feat: US6 SDK Corpus + US3 Language Filtering + US4 LLM Code Summaries

.env.example

@@ -24,8 +24,8 @@ EMBEDDING_MODEL=text-embedding-3-large
 # Get your API key from: https://console.anthropic.com/settings/keys
 ANTHROPIC_API_KEY=sk-ant-your-anthropic-api-key-here
 
-# Claude model for summarization (default: claude-3-5-haiku-20241022)
-CLAUDE_MODEL=claude-3-5-haiku-20241022
+# Claude model for summarization (default: claude-haiku-4-5-20251001)
+CLAUDE_MODEL=claude-haiku-4-5-20251001
 
 # =============================================================================
 # Chroma Vector Store Configuration

.github/workflows/pr-qa-gate.yml

@@ -29,7 +29,7 @@ jobs:
       - name: Install Task
         uses: arduino/setup-task@v2
         with:
-          version: 3.x
+          version: 3.43.3
 
       - name: Install Poetry
         uses: snok/install-poetry@v1

AGENTS.md

@@ -162,7 +162,9 @@ task before-push
 
 This runs format, lint, typecheck, and tests with coverage.
 
-**MANDATORY**: You MUST run `task pr-qa-gate` before checking in or pushing any changes. You should also run `task pr-qa-gate` whenever checking project status or SDD status.
+**IMPORTANT**: You MUST run `task pr-qa-gate` before checking in or pushing any changes. You should also run `task pr-qa-gate` whenever checking project status or SDD status.
+
+**MANDATORY**: Any feature or task is not considered done unless `task pr-qa-gate` passes successfully.
 
 ## Git Workflow
 

CLAUDE.md

@@ -198,6 +198,10 @@ Do NOT push code that fails `task before-push`.
 ## Active Technologies
 - Python 3.10+ + FastAPI, LlamaIndex, ChromaDB, OpenAI, rank-bm25 (100-bm25-hybrid-retrieval)
 - ChromaDB (Vector Store), Local Persistent BM25 Index (LlamaIndex) (100-bm25-hybrid-retrieval)
+- Python 3.10+ + LlamaIndex (CodeSplitter, SummaryExtractor), tree-sitter parsers, ChromaDB (101-code-ingestion)
+- ChromaDB (unified vector store), Disk-based BM25 index (101-code-ingestion)
+- Python 3.10+ + LlamaIndex (CodeSplitter, SummaryExtractor), tree-sitter (AST parsing), OpenAI/Anthropic (embeddings/summaries) (101-code-ingestion)
+- ChromaDB vector store (existing) (101-code-ingestion)
 
 ## Recent Changes
 - 100-bm25-hybrid-retrieval: Added Python 3.10+ + FastAPI, LlamaIndex, ChromaDB, OpenAI, rank-bm25

README.md

@@ -12,13 +12,41 @@ Doc-Serve is a monorepo containing three packages:
 | **doc-svr-ctl** | Command-line interface for managing the server |
 | **doc-serve-skill** | Claude Code skill for AI-powered documentation queries |
 
+## Code Ingestion & Search
+
+Doc-Serve now supports unified search across documentation and source code:
+
+- **10 Programming Languages**: Python, TypeScript, JavaScript, Java, Kotlin, C, C++, Go, Rust, Swift
+- **AST-Aware Chunking**: Intelligent code parsing and chunking using tree-sitter
+- **Cross-Reference Queries**: Search across docs and code simultaneously
+- **Language Filtering**: Filter results by programming language
+- **Source Type Filtering**: Separate results by documentation vs. source code
+- **LLM Code Summaries**: AI-generated summaries improve semantic search quality
+
+### Example: Code-Aware Search
+```bash
+# Index both docs and code
+doc-svr-ctl index ./my-project --include-code
+
+# Search across everything
+doc-svr-ctl query "authentication implementation"
+
+# Filter by code only
+doc-svr-ctl query "API endpoints" --source-types code --languages python
+```
+
 ## Features
 
+- **Code Ingestion**: Index and search across documentation AND source code
+- **Cross-Reference Search**: Unified queries across docs and code with intelligent filtering
+- **Language-Aware Processing**: AST-based chunking for 10+ programming languages
 - **Hybrid Search**: Combines semantic meaning (Vector) with exact keyword matching (BM25)
 - **Semantic Search**: Natural language queries using OpenAI embeddings
 - **Keyword Search**: Precise term matching for technical documentation
+- **Advanced Filtering**: Filter by source type (doc/code) and programming language
 - **Vector Store**: ChromaDB for efficient similarity search
-- **Context-Aware Chunking**: Intelligent document splitting with overlap
+- **Context-Aware Chunking**: Intelligent document and code splitting with overlap
+- **LLM Summaries**: AI-generated summaries for code chunks improve semantic search
 - **REST API**: Full OpenAPI-documented REST interface
 - **CLI Tool**: Comprehensive command-line management
 - **Claude Integration**: Native Claude Code skill for AI workflows
@@ -155,6 +183,7 @@ doc-serve/
 - **Embeddings**: OpenAI text-embedding-3-large
 - **Summarization**: Claude Haiku
 - **Indexing**: LlamaIndex
+- **Code Parsing**: Tree-sitter (AST-aware chunking)
 - **CLI**: Click + Rich
 - **Build System**: Poetry