feat: Add Conversation History and Persistence

[r4kin]

Overview

Adds conversation memory to the GCP Monitoring Bot, enabling users to reference previous interactions naturally and maintain context across sessions.

Key Features

• SQLite-based conversation storage with session management
• Context-aware responses using conversation history
• History commands: history, search history , clear history, stats
• Natural references: "Show me that BigQuery data again" automatically finds previous conversations

Technical Implementation

• New ConversationStore class handles all database operations
• Enhanced GCPMonitoringBot with seamless history integration
• Intelligent context retrieval with fallback word matching
• Session isolation for multi-user support

Usage Examples

User :> Show BigQuery usage for last 7 days
Bot :> [Returns usage data: 2.3TB processed, $11.50 cost]

User :> What about those costs we discussed?
Bot :> Referring back to our BigQuery discussion, the $11.50...

User :> history
Bot :> [Shows formatted conversation history]

Testing

All tests passing: 7/7

• 3 unit tests verify core database functionality
• 4 integration tests ensure components work together
• Mock-based testing enables validation without API keys

Test Results:
Context retrieval: Passed (236+ char contexts found)
Session isolation: Passed (2 vs 1 verification)
Database operations: Passed (CRUD operations working)
Bot integration: Passed (Mocked API integration successful)

Current Limitations

• No live GenAI testing: Unable to verify actual context injection with Gemini API due to lack of API keys
• Context effectiveness unknown: Can't measure if conversation history actually improves bot responses
• Local storage only: No cloud synchronization or backup
• Basic search: Simple pattern matching rather than semantic search

Technical Debt

• Performance optimization: Database queries not yet optimized for large conversation histories
• Configuration: Hard-coded settings should be configurable

Breaking Change Risk

• Changes create_bot() return type from Gemini chat to GCPMonitoringBot wrapper
• Requires SQLite database creation (fails if permissions/disk issues)
• Adds conversation_history.db file to working directory

Mitigation

• All original bot functionality preserved through wrapper class
• Same interface maintained for send_message() method
• Consider adding graceful degradation in future iteration