feat: add chunk splitting, S3 storage, and session transcript copy

[animalnots]

PR: Add Chunk Splitting, S3 Storage & Session Transcript Copy

Summary

This PR adds three features:

• size-based recording chunk splitting to avoid memory-only recording risk and provide incremental transcription,
• S3-compatible temporary storage + presigned URL flow for large-file transcription,
• full-session transcript aggregation with a "Copy Full Session" action in the UI.

Tested on Windows only.

---

What Changed

1. Feature: Size-Based Recording Chunk Splitting

Problem: Previously, recordings were kept entirely in memory until the user stopped recording. For long recordings (e.g., 2+ hours), this created:

• Data loss risk — Browser crash or memory limit = entire recording lost
• Large transcription delays — Users had to wait until the end to get any transcription results
• Inefficient for large files — Single large blob difficult to upload/process

Solution: New chunk splitting feature that automatically splits recordings at a configurable size threshold (default: 24.5 MB):

• Incremental disk saves — Each chunk is saved to disk immediately via new recordingStorage.js module
• Incremental transcription — Chunks transcribe as they complete (live results during recording)
• Size tracking — Added timeslice: 1000 to MediaRecorder.start() calls so ondataavailable fires every ~1s, enabling real-time cumulative size tracking
• Configurable threshold — User can adjust chunk size in Settings → Recording

New files: src/helpers/recordingStorage.js
Modified: src/helpers/audioManager.js (new _startChunkSplitTimer, _splitRecordingChunk, _saveRecordingBackup methods)

---

2. Feature: S3-Compatible Cloud Storage for Large Files

Why: For large audio files (>25 MB), direct upload to transcription providers can fail with HTTP 413 errors. S3-compatible storage provides temporary hosting with presigned URLs that services like Groq can fetch directly.

Changes:

• src/helpers/s3Storage.js — New S3StorageManager with support for any S3-compatible provider (AWS S3, Cloudflare R2, MinIO, Backblaze B2, etc.). Supports custom endpoint URL, region, forcePathStyle, presigned URLs, and a full connection test (write → read via presigned URL → delete).
• main.js — Instantiates S3StorageManager and passes to IPC handlers.
• src/helpers/ipcHandlers.js — S3 IPC handlers for config, test connection, upload, presigned URLs, and cleanup.
• preload.js — S3 IPC channel exposures.
• src/types/electron.ts — Type definitions for S3 config including endpointUrl, region, forcePathStyle, presigned URL on upload.
• src/helpers/audioManager.js — S3 upload/cleanup methods, presigned URL passthrough for Groq large-file transcription.
• src/components/SettingsPage.tsx — S3 settings UI with endpoint URL and region fields, plus a quick-start guide.
• src/locales/en/translation.json — S3 storage i18n strings.

Reliability improvement: The connection test now verifies write + public read (via presigned URL) + delete, not just bucket access. This catches misconfigured CORS/permissions before the user starts recording.

Presigned URL passthrough: For files exceeding 25 MB, the app now passes a presigned S3 URL to Groq's API (url parameter) instead of uploading the blob directly, avoiding HTTP 413 errors.

---

3. Feature: Copy Full Session Transcript

Why: When chunk splitting is enabled, a long recording (e.g. 2 hours) produces many partial transcriptions. Users had no way to get the combined text for the entire session.

How it works:

1. AudioManager tracks _sessionTranscripts[] — an ordered array of {partIndex, text} entries.
2. Each chunk split and the final recording part push their transcription text into this array.
3. When the final part's transcription completes and there are >1 parts, the combined text is:
   • Pasted into the active input field (e.g. Notepad++) as the full session text — not just the last part.
   • Broadcast via IPC (session-transcript-ready) to the ControlPanel window.
4. The ControlPanel shows a dismissible banner with session timing (start → end, duration, character count) and a "Copy Full Session" button.

Files changed:

• src/helpers/audioManager.js — Session tracking, combined text assembly, IPC broadcast with timing data.
• src/hooks/useAudioRecording.js — Uses result.sessionText for paste at session end instead of just the final part's text.
• src/helpers/ipcHandlers.js — New broadcast-session-transcript IPC handler.
• preload.js — broadcastSessionTranscript and onSessionTranscriptReady channels.
• src/types/electron.ts — Type definitions for session transcript IPC (includes timing fields).
• src/components/ControlPanel.tsx — Session transcript banner with Layers icon, timing info, Copy button, and dismiss (X) button.
• src/locales/en/translation.json — i18n strings for session banner (controlPanel.session.*).

---

Files Modified

File	Change Type
src/helpers/audioManager.js	Chunk splitting + session tracking + S3 upload/cleanup
src/helpers/s3Storage.js	New file
src/helpers/recordingStorage.js	New file
src/helpers/ipcHandlers.js	S3 IPC handlers, session transcript broadcast
preload.js	S3 and session transcript IPC channels
src/types/electron.ts	S3 config types, session transcript types
main.js	S3StorageManager instantiation
src/components/SettingsPage.tsx	S3 settings UI with endpoint/region fields
src/components/ControlPanel.tsx	Session transcript banner + copy button
src/hooks/useAudioRecording.js	Combined session paste at session end
src/locales/en/translation.json	S3 + session transcript i18n strings

Testing Notes

• Platform: Windows 11 only. macOS/Linux not tested.
• Chunk splitting: Set chunk size to 1 MB in Settings → Recording, record for a few minutes, verify files in recordings folder are ~1 MB each and transcriptions appear incrementally.
• Session transcript: After a multi-part recording, verify the banner appears in ControlPanel with correct part count, timing, and that "Copy Full Session" copies the combined text.
• S3 storage: Configure any S3-compatible provider, run "Test Connection" — should show write/read/delete steps passing.
• Presigned URL: Record a file >25 MB with Groq provider and S3 enabled — should use URL mode instead of direct upload.

Breaking Changes

None.