claude-memory-agent 2.0.0

package/hooks/session_start.py

@@ -0,0 +1,227 @@
+#!/usr/bin/env python3
+"""Session start hook - auto-loads relevant context.
+
+This hook runs when a Claude Code session starts and:
+- Loads project info and preferences
+- Retrieves recent decisions and patterns
+- Gets unresolved items from previous sessions
+- Injects relevant context into the session
+- Loads daily logs (Moltbot-inspired)
+- Loads MEMORY.md core facts (Moltbot-inspired)
+
+Configure in Claude Code settings:
+{
+  "hooks": {
+    "SessionStart": ["python /path/to/session_start.py"]
+  }
+}
+
+Output is printed to stdout and injected into Claude's context.
+"""
+import os
+import sys
+import json
+import asyncio
+from datetime import datetime
+from pathlib import Path
+from typing import Dict, Any, List, Optional
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+import httpx
+
+MEMORY_AGENT_URL = os.getenv("MEMORY_AGENT_URL", "http://localhost:8102")
+API_KEY = os.getenv("MEMORY_API_KEY", "")
+
+
+async def call_memory_skill(skill_id: str, params: Dict[str, Any]) -> Optional[Dict[str, Any]]:
+    """Call a memory agent skill."""
+    headers = {"Content-Type": "application/json"}
+    if API_KEY:
+        headers["X-Memory-Key"] = API_KEY
+
+    payload = {
+        "jsonrpc": "2.0",
+        "method": "skills/call",
+        "params": {
+            "skill_id": skill_id,
+            "params": params
+        },
+        "id": f"session-start-{datetime.now().isoformat()}"
+    }
+
+    try:
+        async with httpx.AsyncClient(timeout=10.0) as client:
+            response = await client.post(
+                f"{MEMORY_AGENT_URL}/a2a",
+                json=payload,
+                headers=headers
+            )
+            if response.status_code == 200:
+                data = response.json()
+                return data.get("result", {}).get("result", {})
+    except Exception as e:
+        pass
+    return None
+
+
+async def load_session_context(project_path: str) -> str:
+    """Load all relevant context for a session start."""
+    context_parts = []
+
+    # ============================================================
+    # MOLTBOT-INSPIRED: Load MEMORY.md first (core facts)
+    # ============================================================
+    memory_md = await call_memory_skill("read_memory_md", {
+        "project_path": project_path
+    })
+
+    if memory_md and memory_md.get("exists"):
+        context_parts.append("## Core Facts (from MEMORY.md)")
+        # Include the summary or first part of content
+        content = memory_md.get("content", "")
+        # Truncate if too long
+        if len(content) > 2000:
+            content = content[:2000] + "\n...(truncated)"
+        context_parts.append(content)
+
+    # ============================================================
+    # MOLTBOT-INSPIRED: Load recent daily logs
+    # ============================================================
+    daily_logs = await call_memory_skill("daily_log_read", {
+        "project_path": project_path,
+        "days": 2,
+        "max_chars": 3000
+    })
+
+    if daily_logs and daily_logs.get("logs"):
+        context_parts.append("\n## Recent Activity (from Daily Logs)")
+        for log in daily_logs["logs"]:
+            log_date = log.get("date", "Unknown")
+            log_content = log.get("content", "")
+            # Show just the highlights, not full content
+            if len(log_content) > 1500:
+                log_content = log_content[:1500] + "\n...(truncated)"
+            context_parts.append(f"\n### {log_date}")
+            context_parts.append(log_content)
+
+    # ============================================================
+    # ORIGINAL MEMORY SYSTEM CONTEXT
+    # ============================================================
+
+    # 1. Get project info
+    project_info = await call_memory_skill("get_project_context", {
+        "project_path": project_path,
+        "limit": 5
+    })
+
+    if project_info and project_info.get("project"):
+        proj = project_info["project"]
+        context_parts.append(f"\n## Project: {proj.get('name', project_path)}")
+        if proj.get("tech_stack"):
+            context_parts.append(f"Tech Stack: {', '.join(proj['tech_stack'])}")
+        if proj.get("conventions"):
+            context_parts.append(f"Conventions: {json.dumps(proj['conventions'], indent=2)}")
+
+    # 2. Get recent decisions
+    decisions = await call_memory_skill("semantic_search", {
+        "query": "decision architecture approach",
+        "project_path": project_path,
+        "type": "decision",
+        "limit": 5
+    })
+
+    if decisions and decisions.get("results"):
+        context_parts.append("\n## Recent Decisions")
+        for d in decisions["results"][:3]:
+            context_parts.append(f"- {d['content'][:150]}")
+
+    # 3. Get recent errors (to avoid repeating)
+    errors = await call_memory_skill("semantic_search", {
+        "query": "error bug fix problem",
+        "project_path": project_path,
+        "type": "error",
+        "success_only": True,  # Only get solved errors
+        "limit": 5
+    })
+
+    if errors and errors.get("results"):
+        context_parts.append("\n## Past Errors & Solutions")
+        for e in errors["results"][:3]:
+            context_parts.append(f"- {e['content'][:150]}")
+
+    # 4. Get session handoff (unresolved items)
+    handoff = await call_memory_skill("get_session_handoff", {
+        "project_path": project_path,
+        "include_last_n_sessions": 2
+    })
+
+    if handoff:
+        if handoff.get("unresolved_questions"):
+            context_parts.append("\n## Unresolved from Previous Sessions")
+            for q in handoff["unresolved_questions"][:3]:
+                context_parts.append(f"- {q}")
+
+        if handoff.get("recent_summaries"):
+            context_parts.append("\n## Recent Session Summaries")
+            for s in handoff["recent_summaries"][:2]:
+                context_parts.append(f"- {s.get('summary', '')[:200]}")
+
+    # 5. Get relevant patterns
+    patterns = await call_memory_skill("search_patterns", {
+        "query": "common patterns solutions",
+        "limit": 3
+    })
+
+    if patterns and patterns.get("patterns"):
+        context_parts.append("\n## Useful Patterns")
+        for p in patterns["patterns"][:2]:
+            context_parts.append(f"- **{p['name']}**: {p['solution'][:100]}")
+
+    # 6. Check for anchor conflicts
+    conflicts = await call_memory_skill("get_unresolved_conflicts", {
+        "project_path": project_path,
+        "limit": 3
+    })
+
+    if conflicts and conflicts.get("conflicts"):
+        context_parts.append("\n## Unresolved Fact Conflicts")
+        for c in conflicts["conflicts"]:
+            context_parts.append(f"- {c.get('anchor1_summary', '')} vs {c.get('anchor2_summary', '')}")
+
+    if context_parts:
+        return "\n".join(context_parts)
+    return ""
+
+
+async def main():
+    # Get project path from environment or current directory
+    project_path = os.getenv("PROJECT_PATH") or os.getcwd()
+
+    # Try to read from stdin if available
+    try:
+        if not sys.stdin.isatty():
+            data = sys.stdin.read()
+            if data:
+                hook_data = json.loads(data)
+                project_path = hook_data.get("project_path", project_path)
+    except:
+        pass
+
+    context = await load_session_context(project_path)
+
+    if context:
+        # Output context for Claude to see
+        print("\n<memory-context>")
+        print("# Loaded from Memory System")
+        print(context)
+        print("</memory-context>\n")
+    else:
+        print("\n<memory-context>")
+        print("# No prior context found for this project")
+        print("Starting fresh session.")
+        print("</memory-context>\n")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())