Add chunk-level state tracking with retry and crash recovery

[mohi-devhub]

Overview

This PR implements a robust chunk-level state tracking system that ensures reliable audio processing with automatic retry and crash recovery capabilities.

What's New

Core Components:

• ChunkStateManager - Manages chunk lifecycle with three states: RECORDED → SENT → TRANSCRIBED
• RecoveryManager - Handles bot restart recovery and resumes incomplete processing
• Meeting state integration - Track chunks per meeting session

Features:

• ✅ Per-chunk state tracking with JSON persistence
• ✅ Independent retry with exponential backoff (max 5 attempts)
• ✅ Crash recovery - state survives process/VM restarts
• ✅ Meeting completion guarantee - all chunks must be TRANSCRIBED
• ✅ No silent failures - comprehensive logging

Code Quality:

• Removed decorative emojis from bot messages
• Clear error/warning text prefixes
• Validated state transitions
• Proper error handling throughout

Documentation

• README.md - Updated with setup, testing, and debugging guides
• CONTRIBUTING.md - Developer guidelines with integration patterns
• docs/CHUNK_STATE_GUIDE.md - Quick start integration reference

Testing

Includes comprehensive test suite and integration examples:

• examples/test_chunk_state.py - Unit tests for all core functionality
• examples/integration_example.py - Real-world usage with mock API

All tests pass ✓

Files Changed

New:

• bot/chunk_state_manager.py (467 lines)
• bot/recovery_manager.py (294 lines)
• docs/CHUNK_STATE_GUIDE.md
• CONTRIBUTING.md

Modified:

• bot/main.py - Cleaned up messages
• bot/meeting_state.py - Added chunk tracking
• README.md - Comprehensive documentation

Integration Points

Ready for immediate integration with:

• Transcription API - Drop-in retry handling
• Summarization pipeline - Completion validation and transcript assembly

Fixes : #5