Summary

This PR introduces a comprehensive AI Assistant feature with an integrated chat window, enabling users to interact with AI models (Anthropic Claude & OpenAI GPT) directly within the QuestDB Web Console for SQL assistance, query explanations, and error fixing.

Features

AI Chat Window

• Conversational AI interface in the right sidebar for natural language interactions
• Chat history with persistent conversations stored in IndexedDB
• Message streaming with real-time status indicators
• Conversation compaction using LLM summarization to manage context window limits

Query Assistance

• Explain Query - Get AI-powered explanations of SQL queries with a single click
• Fix Query - Automatic error detection and AI-suggested fixes for failed queries
• Schema Explanation - Explain table schemas with column descriptions and storage details
• SQL Generation - Generate queries from natural language descriptions

Monaco Editor Integration

• Glyph margin icons - AI sparkle icons in the editor gutter for quick access to AI features
• Inline diff view - Review AI-suggested SQL changes before applying
• Accept/Reject workflow - Easy application of AI suggestions to the editor
• Query-specific conversations - Conversations are tied to specific queries in buffers

AI Provider Configuration

• Multi-provider support - Anthropic (Claude) and OpenAI (GPT) models
• Model selection - Choose from multiple models (Claude Sonnet 4.5, Opus 4.5, GPT-5.1, o3, o4-mini, etc.)
• Settings and Configuration modals - Configure providers, enable/disable models, set defaults

Status & Feedback

• AI Status Indicator - Shows current operation status (Processing, Compacting, Tool Use, etc.) when chat window is closed
• Token usage display - Track input/output tokens per conversation

New Providers

• AIConversationProvider - Manages conversation state, persistence, and operations
• AIStatusProvider - Handles AI operation status and abort controls

Dependencies Added

• @anthropic-ai/sdk - Anthropic Claude API client
• openai - OpenAI API client
• js-tiktoken - Token counting for context management
• fflate - Compression for conversation storage
• remark-gfm - GitHub-flavored markdown rendering
• vite-plugin-wasm + vite-plugin-top-level-await - WASM support for tiktoken

Data Persistence

• Conversation metas and messages are stored in IndexedDB. Messages are compressed
• Settings stored in localStorage

QuestDB Integration

• Tool use for SQL validation via QuestDB backend (/api/v1/sql/validate)
• Schema access for table/column information, using SHOW CREATE ... queries
• Documentation retrieval from https://questdb.com/docs for context

Testing

• Comprehensive E2E test suite (aiAssistant.spec.js)