisor: return flow to supervisor for final response
Supervisor-->>Frontend: final message (includes pending_audio metadata)
Frontend-->>User: render audio player with signed URL
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
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. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@backend/app/features/audio/schemas.py`:
- Around line 9-11: The text_preview field in the schema lacks length
constraints; update the schema in schemas.py to import pydantic.Field and change
text_preview: str to text_preview: str = Field(..., min_length=1,
max_length=1024) (choose an appropriate max if 1024 is too large) so
user-derived content is bounded; ensure the Field import is added and leave
signed_url / filename unchanged unless they are also user-input — in that case
apply similar Field(..., min_length=1, max_length=N) constraints to those
symbols.
In `@backend/app/features/audio/service.py`:
- Around line 68-73: Wrap the HTTP call using httpx.AsyncClient (the block that
posts to f"{settings.openrouter_base_url}/audio/speech" with json=payload and
headers=headers) in a try/except that catches httpx.RequestError and
httpx.HTTPStatusError (or Exception), obtain a module logger via
logging.getLogger(__name__), log the exception with context (include URL,
payload summary or headers as appropriate), and then convert or re-raise into
the service-level error contract (e.g., raise a custom exception or return an
error result) so callers receive a consistent error; update the async function
containing this request to use that try/except around the client.post call and
ensure response handling occurs only on success.
- Around line 57-62: The payload currently sends raw user `text`; call the
privacy masker and send the masked value instead: import and use mask_pii(text)
(from core/privacy.py) before building the `payload`, assign the result to a new
variable (e.g., `masked_text`) and replace `input: text` with `input:
masked_text`; ensure any downstream uses in the same function (e.g., in the TTS
call that builds `payload`) reference the masked variable so no raw user input
is forwarded to OpenRouter.
In `@backend/app/features/audio/tool.py`:
- Around line 91-95: The success log in generate_tts_node currently logs a raw
user identifier via logger.info("[generate_tts_node] audio ready filename=%s
user=%s", filename, user_id); remove or redact the sensitive user_id from the
log (e.g., log only filename and a non-identifying status or a hashed/obfuscated
id) so that generate_tts_node no longer emits raw PII; update the logger.info
call to exclude user_id (or replace with a safe token) and ensure any callers or
tests expecting the old message are adjusted accordingly.
In `@supabase/migrations/20260512100000_create_audio_bucket_and_policies.sql`:
- Line 20: Update the RLS policy predicates to use a subquery form of auth.uid()
instead of the bare function: replace occurrences of (auth.uid()) with (select
auth.uid()) in the policy expressions (e.g., the clause comparing
(storage.foldername(name))[1] to auth.uid()). Specifically modify the
predicate(s) involving storage.foldername(name) and auth.uid() so they read like
(storage.foldername(name))[1] = (select auth.uid())::text; apply the same
replacement at the other affected predicates referenced in the comment.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: bfbcd21f-1fe5-41ad-b29f-a8ee3147a095
📒 Files selected for processing (12)
backend/app/agents/orchestrator.pybackend/app/agents/state.pybackend/app/agents/supervisor.pybackend/app/core/config.pybackend/app/features/audio/__init__.pybackend/app/features/audio/schemas.pybackend/app/features/audio/service.pybackend/app/features/audio/tool.pybackend/app/features/chat/service.pyfrontend/features/chat/components/ChatWindow.tsxfrontend/features/chat/components/MessageBubble.tsxsupabase/migrations/20260512100000_create_audio_bucket_and_policies.sql
- schemas.py: add Field(min_length=1, max_length=1024) to text_preview - service.py: mask_pii(text) before sending to OpenRouter TTS endpoint - service.py: wrap httpx call in try/except HTTPStatusError+RequestError with structured logging; use response.raise_for_status() on success path - tool.py: remove raw user_id from logger.info (PII leak) - migration: replace auth.uid() with (select auth.uid()) in all three RLS policy predicates to avoid per-row re-evaluation (Supabase best practice)
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@backend/app/features/audio/service.py`:
- Around line 92-95: The log statements currently emit the constructed filename
which includes raw user_id (see the filename variable and logger.info calls in
generate_tts and related places), leaking PII; change these logger.info calls to
omit or redact user_id — e.g., log only the file size and the generated uuid (or
a redacted filename like "<redacted_user>/{uuid}.mp3" or just the uuid string
from uuid4()), and ensure any other logger.info calls that reference filename or
