claude-memory-agent 2.0.1 → 2.2.0

package/hooks/grounding-hook.py

@@ -1,348 +1,422 @@
-#!/usr/bin/env python3
-"""
-Grounding Hook for Claude Code - Automatic Context Injection
-
-This script is called by Claude Code's UserPromptSubmit hook.
-It fetches the current session context and outputs it to stdout,
-which Claude Code automatically injects into Claude's context.
-
-This is the REAL anti-hallucination layer - automatic, not relying on Claude to call tools.
-
-Moltbot-inspired additions:
-- Checks flush conditions (events > 50 or time > 30min)
-- Loads MEMORY.md content into grounding context
-- Loads today's daily log highlights
-"""
-
-import os
-import sys
-import json
-import logging
-import requests
-from pathlib import Path
-from typing import Any, Optional
-
-# Configure logging to stderr (important for Claude Code hooks)
-logging.basicConfig(
-    level=logging.INFO,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
-    stream=sys.stderr
-)
-logger = logging.getLogger(__name__)
-
-# Configuration from environment
-MEMORY_AGENT_URL = os.getenv("MEMORY_AGENT_URL", "http://localhost:8102")
-API_TIMEOUT = int(os.getenv("API_TIMEOUT", "30"))
-
-
-def safe_get(data: Any, *keys, default: Any = None) -> Any:
-    """
-    Safely navigate nested data structures (dicts and lists).
-
-    Args:
-        data: The data structure to navigate
-        *keys: Keys (str for dict) or indices (int for list) to traverse
-        default: Value to return if path doesn't exist
-
-    Returns:
-        The value at the path, or default if not found
-
-    Example:
-        safe_get(result, "result", "artifacts", 0, "parts", 0, "text")
-    """
-    for key in keys:
-        if data is None:
-            return default
-        if isinstance(data, dict):
-            data = data.get(key, default)
-        elif isinstance(data, list) and isinstance(key, int):
-            if 0 <= key < len(data):
-                data = data[key]
-            else:
-                return default
-        else:
-            return default
-    return data
-
-
-def get_project_path():
-    """Get current working directory as project path."""
-    return os.getcwd()
-
-
-def load_session_data():
-    """Load session data from JSON file."""
-    session_file = Path(get_project_path()) / ".claude_session"
-    if session_file.exists():
-        try:
-            content = session_file.read_text().strip()
-            # Try JSON format first
-            return json.loads(content)
-        except json.JSONDecodeError as e:
-            logger.debug(f"JSON decode error, trying legacy format: {e}")
-            # Fall back to legacy plain text format (just session_id)
-            try:
-                content = session_file.read_text().strip()
-                return {"session_id": content}
-            except (IOError, OSError) as read_err:
-                logger.warning(f"Failed to read session file: {read_err}")
-                return None
-        except (IOError, OSError) as e:
-            logger.warning(f"Failed to read session file: {e}")
-            return None
-    return None
-
-
-def save_session_data(data: dict):
-    """Save session data to JSON file."""
-    session_file = Path(get_project_path()) / ".claude_session"
-    try:
-        session_file.write_text(json.dumps(data, indent=2))
-    except (IOError, OSError) as e:
-        logger.warning(f"Failed to save session data: {e}")
-
-
-def get_session_id():
-    """Get or create session ID from environment or file."""
-    # Try environment variable first
-    session_id = os.getenv("CLAUDE_SESSION_ID")
-    if session_id:
-        return session_id
-
-    # Try session file in project
-    data = load_session_data()
-    return data.get("session_id") if data else None
-
-
-def call_memory_agent(skill_id: str, params: dict) -> Optional[dict]:
-    """Call the memory agent API."""
