feat: Add MCP server stubs, development docs, and Docker updates

- Add MCP server skeleton implementations for all 7 planned servers
  (llm-gateway, knowledge-base, git, issues, filesystem, code-analysis, cicd)
- Add comprehensive DEVELOPMENT.md with setup and usage instructions
- Add BACKLOG.md with detailed phase planning
- Update docker-compose.dev.yml with Redis and Celery workers
- Update CLAUDE.md with Syndarix-specific context

Addresses issues #16, #20, #21

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

mcp-servers/knowledge-base/server.py

@@ -0,0 +1,162 @@
+"""
+Syndarix Knowledge Base MCP Server.
+
+Provides RAG capabilities with:
+- pgvector for semantic search
+- Per-project collection isolation
+- Hybrid search (vector + keyword)
+- Chunking strategies for code, markdown, and text
+
+Per ADR-008: Knowledge Base RAG Architecture.
+"""
+
+import os
+
+from fastmcp import FastMCP
+
+# Create MCP server
+mcp = FastMCP(
+    "syndarix-knowledge-base",
+    description="RAG with pgvector for semantic search",
+)
+
+# Configuration
+DATABASE_URL = os.getenv("DATABASE_URL")
+REDIS_URL = os.getenv("REDIS_URL", "redis://localhost:6379/0")
+
+
+@mcp.tool()
+async def search_knowledge(
+    project_id: str,
+    query: str,
+    top_k: int = 10,
+    search_type: str = "hybrid",
+    filters: dict | None = None,
+) -> dict:
+    """
+    Search the project knowledge base.
+
+    Args:
+        project_id: UUID of the project (scopes to project collection)
+        query: Search query text
+        top_k: Number of results to return
+        search_type: Search type (semantic, keyword, hybrid)
+        filters: Optional filters (file_type, path_prefix, etc.)
+
+    Returns:
+        List of matching documents with scores
+    """
+    # TODO: Implement pgvector search
+    # 1. Generate query embedding via LLM Gateway
+    # 2. Search project-scoped collection
+    # 3. Apply filters
+    # 4. Return results with scores
+    return {
+        "status": "not_implemented",
+        "project_id": project_id,
+        "query": query,
+    }
+
+
+@mcp.tool()
+async def ingest_document(
+    project_id: str,
+    content: str,
+    source_path: str,
+    doc_type: str = "text",
+    metadata: dict | None = None,
+) -> dict:
+    """
+    Ingest a document into the knowledge base.
+
+    Args:
+        project_id: UUID of the project
+        content: Document content
+        source_path: Original file path for reference
+        doc_type: Document type (code, markdown, text)
+        metadata: Additional metadata
+
+    Returns:
+        Ingestion result with chunk count
+    """
+    # TODO: Implement document ingestion
+    # 1. Apply chunking strategy based on doc_type
+    # 2. Generate embeddings for chunks
+    # 3. Store in project collection
+    return {
+        "status": "not_implemented",
+        "project_id": project_id,
+        "source_path": source_path,
+    }
+
+
+@mcp.tool()
+async def ingest_repository(
+    project_id: str,
+    repo_path: str,
+    include_patterns: list[str] | None = None,
+    exclude_patterns: list[str] | None = None,
+) -> dict:
+    """
+    Ingest an entire repository into the knowledge base.
+
+    Args:
+        project_id: UUID of the project
+        repo_path: Path to the repository
+        include_patterns: Glob patterns to include (e.g., ["*.py", "*.md"])
+        exclude_patterns: Glob patterns to exclude (e.g., ["node_modules/*"])
+
+    Returns:
+        Ingestion summary with file and chunk counts
+    """
+    # TODO: Implement repository ingestion
+    return {
+        "status": "not_implemented",
+        "project_id": project_id,
+        "repo_path": repo_path,
+    }
+
+
+@mcp.tool()
+async def delete_document(
+    project_id: str,
+    source_path: str,
+) -> dict:
+    """
+    Delete a document from the knowledge base.
+
+    Args:
+        project_id: UUID of the project
+        source_path: Original file path
+
+    Returns:
+        Deletion result
+    """
+    # TODO: Implement document deletion
+    return {
+        "status": "not_implemented",
+        "project_id": project_id,
+        "source_path": source_path,
+    }
+
+
+@mcp.tool()
+async def get_collection_stats(project_id: str) -> dict:
+    """
+    Get statistics for a project's knowledge base collection.
+
+    Args:
+        project_id: UUID of the project
+
+    Returns:
+        Collection statistics (document count, chunk count, etc.)
+    """
+    # TODO: Implement collection stats
+    return {
+        "status": "not_implemented",
+        "project_id": project_id,
+    }
+
+
+if __name__ == "__main__":
+    mcp.run()