feat(tts): prioritize regional scripts over English in Sarvam TTS

[manas-narra]

• Prioritize regional languages when mixed with English words

Summary by CodeRabbit

• Bug Fixes

  • Improved text-to-speech language detection to prioritize regional Indian language scripts when present in mixed-language input.

• Chores

  • Enhanced logging filters to exclude additional binary data patterns from application logs.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

Walkthrough

Two targeted logic improvements were made: the script detection function now prioritizes regional Indian scripts over English, returning any detected regional script immediately; meanwhile, logging spam filters were expanded to exclude additional binary data patterns from specific loggers.

Changes

Cohort / File(s)	Summary
Script Detection Enhancement app/ai/voice/tts/sarvam.py	Modified detect_script() function to prioritize regional Indian scripts (Telugu, Devanagari, Tamil, etc.) over English. Now returns a regional script immediately upon detection, rather than selecting based on highest character count. Only returns English when no regional scripts are present. Improves handling of mixed-language input.
Logging Spam Filter Expansion app/core/logger/__init__.py	Extended spam-filter logic to block additional logging patterns: filters records where logger name starts with "logging" and message begins with "> BINARY" or "< BINARY". Applied across both production and development paths.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Poem

🐰 A script dance unfolds so bright,
Regional voices take their flight!
Telugu, Tamil, standing tall,
No more English drowns them all—
And filters hush the binary chatter,
Making logs that truly matter! ✨

Pre-merge checks and finishing touches

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 66.67% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: prioritizing regional scripts over English in Sarvam TTS detection logic, which aligns with the primary modification in app/ai/voice/tts/sarvam.py.

✨ Finishing touches

• 📝 Generate docstrings

---

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 and usage tips.}

[Copilot]

Pull request overview

This PR enhances the Sarvam TTS language detection to prioritize regional Indian languages over English when both scripts are present in the text, addressing the common pattern of mixing English words with regional languages in Indian speech.

Key Changes:

• Modified script detection logic to prioritize any regional language characters over English
• Added logging filters to suppress verbose binary websocket messages

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
app/ai/voice/tts/sarvam.py	Updated detect_script() function to return regional language when any regional script is detected, even if mixed with English words
app/core/logger/__init__.py	Added filters to suppress "> BINARY" and "< BINARY" log messages from websocket logging spam

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The condition if regional_script and max_regional_count > 0 is redundant. If max_regional_count > 0, then regional_script will always be set to a non-None value based on the loop logic in lines 78-81. The check for regional_script being truthy is unnecessary and can be simplified to just if max_regional_count > 0.

Suggested change

	if regional_script and max_regional_count > 0:
	if max_regional_count > 0:

[coderabbitai]

Actionable comments posted: 0

🧹 Nitpick comments (1)

app/ai/voice/tts/sarvam.py (1)

85-85: Optional: Simplify the conditional check.

The condition regional_script and max_regional_count > 0 includes a redundant check. If max_regional_count > 0, then regional_script must be non-None (it was set in the loop). The simplified condition max_regional_count > 0 would be sufficient.

🔎 Proposed simplification

-        if regional_script and max_regional_count > 0:
+        if max_regional_count > 0:
             return regional_script

📜 Review details

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between ef71043 and a18aef7.

📒 Files selected for processing (2)

• app/ai/voice/tts/sarvam.py
• app/core/logger/__init__.py

🧰 Additional context used

🪛 Ruff (0.14.10)

app/ai/voice/tts/sarvam.py

93-93: Consider moving this statement to an else block

(TRY300)

🔇 Additional comments (3)

app/core/logger/__init__.py (1)

81-82: LGTM: Binary data spam filters added correctly.

The new filter conditions for "> BINARY" and "< BINARY" message prefixes are consistent with existing spam filtering patterns and will effectively reduce logging noise.

app/ai/voice/tts/sarvam.py (2)

54-59: Clear documentation of the new prioritization behavior.

The updated docstring accurately describes the regional-first detection logic and provides a helpful example.

---

75-93: No breaking changes identified.

The behavioral change is safe. The detect_script function has only one consumer (_switch_language_if_needed), which uses SCRIPT_TO_SARVAM_LANG.get(detected_script, "en-IN") with a fallback, so it's protected against any return value. The mapping covers all 10 possible script return values (9 regional scripts + "english"), and the implementation is correct and well-documented.

Note: There are currently no unit tests for detect_script, so there are no existing tests to break.

[manas-narra]

Not needed anymore, using cartesia TTS for regional languages now.