-    try:
-        response = requests.post(
-            f"{MEMORY_AGENT_URL}/a2a",
-            json={
-                "jsonrpc": "2.0",
-                "id": "grounding-hook",
-                "method": "tasks/send",
-                "params": {
-                    "message": {"parts": [{"type": "text", "text": ""}]},
-                    "metadata": {
-                        "skill_id": skill_id,
-                        "params": params
-                    }
-                }
-            },
-            timeout=API_TIMEOUT
-        )
-        result = response.json()
-
-        # Safely extract the artifact text using safe_get
-        artifact_text = safe_get(result, "result", "artifacts", 0, "parts", 0, "text")
-        if artifact_text:
-            try:
-                return json.loads(artifact_text)
-            except json.JSONDecodeError as e:
-                logger.debug(f"Failed to parse artifact text as JSON for skill '{skill_id}': {e}")
-                return None
-        return None
-
-    except requests.RequestException as e:
-        # Silently fail - don't break Claude Code if memory agent is down
-        logger.debug(f"Memory agent request failed for skill '{skill_id}': {e}")
-        return None
-    except json.JSONDecodeError as e:
-        logger.debug(f"Failed to decode memory agent response for skill '{skill_id}': {e}")
-        return None
-
-
-def format_grounding_context(context: dict) -> str:
-    """Format the grounding context for injection."""
-    if not context or not context.get("success"):
-        return ""
-
-    grounding = context.get("grounding", {})
-
-    lines = ["[GROUNDING CONTEXT - VERIFY BEFORE RESPONDING]"]
-
-    # Current goal
-    if grounding.get("current_goal"):
-        lines.append(f"CURRENT GOAL: {grounding['current_goal']}")
-
-    # Entity registry
-    registry = grounding.get("entity_registry", {})
-    if registry:
-        lines.append("ENTITY REGISTRY (use these exact references):")
-        for key, value in list(registry.items())[:5]:
-            lines.append(f"  - {key}: {value}")
-
-    # Anchors (verified facts)
-    anchors = grounding.get("anchors", [])
-    if anchors:
-        lines.append("ANCHORS (verified facts - DO NOT CONTRADICT):")
-        for anchor in anchors[:5]:
-            lines.append(f"  - {anchor}")
-
-    # Recent decisions
-    decisions = grounding.get("decisions", [])
-    if decisions:
-        lines.append("RECENT DECISIONS:")
-        for decision in decisions[:3]:
-            lines.append(f"  - {decision}")
-
-    # Recent events
-    events = grounding.get("recent_events", [])
-    if events:
-        lines.append("RECENT EVENTS:")
-        for event in events[:5]:
-            lines.append(f"  - [{event.get('type', '?')}] {event.get('summary', '')}")
-
-    # Contradictions warning
-    contradictions = grounding.get("contradictions", [])
-    if contradictions:
-        lines.append("WARNING - POTENTIAL CONTRADICTIONS DETECTED:")
-        for c in contradictions[:3]:
-            lines.append(f"  - {c.get('content', '')[:100]}")
-
-    # Pending questions
-    pending = grounding.get("pending_questions", [])
-    if pending:
-        lines.append("PENDING QUESTIONS:")
-        for q in pending[:3]:
-            lines.append(f"  - {q}")
-
-    lines.append("[/GROUNDING CONTEXT]")
-    lines.append("")  # Empty line after
-
-    return "\n".join(lines)
-
-
-def format_memory_md_context(memory_md: dict) -> str:
-    """Format MEMORY.md content for injection."""
-    if not memory_md or not memory_md.get("exists"):
-        return ""
-
-    summary = memory_md.get("summary", "")
-    if not summary:
-        return ""
-
-    lines = ["[CORE FACTS from MEMORY.md]"]
-    lines.append(summary)
-    lines.append("[/CORE FACTS]")
-    lines.append("")
-
-    return "\n".join(lines)
-
-
-def format_daily_highlights(highlights: dict) -> str:
-    """Format daily log highlights for injection."""
-    if not highlights or not highlights.get("highlights"):
-        return ""
-
-    entries = highlights.get("highlights", [])
-    if not entries:
-        return ""
-
-    lines = ["[TODAY'S HIGHLIGHTS from Daily Log]"]
-    for entry in entries[:5]:
-        lines.append(f"  - {entry}")
-    lines.append("[/TODAY'S HIGHLIGHTS]")
-    lines.append("")
-
-    return "\n".join(lines)
-
-
-def check_and_trigger_flush(session_id: str, project_path: str):
-    """Check if flush is needed and trigger it."""
