#!/usr/bin/env python3
"""Memora SessionStart hook - inject relevant memories into Claude Code context.
This script:
1. Reads session info from stdin (cwd, session_id)
2. Extracts project context from working directory
3. Searches memora for relevant memories
4. Returns additionalContext for Claude's system prompt
"""
import json
import os
import sys
from pathlib import Path
def load_memora_env():
"""Load memora environment variables from .mcp.json if available."""
# Try to find .mcp.json in common locations
search_paths = [
Path.home() / "repos" / "agentic-mcp-tools" / ".mcp.json",
Path.home() / ".mcp.json",
Path.cwd() / ".mcp.json",
]
for mcp_path in search_paths:
if mcp_path.exists():
try:
with open(mcp_path) as f:
config = json.load(f)
memora_config = config.get("mcpServers", {}).get("memora", {})
env_vars = memora_config.get("env", {})
for key, value in env_vars.items():
if key not in os.environ:
os.environ[key] = value
return True
except Exception:
pass
return False
def extract_project_context(cwd: str) -> dict:
"""Extract project identifiers from working directory."""
path = Path(cwd)
# Get project name (directory name)
project_name = path.name
# Build search queries from path components
queries = [project_name]
# Add parent directory names that might be relevant
for parent in list(path.parents)[:2]:
if parent.name and parent.name not in ("", "Users", "home", "repos", "src"):
queries.append(parent.name)
return {
"project_name": project_name,
"cwd": cwd,
"search_query": " ".join(queries[:3]),
}
def search_memora(query: str, top_k: int = 5) -> list:
"""Search memora for relevant memories using direct storage import."""
try:
# Try importing memora (installed as uv tool or in path)
from memora import storage
conn = storage.connect()
results = storage.hybrid_search(
conn,
query=query,
top_k=top_k,
min_score=0.02,
)
conn.close()
return results
except ImportError:
# Try adding local repo to path
memora_path = Path.home() / "repos" / "agentic-mcp-tools" / "memora"
if memora_path.exists():
sys.path.insert(0, str(memora_path))
try:
from memora import storage
conn = storage.connect()
results = storage.hybrid_search(conn, query=query, top_k=top_k, min_score=0.02)
conn.close()
return results
except Exception:
return []
return []
except Exception:
return []
def format_memories(memories: list, max_chars: int = 1500) -> str:
"""Format memories concisely for context injection."""
if not memories:
return ""
lines = ["## Relevant Memories\n"]
total_chars = len(lines[0])
for item in memories:
memory = item.get("memory", item)
score = item.get("score", 0)
# Extract key info
mid = memory.get("id", "?")
content = memory.get("content", "")
tags = memory.get("tags", [])
# Truncate long content
if len(content) > 150:
content = content[:150] + "..."
# Format as compact entry
tags_str = ", ".join(tags[:3]) if tags else ""
entry = f"- [#{mid}] {content}"
if tags_str:
entry += f" ({tags_str})"
entry += "\n"
# Check length limit
if total_chars + len(entry) > max_chars:
lines.append("- ... more available via `memory_hybrid_search`\n")
break
lines.append(entry)
total_chars += len(entry)
lines.append("\nUse memora tools to search for more context.\n")
return "".join(lines)
def main():
"""Main entry point for SessionStart hook."""
try:
# Load memora env from .mcp.json
load_memora_env()
# Read input from stdin
input_data = json.load(sys.stdin)
cwd = input_data.get("cwd", os.getcwd())
# Extract project context
context = extract_project_context(cwd)
# Search memora
memories = search_memora(context["search_query"], top_k=5)
# Format output
if memories:
additional_context = format_memories(memories)
output = {
"hookSpecificOutput": {
"hookEventName": "SessionStart",
"additionalContext": additional_context
}
}
else:
output = {}
print(json.dumps(output))
except Exception:
# On error, allow session to continue without blocking
print(json.dumps({}))
sys.exit(0)
if __name__ == "__main__":
main()
