feat: add CAMB AI text-to-speech for note summaries

[neilruaro-camb]

Summary

• Adds CAMB AI text-to-speech so users can listen to their note summaries as audio
• After a summary is generated, CAMB AI's streaming TTS creates a spoken audio version
• A "Listen to summary" audio player appears in the Summary view (desktop and mobile)
• Uses CAMB AI's MARS speech models for low-latency, high-quality speech synthesis

About CAMB AI

CAMB AI is a voice AI and localization platform trusted by brands like the Premier League, the NBA, NASCAR, and the Australian Open. We'd love for CAMB AI to power the audio experience in notesGPT.

Note on testing

During testing, we found that the @instructor-ai/instructor library's JSON_SCHEMA mode no longer works with Together.ai's current serverless models. We had to switch to using the Together SDK directly with meta-llama/Llama-3.3-70B-Instruct-Turbo to get summaries working, which then allowed the CAMB TTS feature to trigger. This is not included in this PR but worth noting.

Usage

1. Set CAMB_API_KEY in Convex environment variables (get a key from https://studio.camb.ai)
2. Record a voice note as usual
3. After the summary generates, toggle to "Summary" view
4. Click play on the "Listen to summary" audio player

Files changed

• convex/camb.ts - New file: CAMB TTS generation action and audio storage
• convex/schema.ts - Added summaryAudioFileId and summaryAudioUrl fields to notes table
• convex/together.ts - Schedules CAMB TTS generation after summary is saved
• components/pages/recording/RecordingDesktop.tsx - Audio player in summary view
• components/pages/recording/RecordingMobile.tsx - Audio player in summary view (mobile)

[vercel]

@neilruaro-camb is attempting to deploy a commit to the Together AI Team on Vercel.

A member of the Team first needs to authorize it.