-    # Check flush conditions
-    flush_check = call_memory_agent("check_flush_needed", {
-        "session_id": session_id
-    })
-
-    if flush_check and flush_check.get("flush_needed"):
-        reasons = flush_check.get("reasons", [])
-        logger.info(f"Flush needed: {', '.join(reasons)}")
-
-        # Trigger flush
-        flush_result = call_memory_agent("pre_compaction_flush", {
-            "project_path": project_path,
-            "session_id": session_id
-        })
-
-        if flush_result and flush_result.get("success"):
-            logger.info(f"Flush completed: {flush_result.get('file_path')}")
-
-
-def main():
-    """Main entry point for the hook."""
-    project_path = get_project_path()
-    session_id = get_session_id()
-
-    # If no session, try to init one
-    if not session_id:
-        init_result = call_memory_agent("state_init_session", {
-            "project_path": project_path
-        })
-        if init_result and init_result.get("session_id"):
-            session_id = init_result["session_id"]
-            # Save session data as JSON
-            save_session_data({"session_id": session_id})
-
-    if not session_id:
-        # No session, no grounding - exit silently
-        sys.exit(0)
-
-    # ============================================================
-    # MOLTBOT-INSPIRED: Check flush conditions
-    # ============================================================
-    check_and_trigger_flush(session_id, project_path)
-
-    # ============================================================
-    # MOLTBOT-INSPIRED: Load MEMORY.md summary
-    # ============================================================
-    memory_md = call_memory_agent("get_memory_md_summary", {
-        "project_path": project_path
-    })
-
-    memory_md_context = format_memory_md_context(memory_md) if memory_md else ""
-
-    # ============================================================
-    # MOLTBOT-INSPIRED: Load today's daily log highlights
-    # ============================================================
-    daily_highlights = call_memory_agent("daily_log_highlights", {
-        "project_path": project_path
-    })
-
-    daily_context = format_daily_highlights(daily_highlights) if daily_highlights else ""
-
-    # ============================================================
-    # ORIGINAL: Get grounding context
-    # ============================================================
-    context = call_memory_agent("context_refresh", {
-        "session_id": session_id,
-        "include_recent_events": 5,
-        "include_state": True,
-        "include_checkpoint": True,
-        "check_contradictions": True
-    })
-
-    grounding_context = format_grounding_context(context) if context else ""
-
-    # Combine all context
-    output_parts = []
-
-    if memory_md_context:
-        output_parts.append(memory_md_context)
-
-    if daily_context:
-        output_parts.append(daily_context)
-
-    if grounding_context:
-        output_parts.append(grounding_context)
-
-    if output_parts:
-        print("\n".join(output_parts))
-
-    sys.exit(0)
-
-
-if __name__ == "__main__":
-    main()
+#!/usr/bin/env python3
+"""
+Grounding Hook for Claude Code - Automatic Context Injection
+
+This script is called by Claude Code's UserPromptSubmit hook.
+It fetches the current session context and outputs it to stdout,
+which Claude Code automatically injects into Claude's context.
+
+This is the REAL anti-hallucination layer - automatic, not relying on Claude to call tools.
+
+Moltbot-inspired additions:
+- Checks flush conditions (events > 50 or time > 30min)
+- Loads MEMORY.md content into grounding context
+- Loads today's daily log highlights
+"""
+
+import os
+import sys
+import json
+import logging
+import requests
+from pathlib import Path
+from typing import Any, Optional
+
+# Configure logging to stderr (important for Claude Code hooks)
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
+    stream=sys.stderr
+)
+logger = logging.getLogger(__name__)
+
+# Configuration from environment
+MEMORY_AGENT_URL = os.getenv("MEMORY_AGENT_URL", "http://localhost:8102")
+API_TIMEOUT = int(os.getenv("API_TIMEOUT", "30"))
+
+
+def safe_get(data: Any, *keys, default: Any = None) -> Any:
+    """
+    Safely navigate nested data structures (dicts and lists).
+
+    Args:
+        data: The data structure to navigate
+        *keys: Keys (str for dict) or indices (int for list) to traverse
+        default: Value to return if path doesn't exist
+
+    Returns:
+        The value at the path, or default if not found
+
+    Example:
+        safe_get(result, "result", "artifacts", 0, "parts", 0, "text")
+    """
+    for key in keys:
+        if data is None:
+            return default
+        if isinstance(data, dict):
+            data = data.get(key, default)
+        elif isinstance(data, list) and isinstance(key, int):
+            if 0 <= key < len(data):
+                data = data[key]
+            else:
+                return default
+        else:
+            return default
+    return data
+
+
+def get_project_path():
+    """Get current working directory as project path."""
+    return os.getcwd()
+
+
+def load_session_data():
+    """Load session data from JSON file."""
+    session_file = Path(get_project_path()) / ".claude_session"
+    if session_file.exists():
+        try:
+            content = session_file.read_text().strip()
+            # Try JSON format first
+            return json.loads(content)
+        except json.JSONDecodeError as e:
+            logger.debug(f"JSON decode error, trying legacy format: {e}")
+            # Fall back to legacy plain text format (just session_id)
+            try:
+                content = session_file.read_text().strip()
+                return {"session_id": content}
+            except (IOError, OSError) as read_err:
+                logger.warning(f"Failed to read session file: {read_err}")
+                return None
+        except (IOError, OSError) as e:
+            logger.warning(f"Failed to read session file: {e}")
+            return None
+    return None
+
+
+def save_session_data(data: dict):
+    """Save session data to JSON file."""
+    session_file = Path(get_project_path()) / ".claude_session"
+    try:
+        session_file.write_text(json.dumps(data, indent=2))
+    except (IOError, OSError) as e:
+        logger.warning(f"Failed to save session data: {e}")
+
+
+def get_session_id():
+    """Get or create session ID from environment or file."""
+    # Try environment variable first
+    session_id = os.getenv("CLAUDE_SESSION_ID")
+    if session_id:
+        return session_id
+
+    # Try session file in project
+    data = load_session_data()
+    return data.get("session_id") if data else None
+
+
+def call_memory_agent(skill_id: str, params: dict) -> Optional[dict]:
+    """Call the memory agent API."""
+    try:
+        response = requests.post(
+            f"{MEMORY_AGENT_URL}/a2a",
+            json={
+                "jsonrpc": "2.0",
+                "id": "grounding-hook",
+                "method": "tasks/send",
+                "params": {
+                    "message": {"parts": [{"type": "text", "text": ""}]},
+                    "metadata": {
+                        "skill_id": skill_id,
+                        "params": params
+                    }
+                }
+            },
+            timeout=API_TIMEOUT
+        )
+        result = response.json()
+
+        # Safely extract the artifact text using safe_get
+        artifact_text = safe_get(result, "result", "artifacts", 0, "parts", 0, "text")
+        if artifact_text:
+            try:
+                return json.loads(artifact_text)
+            except json.JSONDecodeError as e:
+                logger.debug(f"Failed to parse artifact text as JSON for skill '{skill_id}': {e}")
+                return None
+        return None
+
+    except requests.RequestException as e:
+        # Silently fail - don't break Claude Code if memory agent is down
+        logger.debug(f"Memory agent request failed for skill '{skill_id}': {e}")
+        return None
+    except json.JSONDecodeError as e:
+        logger.debug(f"Failed to decode memory agent response for skill '{skill_id}': {e}")
+        return None
+
+
+def format_grounding_context(context: dict) -> str:
+    """Format the grounding context for injection."""
+    if not context or not context.get("success"):
+        return ""
+
+    grounding = context.get("grounding", {})
+
+    lines = ["[GROUNDING CONTEXT - VERIFY BEFORE RESPONDING]"]
+
+    # Current goal
+    if grounding.get("current_goal"):
+        lines.append(f"CURRENT GOAL: {grounding['current_goal']}")
+
+    # Entity registry
+    registry = grounding.get("entity_registry", {})
+    if registry:
+        lines.append("ENTITY REGISTRY (use these exact references):")
+        for key, value in list(registry.items())[:5]:
+            lines.append(f"  - {key}: {value}")
+
+    # Anchors (verified facts)
+    anchors = grounding.get("anchors", [])
+    if anchors:
+        lines.append("ANCHORS (verified facts - DO NOT CONTRADICT):")
+        for anchor in anchors[:5]:
+            lines.append(f"  - {anchor}")
+
+    # Recent decisions
+    decisions = grounding.get("decisions", [])
+    if decisions:
+        lines.append("RECENT DECISIONS:")
+        for decision in decisions[:3]:
+            lines.append(f"  - {decision}")
+
+    # Recent events
+    events = grounding.get("recent_events", [])
+    if events:
+        lines.append("RECENT EVENTS:")
+        for event in events[:5]:
+            lines.append(f"  - [{event.get('type', '?')}] {event.get('summary', '')}")
+
+    # Contradictions warning
+    contradictions = grounding.get("contradictions", [])
+    if contradictions:
+        lines.append("WARNING - POTENTIAL CONTRADICTIONS DETECTED:")
+        for c in contradictions[:3]:
+            lines.append(f"  - {c.get('content', '')[:100]}")
+
+    # Pending questions
+    pending = grounding.get("pending_questions", [])
+    if pending:
+        lines.append("PENDING QUESTIONS:")
+        for q in pending[:3]:
+            lines.append(f"  - {q}")
+
+    lines.append("[/GROUNDING CONTEXT]")
+    lines.append("")  # Empty line after
+
+    return "\n".join(lines)
+
+
+def format_memory_md_context(memory_md: dict) -> str:
+    """Format MEMORY.md content for injection."""
+    if not memory_md or not memory_md.get("exists"):
+        return ""
+
+    summary = memory_md.get("summary", "")
+    if not summary:
+        return ""
+
+    lines = ["[CORE FACTS from MEMORY.md]"]
+    lines.append(summary)
+    lines.append("[/CORE FACTS]")
+    lines.append("")
+
+    return "\n".join(lines)
+
+
+def format_daily_highlights(highlights: dict) -> str:
+    """Format daily log highlights for injection."""
+    if not highlights or not highlights.get("highlights"):
+        return ""
+
+    entries = highlights.get("highlights", [])
+    if not entries:
+        return ""
+
+    lines = ["[TODAY'S HIGHLIGHTS from Daily Log]"]
+    for entry in entries[:5]:
+        lines.append(f"  - {entry}")
+    lines.append("[/TODAY'S HIGHLIGHTS]")
+    lines.append("")
+
+    return "\n".join(lines)
+
+
+def format_curator_context(curator_summary: dict, curator_status: dict) -> str:
+    """Format curator context for injection."""
+    if not curator_summary and not curator_status:
+        return ""
+
+    lines = ["[CURATOR CONTEXT]"]
+
+    # Knowledge graph summary
+    if curator_summary:
+        context = curator_summary.get("context", "")
+        if context:
+            lines.append("Relevant Knowledge:")
+            for line in context.split("\n")[:10]:
+                if line.strip():
+                    lines.append(f"  {line}")
+
+        # Graph relationships
+        graph_context = curator_summary.get("graph_context")
+        if graph_context and graph_context.get("summary"):
+            lines.append("")
+            lines.append(f"Graph: {graph_context['summary']}")
+
+        # Pending reviews
+        pending = curator_summary.get("pending_reviews", {})
+        if pending.get("total_pending", 0) > 0:
+            lines.append("")
+            lines.append("Pending Reviews:")
+            if pending.get("duplicate_clusters", 0) > 0:
+                lines.append(f"  - {pending['duplicate_clusters']} duplicate clusters")
+            if pending.get("suggested_links", 0) > 0:
+                lines.append(f"  - {pending['suggested_links']} suggested links")
+            if pending.get("orphan_memories", 0) > 0:
+                lines.append(f"  - {pending['orphan_memories']} orphan memories")
+
+    # Curator status summary
+    if curator_status:
+        orphan_count = curator_status.get("orphan_count", 0)
+        connection_ratio = curator_status.get("connection_ratio", 0)
+        if orphan_count > 10:
+            lines.append(f"Warning: {orphan_count} orphan memories need linking")
+        if connection_ratio < 0.5:
+            lines.append(f"Note: Low graph connectivity ({connection_ratio:.1%})")
+
+    lines.append("[/CURATOR CONTEXT]")
+    lines.append("")
+
+    return "\n".join(lines)
+
+
+def check_and_trigger_flush(session_id: str, project_path: str):
+    """Check if flush is needed and trigger it."""
