feat: add Amazon Bedrock as transcription provider

[devin-ai-integration]

feat: add Amazon Bedrock as transcription provider

Summary

Adds Amazon Bedrock as a new STT provider using the bedrock-mantle OpenAI-compatible endpoint (https://bedrock-mantle.us-east-1.api.aws/v1). Supports both real-time (WebSocket) and batch transcription.

Key decisions:

• Uses bedrock-mantle's OpenAI-compatible API rather than native AWS SDK/SigV4 — authentication is Bearer token (API key)
• Real-time adapter follows the OpenAI Realtime protocol (session.update, input_audio_buffer.append/commit)
• Default live model: openai.gpt-4o-transcribe, default batch model: amazon.nova-sonic-v1:0
• Live sample rate: 24000 Hz (matching OpenAI)
• Hardcoded to us-east-1 region; users can override via custom base URL

Files added: bedrock/mod.rs, bedrock/batch.rs, bedrock/live.rs

Registration across dispatch sites: listener-core, listener2-core, transcribe-proxy (batch + streaming + env + tests), desktop settings UI

Closes #3000

Review & Testing Checklist for Human

• Batch model mismatch: default_batch_model() in providers.rs returns "amazon.nova-sonic-v1:0" but the desktop UI lists "openai.whisper-1" as a model option. These need to be aligned to whatever bedrock-mantle actually expects for batch transcription.
• WebSocket protocol correctness: The live adapter assumes bedrock-mantle implements the OpenAI Realtime API (session.update with nested audio.input.transcription config, input_audio_buffer.append, conversation.item.input_audio_transcription.completed/delta, etc.). Verify this matches bedrock-mantle's actual WebSocket protocol — especially the include: ["item.input_audio_transcription.logprobs"] field in the session config, which may be OpenAI-specific.
• WebSocket path: ws_path() returns /v1/realtime for Bedrock. Confirm this is the correct path for bedrock-mantle's WebSocket endpoint.
• End-to-end test with a real Bedrock API key: This code was not tested against a live bedrock-mantle endpoint. Test both batch and real-time transcription paths to verify the protocol assumptions are correct.
• Icon availability: Desktop UI uses <Icon icon="simple-icons:amazonaws" /> — confirm this icon resolves correctly in the app's iconify setup.

Suggested test plan:

1. Configure a Bedrock API key in the desktop settings UI and verify the provider appears correctly
2. Run a batch transcription against a real audio file and verify the response parses correctly
3. Run a real-time transcription session and verify WebSocket connection, session update, audio streaming, and transcript responses all work

Notes

• Link to Devin run: https://app.devin.ai/sessions/8c5bed128211473a87e5a475802b4b39
• Requested by: @ComputelessComputer

[netlify]

✅ Deploy Preview for hyprnote-storybook canceled.

Name	Link
🔨 Latest commit	4db7e82
🔍 Latest deploy log	https://app.netlify.com/projects/hyprnote-storybook/deploys/699ed208fcb82b0008af1920

[netlify]

✅ Deploy Preview for hyprnote canceled.

Name	Link
🔨 Latest commit	4db7e82
🔍 Latest deploy log	https://app.netlify.com/projects/hyprnote/deploys/699ed208a7a48f0008a97572

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring