docs: Enhance module docstrings for core engine components

Improve module-level documentation for MemDocs core functionality:

- MCP Server: Clarify flagship feature for Claude Desktop integration
- Extract: Document multi-language symbol extraction capabilities
- Embeddings: Highlight zero-cost ($0) local embedding generation
- Search: Detail FAISS-based offline vector similarity search
- Summarize: Describe Claude-powered structured documentation
- Policy: Explain intelligent scope escalation logic

Also update .gitignore to commit .memdocs/ directory, demonstrating
dogfooding of MemDocs on itself with git-committed memories.

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

Co-Authored-By: Claude <[email protected]>

Author: Patrick Roebuck

.gitignore

@@ -58,8 +58,9 @@ dmypy.json
 .env.local
 .env.*.local
 
-# MemDocs output (for testing)
-.memdocs/
+# MemDocs output
+# NOTE: Committing .memdocs/ to demonstrate dogfooding
+# .memdocs/  # Uncomment to exclude from git (for testing only)
 *.memdocs.yml.backup
 
 # Distribution

memdocs/embeddings.py

@@ -1,8 +1,10 @@
 """
 Local embeddings generation using sentence-transformers.
 
-Zero-cost alternative to OpenAI embeddings API.
-Model downloads once (~90MB), runs 100% locally.
+Zero-cost alternative to OpenAI embeddings API ($0.13/1M tokens → $0).
+Model downloads once (~90MB), runs 100% locally for privacy-first operation.
+
+Enables semantic search over code documentation without external API dependencies.
 """
 
 import json

memdocs/extract.py

@@ -1,10 +1,13 @@
 """
 Extract context from git diffs, code files, and repository metadata.
 
+Core extraction engine that powers MemDocs documentation generation.
+
 Responsibilities:
 - Parse git diffs to identify changed files
 - Extract code symbols (functions, classes) using AST parsing
 - Gather metadata (commit info, file stats, dependencies)
+- Support multi-language symbol extraction (Python, TypeScript, JavaScript)
 """
 
 import json

memdocs/mcp_server.py

@@ -2,6 +2,10 @@
 MCP (Model Context Protocol) server for Claude Desktop integration.
 
 Enables Claude Desktop to query project memory stored in .memdocs/
+via semantic search, symbol lookup, and documentation retrieval.
+
+This is MemDocs' flagship feature - allowing AI assistants to autonomously
+query git-committed memory without file system access.
 """
 
 import json

memdocs/policy.py

@@ -1,11 +1,14 @@
 """
-Policy engine for scope determination and escalation rules.
+Policy engine for intelligent scope determination and escalation.
 
-Determines when to expand scope from file → module → repo based on:
+Automatically expands documentation scope from file → module → repo based on:
 - Security-sensitive paths (auth/*, security/*)
-- Cross-module dependencies
-- Public API changes
-- File count thresholds
+- Cross-module dependencies and import chains
+- Public API signature changes
+- File count thresholds (configurable)
+
+Ensures critical changes get comprehensive documentation while keeping
+costs low for routine single-file updates.
 """
 
 from pathlib import Path

memdocs/search.py

@@ -1,8 +1,11 @@
 """
-Local vector search using FAISS.
+Local vector search using FAISS (Facebook AI Similarity Search).
 
-Zero-cost alternative to Pinecone ($70-200/month).
-Index stored in .memdocs/memory/ (git-committed).
+Zero-cost alternative to Pinecone ($70-200/month → $0).
+Index stored in .memdocs/memory/ (git-committed, version-controlled).
+
+Provides fast L2 distance similarity search for semantic code queries.
+Works offline with no external dependencies or API latency.
 """
 
 import json

memdocs/summarize.py

@@ -1,7 +1,9 @@
 """
-AI-powered summarization using Claude.
+AI-powered summarization using Claude Sonnet 4.5.
 
-Generates structured documentation from extracted code context.
+Generates structured YAML documentation from extracted code context,
+transforming git diffs and AST symbols into human-readable summaries
+with risk analysis, impact tracking, and feature descriptions.
 """
 
 import os