get-claudia 1.42.1 → 1.42.3

package/README.md

@@ -5,7 +5,7 @@
 <h3 align="center">A thinking partner who tracks relationships, not just tasks.</h3>
 
 <p align="center">
-Catches commitments. Remembers context. Connects the dots across your network.
+Remembers your people. Catches your commitments. Learns how you work.
 </p>
 
 <p align="center">
@@ -17,6 +17,7 @@ Catches commitments. Remembers context. Connects the dots across your network.
 <p align="center">
 <a href="#try-it-in-30-seconds"><strong>Try the Demo</strong></a> ·
 <a href="#what-makes-claudia-different">Why Claudia</a> ·
+<a href="#how-her-mind-works">Her Mind</a> ·
 <a href="#quick-start">Install</a> ·
 <a href="#how-it-works">How It Works</a>
 </p>
@@ -42,26 +43,53 @@ You make a promise in a meeting. Nobody tracks it. You promise a deliverable on
 <p>Say "I'll send that by Friday" and she tracks it. On Friday morning, she reminds you.</p>
 </td>
 <td width="33%" align="center">
-<h3>🔗 Remembers Relationships</h3>
-<p>Mention Sarah from Acme and Claudia surfaces: last conversation, open promises, sentiment.</p>
+<h3>🔗 Knows Your People</h3>
+<p>Every person she meets gets a living profile: relationship health, contact trends, connected entities. Ask about anyone and she has the full picture.</p>
 </td>
 <td width="33%" align="center">
-<h3>⚠️ Warns Before Things Slip</h3>
-<p>Haven't talked to your best client in 3 weeks? She tells you before it's a problem.</p>
+<h3>⚠️ Spots Patterns You Miss</h3>
+<p>Overcommitting again? A key relationship going cold? The same mistake twice? She sees it forming and speaks up.</p>
 </td>
 </tr>
 <tr>
 <td width="33%" align="center">
-<h3>📄 Shows Her Work</h3>
+<h3>📄 Shows Her Sources</h3>
 <p>Every fact traces to its source. Ask "how do you know that?" and she shows the receipt.</p>
 </td>
 <td width="33%" align="center">
-<h3>🧠 Obsidian Vault Sync</h3>
-<p>Memory syncs to an Obsidian vault as markdown notes with wikilinks. Graph view visualizes your entire network. Plain files you own forever.</p>
+<h3>🧠 Second Brain in Obsidian</h3>
+<p>Memory syncs to an Obsidian vault organized by activity: Active projects, Relationships, Reference, Archive. Graph view maps your world. Plain markdown you own forever.</p>
 </td>
 <td width="33%" align="center">
-<h3>🔒 Fully Local</h3>
-<p>Everything runs on your machine. Your data never leaves. Delete anything, anytime.</p>
+<h3>🌙 Learns in the Background</h3>
+<p>Overnight, old memories fade, near-duplicates merge, and patterns surface. Each morning she knows a little more than yesterday.</p>
+</td>
+</tr>
+</table>
+
+---
+
+## How Her Mind Works
+
+<table>
+<tr>
+<td width="50%" align="center">
+<h3>💾 Remember</h3>
+<p>Every fact is stored with who said it, when, and how confident she is. Embeddings capture <em>meaning</em>, not just keywords, so "we pushed the launch" and "timeline shifted" connect naturally.</p>
+</td>
+<td width="50%" align="center">
+<h3>🔍 Recall</h3>
+<p>Search blends meaning similarity, importance, recency, and full-text matching. Accessing a memory strengthens it, just like the rehearsal effect in human cognition.</p>
+</td>
+</tr>
+<tr>
+<td width="50%" align="center">
+<h3>🌙 Consolidate</h3>
+<p>Overnight background jobs fade old memories, merge near-duplicates, and surface patterns: cooling relationships, overdue commitments, repeated behaviors. She wakes up sharper.</p>
+</td>
+<td width="50%" align="center">
+<h3>📓 Vault</h3>
+<p>Memory syncs to an Obsidian vault organized with PARA: Active projects, Relationships, Reference material, Archive. Graph view shows how your world connects. Plain markdown you own forever.</p>
 </td>
 </tr>
 </table>
@@ -153,7 +181,37 @@ Claudia: Sarah Chen is CEO at Meridian Ventures. You met at SaaStr 2025.
          Source: Call notes from Jan 28, Email thread Jan 30
 ```
 
-From here, every conversation builds on the last. She remembers.
+**Morning brief pulls everything together:**
+
+```
