Integrate LLM Enhanced Commands Implementation for Query Generation

[xingfan-git]

No description provided.

[Copilot]

Pull Request Overview

This PR integrates LLM-enhanced query generation functionality into the DocumentDB extension, enabling AI-powered MongoDB query creation from natural language prompts. The implementation leverages GitHub Copilot's language models to generate queries based on collection schemas and user requests.

Key Changes:

• Replaced mock AI query generation with real LLM integration using GitHub Copilot
• Added comprehensive infrastructure for schema inference and prompt template management
• Implemented MongoDB client APIs for gathering collection metadata to support LLM context

Reviewed Changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/webviews/documentdb/collectionView/collectionViewRouter.ts	Replaces mock AI logic with actual LLM query generation using telemetry and error handling
src/utils/schemaInference.ts	Implements schema inference from MongoDB documents including type detection and vector array support
src/services/promptTemplateService.ts	Provides template loading service with caching and custom template file support
src/services/copilotService.ts	Wraps VS Code's language model API for GitHub Copilot interaction
src/documentdb/LlmEnhancedFeatureApis.ts	Implements MongoDB APIs for explain plans, index operations, and sample document retrieval
src/documentdb/ClustersClient.ts	Exposes LLM-enhanced feature APIs through the cluster client and adds metadata caching
src/commands/llmEnhancedCommands/queryGenerationCommands.ts	Orchestrates query generation from natural language using schema inference and LLM calls
src/commands/llmEnhancedCommands/promptTemplates.ts	Contains prompt templates for different query types and model configuration
package.json	Adds configuration settings for custom prompt templates and caching
l10n/bundle.l10n.json	Adds localized strings for new error messages and user-facing text

[tnaum-ms]

• lets pretty-print our queries when there are > 1 fields in use.
• llm config: I wonder why for this prompt "find the top 10 stores, look at store features" it added a long list of project fields instead of just leaving it empty.
• we need to add cancel support
• we need to add more trace messages to the output channel

---

• let's address the metadata topic before merging it, otherwise we'll forget.

the other items can be scheduled for 0.6.1 and 0.6.2

[tnaum-ms]

let's find a better name for this category "LLM", @khelanmodi this is about the settings the user can configure. How about "DocumentDB Copilot", "DocumentDB AI Assistant", "AI Assistant"?

[tnaum-ms]

@xingfan-git FYI: I changed it to AiAssistant. Your test env can break.

[Copilot]

@tnaum-ms I've opened a new pull request, #319, to work on those changes. Once the pull request is ready, I'll request review from you.

[tnaum-ms]

Requested changes were implemented.
Follow-up tickets/issues were created.