Give Your AI Long-Term Memory

# Project Name

## 1. Context
Brief description of you and this project.

## 2. Tech Stack
- Language/Framework
- Database
- Deployment

## 3. Current State
What's built, what's in progress.

## 4. Learnings
### Patterns That Work
- (none yet)

### Avoid
- (none yet)

## 5. Session Log

**YYYY-MM-DD:** What happened

---
name: close
description: Session closer. Updates CLAUDE.md and commits.
---

# Session Close

**AUTO-TRIGGER** on: bye, thanks, done, goodnight, etc.

## Steps

1. **Capture Learnings** — Scan session for corrections,
   preferences, patterns. Add to Section 9 with date.

2. **Update Section 12** — Replace with current project state:
   - What's built/working
   - Open threads with checkboxes
   - Key decisions made
   - Recent session summary

3. **Archive** — Save to /sessions/YYYY-MM-DD-HHMMSS.md

4. **Commit & Push**
   ```
   git add CLAUDE.md sessions/
   git commit -m "session: [date] [summary]"
   git push
   ```

5. **Confirm** — Output summary of learnings and open threads

# Create virtual environment
python -m venv ~/.claude/venv
source ~/.claude/venv/bin/activate

# Install packages
pip install chromadb sentence-transformers

import chromadb
from sentence_transformers import SentenceTransformer
from pathlib import Path

# Initialize
model = SentenceTransformer('all-MiniLM-L6-v2')
client = chromadb.PersistentClient(path=str(Path.home() / '.claude/chroma_db'))
collection = client.get_or_create_collection('sessions')

# Index sessions
sessions_dir = Path.home() / '.claude/compacted-summaries'
for file in sessions_dir.glob('*.md'):
    content = file.read_text()
    embedding = model.encode(content).tolist()
    collection.upsert(
        ids=[file.name],
        documents=[content],
        embeddings=[embedding]
    )
    print(f"Indexed: {file.name}")

print(f"Total: {collection.count()} documents")

import sys
import chromadb
from sentence_transformers import SentenceTransformer
from pathlib import Path

query = ' '.join(sys.argv[1:])
model = SentenceTransformer('all-MiniLM-L6-v2')
client = chromadb.PersistentClient(path=str(Path.home() / '.claude/chroma_db'))
collection = client.get_collection('sessions')

# Search
results = collection.query(
    query_embeddings=[model.encode(query).tolist()],
    n_results=5
)

# Display
for i, (doc, dist) in enumerate(zip(results['documents'][0], results['distances'][0])):
    score = 1 - dist  # Convert distance to similarity
    print(f"\n[{score:.2f}] {results['ids'][0][i]}")
    print(doc[:200] + "...")

# Point vector search to your sessions
ln -s /path/to/project/sessions ~/.claude/compacted-summaries

# Create search alias
echo 'alias memory-search="python ~/.claude/search.py"' >> ~/.zshrc