+    # Check flush conditions
+    flush_check = call_memory_agent("check_flush_needed", {
+        "session_id": session_id
+    })
+
+    if flush_check and flush_check.get("flush_needed"):
+        reasons = flush_check.get("reasons", [])
+        logger.info(f"Flush needed: {', '.join(reasons)}")
+
+        # Trigger flush
+        flush_result = call_memory_agent("pre_compaction_flush", {
+            "project_path": project_path,
+            "session_id": session_id
+        })
+
+        if flush_result and flush_result.get("success"):
+            logger.info(f"Flush completed: {flush_result.get('file_path')}")
+
+
+def main():
+    """Main entry point for the hook."""
+    project_path = get_project_path()
+    session_id = get_session_id()
+
+    # If no session, try to init one
+    if not session_id:
+        init_result = call_memory_agent("state_init_session", {
+            "project_path": project_path
+        })
+        if init_result and init_result.get("session_id"):
+            session_id = init_result["session_id"]
+            # Save session data as JSON
+            save_session_data({"session_id": session_id})
+
+    if not session_id:
+        # No session, no grounding - exit silently
+        sys.exit(0)
+
+    # ============================================================
+    # MOLTBOT-INSPIRED: Check flush conditions
+    # ============================================================
+    check_and_trigger_flush(session_id, project_path)
+
+    # ============================================================
+    # MOLTBOT-INSPIRED: Load MEMORY.md summary
+    # ============================================================
+    memory_md = call_memory_agent("get_memory_md_summary", {
+        "project_path": project_path
+    })
+
+    memory_md_context = format_memory_md_context(memory_md) if memory_md else ""
+
+    # ============================================================
+    # MOLTBOT-INSPIRED: Load today's daily log highlights
+    # ============================================================
+    daily_highlights = call_memory_agent("daily_log_highlights", {
+        "project_path": project_path
+    })
+
+    daily_context = format_daily_highlights(daily_highlights) if daily_highlights else ""
+
+    # ============================================================
+    # ORIGINAL: Get grounding context
+    # ============================================================
+    context = call_memory_agent("context_refresh", {
+        "session_id": session_id,
+        "include_recent_events": 5,
+        "include_state": True,
+        "include_checkpoint": True,
+        "check_contradictions": True
+    })
+
+    grounding_context = format_grounding_context(context) if context else ""
+
+    # ============================================================
+    # CURATOR: Get curated context and status
+    # ============================================================
+    # Get curator summary for current context (lightweight)
+    curator_summary = None
+    curator_status = None
+
+    # Only fetch curator context if there's user input to contextualize
+    user_input = os.getenv("CLAUDE_USER_INPUT", "")
+    if user_input and len(user_input) > 10:
+        curator_summary = call_memory_agent("curator_get_summary", {
+            "query": user_input[:500],  # Limit query length
+            "project_path": project_path,
+            "max_memories": 5,
+            "include_graph": True
+        })
+
+    # Always get curator status for warnings
+    curator_status = call_memory_agent("curator_get_status", {})
+
+    curator_context = format_curator_context(curator_summary, curator_status)
+
+    # Combine all context
+    output_parts = []
+
+    if memory_md_context:
+        output_parts.append(memory_md_context)
+
+    if daily_context:
+        output_parts.append(daily_context)
+
+    if grounding_context:
+        output_parts.append(grounding_context)
+
+    if curator_context:
+        output_parts.append(curator_context)
+
+    if output_parts:
+        print("\n".join(output_parts))
+
+    sys.exit(0)
+
+
+if __name__ == "__main__":
+    main()