user_id (the occurrences around the other noted lines) are updated similarly to
avoid logging raw user identifiers.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 9b22c5e9-f675-4f0e-bb97-97a74a4ba310
📒 Files selected for processing (4)
backend/app/features/audio/schemas.pybackend/app/features/audio/service.pybackend/app/features/audio/tool.pysupabase/migrations/20260512100000_create_audio_bucket_and_policies.sql
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
backend/app/features/audio/service.py (1)
123-137:⚠️ Potential issue | 🟠 Major | ⚡ Quick winPII leakage in storage logs (
filenamecontainsuser_id)Line 123, Line 133, and Line 137 still log
filename, which includes rawuser_id({user_id}/{uuid}.mp3).Suggested fix
async def upload_audio(audio_bytes: bytes, filename: str) -> str: @@ - try: + object_name = filename.rsplit("/", 1)[-1] + try: await client.storage.from_(bucket).upload( path=filename, file=audio_bytes, file_options={"content-type": "audio/mpeg"}, ) except Exception as exc: - logger.exception("[upload_audio] Storage upload failed filename=%s", filename) + logger.exception("[upload_audio] Storage upload failed object=%s", object_name) raise RuntimeError("Failed to upload audio to storage.") from exc @@ except Exception as exc: logger.exception( - "[upload_audio] Failed to create signed URL filename=%s", filename + "[upload_audio] Failed to create signed URL object=%s", object_name ) raise RuntimeError("Audio uploaded but could not generate a signed URL.") from exc - logger.info("[upload_audio] signed URL created filename=%s", filename) + logger.info("[upload_audio] signed URL created object=%s", object_name) return signed_urlAs per coding guidelines,
backend/**/*.py:Never log sensitive data (tokens, passwords, PII).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@backend/app/features/audio/service.py` around lines 123 - 137, The logs in upload_audio (the logger.exception calls around the storage upload and create_signed_url and the logger.info after signed URL creation) currently include filename which embeds user_id (PII); remove or replace raw filename in those log calls with a non-PII value — e.g., compute a safe identifier (use only the basename/UUID part or replace the leading user_id segment with "<redacted>") and log that instead, or omit the filename entirely in the logger.exception and logger.info calls for create_signed_url and storage upload so no user_id is emitted.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@backend/app/features/audio/service.py`:
- Around line 90-97: The code assigns response.content to audio_bytes and
returns it without validation; update the block around response, audio_bytes and
filename (where uuid4() and logger.info are used) to validate the TTS payload:
ensure response.status_code is 200, response.content is non-empty, and
Content-Type indicates audio (e.g., startswith "audio/") and/or verify MP3 magic
bytes (ID3 or 0xFF 0xFB) before constructing filename and returning; on
validation failure log a descriptive error with response.status_code and headers
(using logger.error) and raise or return an explicit error/None instead of
returning/uploading a broken file.
---
Duplicate comments:
In `@backend/app/features/audio/service.py`:
- Around line 123-137: The logs in upload_audio (the logger.exception calls
around the storage upload and create_signed_url and the logger.info after signed
URL creation) currently include filename which embeds user_id (PII); remove or
replace raw filename in those log calls with a non-PII value — e.g., compute a
safe identifier (use only the basename/UUID part or replace the leading user_id
segment with "<redacted>") and log that instead, or omit the filename entirely
in the logger.exception and logger.info calls for create_signed_url and storage
upload so no user_id is emitted.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 7196585c-9bb5-4525-adc4-7265a3e19b3e
📒 Files selected for processing (1)
backend/app/features/audio/service.py
…udio logs - Validate response content: check non-empty + content-type starts with 'audio/' or MP3 magic bytes (ID3 / 0xFF sync word); raise RuntimeError with status/content-type/size logged on failure - Replace raw filename (embeds user_id) in all upload_audio logger calls with safe_log_name (UUID basename only)
This was referenced May 14, 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
1 participant
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.
Add generate_tts LangGraph tool + node (features/audio/)
Wire generate_tts into supervisor + orchestrator
Config: tts_model=openai/gpt-4o-mini-tts-2025-12-15, tts_voice, audio_bucket
State: add pending_audio field for cross-turn save confirmation (Phase 10D)
Emit Generating audio... SSE tool_use event from chat service
Frontend: inline player for .mp3 links in MessageBubble
Frontend: purple Volume2 icon in ChatWindow during TTS tool_use status
Supabase migration: audio bucket with RLS policies scoped to {user_id}/*
Summary by CodeRabbit
New Features
Improvements