fix: text_to_speech requires output_path, removes inline audio

[maharshi-smallest]

Summary

Fixes TTS tool creating 3 duplicate files on Desktop when Claude Desktop can't render inline audio.

Changes

• output_path required — LLM asks user where to save (suggests ~/Desktop by default)
• Removed inline AudioContent — Claude Desktop doesn't support it, causing retries
• Expands ~ to home directory
• Description tells LLM not to retry on success
• Added duration estimate

🤖 Generated with Claude Code

[entelligence-ai-pr-reviews]

EntelligenceAI PR Summary

Refactors text_to_speech tool in src/tools/text-to-speech.ts to enforce caller-supplied output paths and remove inline audio data from responses.

• output_path promoted to required parameter at top of schema with tilde (~) expansion support
• Tool description updated to instruct LLM to always prompt the user for a save location before invoking the tool
• Removed inline base64 audio response block and associated mimeMap lookup
• Added durationEstimate field to the success response payload
• Removed default Desktop path fallback logic, now redundant since output_path is always provided

---

Confidence Score: 5/5 - Safe to Merge

Safe to merge — this PR cleanly refactors text_to_speech in src/tools/text-to-speech.ts by promoting output_path to a required parameter with tilde expansion support, removing the inline base64 audio response and associated mimeMap lookup, and adding a durationEstimate field to the success payload. The automated review found zero issues across the changed file, and the changes reflect a coherent, well-scoped improvement to the tool's interface contract. No logic regressions, security concerns, or unresolved prior comments were identified.

Key Findings:

• Promoting output_path to a required parameter in the schema is a correctness improvement — it eliminates a code path where the tool could silently omit saving audio, ensuring callers always supply an explicit destination.
• Removing the inline base64 audio block and mimeMap lookup simplifies the response contract and reduces unnecessary data transfer in the tool's output, with no apparent loss of functionality since the file is written to disk.
• The addition of durationEstimate to the success response is a backward-compatible enrichment with no risk of breaking existing consumers.
• Zero automated review comments were generated and coverage spans the single changed file, giving high confidence that the change is internally consistent.

Files requiring special attention

• src/tools/text-to-speech.ts