+You:     /morning-brief
+
+Claudia: ☀️ Morning Brief — Wednesday, Feb 19
+
+         ⚠️ Needs Attention:
+         • Send updated pitch deck to Sarah Chen [OVERDUE]
+         • Quarterly review with Acme Corp is tomorrow
+
+         👀 Watch:
+         • Haven't spoken to Marcus in 18 days (usually weekly)
+         • Two new commitments this week, zero closed
+
+         📊 Today: 3 meetings, 2 open commitments
+```
+
+**Pattern detection works across weeks, not just moments:**
+
+```
+Claudia: I've noticed something across the last few weeks.
+         You've taken on 4 new commitments without closing any.
+         Last time this happened (early January), two slipped
+         past deadline.
+
+         Want me to flag the lowest-priority ones so you can
+         decide what to defer?
+```
+
+Every conversation builds on the last. She remembers.
 
 ---
 
@@ -180,8 +238,9 @@ Claudia detects your work style and generates structure that fits:
 | `/meeting-prep [person]` | One-page briefing before a call |
 | `/capture-meeting` | Process notes into decisions, commitments, action items |
 | `/what-am-i-missing` | Surface risks, overdue items, cooling relationships |
-| `/inbox-check` | Review Telegram and webhook messages |
-| `/gateway` | Manage external message gateway |
+| `/sync-vault` | Sync memory to Obsidian vault |
+| `/brain` | Launch 3D brain visualizer |
+| `/deep-context [topic]` | Full-context deep analysis |
 | `/memory-audit` | See everything Claudia knows, with source chains |
 
 <details>
@@ -196,8 +255,10 @@ Claudia detects your work style and generates structure that fits:
 | `/new-person [name]` | Create a relationship file |
 | `/pipeline-review` | Active deals, capacity, pipeline health |
 | `/client-health` | Status across all client relationships |
-| `/setup-gateway` | Configure external message gateway |
-| `/setup-telegram` | Connect Telegram bot integration |
+| `/inbox-check` | Review messages from connected channels |
+| `/meditate` | End-of-session reflection and persistent learnings |
+| `/fix-duplicates` | Find and merge duplicate entities |
+| `/memory-health` | Check memory system health |
 | `/diagnose` | Check memory daemon health and troubleshoot |
 
 </details>
@@ -224,12 +285,14 @@ You ──► Claude Code ──► Reads Claudia's templates ──► Becomes
                                                            ▼
                               Memory daemon (local) ◄── MCP tools
                                       │
-                               ┌──────┴──────┐
-                               ▼              ▼
-              SQLite + vectors + Ollama    Obsidian vault
-                   (all local)           (markdown notes)
+                               ┌──────┼──────┐
+                               ▼      ▼      ▼
+                           SQLite  Ollama  Obsidian vault
+                          +vectors         (PARA structure)
 ```
 
+**Local extraction, zero API calls.** Paste a meeting transcript or email thread. A local language model (Qwen3, SmolLM3, or Llama 3.2 via Ollama) extracts entities, commitments, and decisions in seconds. Claude reviews the extractions and applies judgment. No data leaves your machine.
+
 <details>
 <summary><strong>Technical deep dive</strong></summary>
 
@@ -249,26 +312,13 @@ For full architecture diagrams, see [ARCHITECTURE.md](ARCHITECTURE.md).
 
 </details>
 
-<details>
-<summary><strong>Cognitive tools (local LLM extraction)</strong></summary>
-
-Paste a meeting transcript. A local language model extracts structured data (entities, facts, commitments) in seconds. Claude reviews and applies judgment.
-
-- Runs locally via [Ollama](https://ollama.com), no API keys
-- Models: Qwen3-4B (recommended), SmolLM3-3B, Llama 3.2-3B
-- Falls back gracefully when no model installed
-
-Four extraction modes: **meeting**, **email**, **document**, **general**.
-
-</details>
-
 ---
 
 ## Privacy and Safety
 
 - **Fully local.** Memory, embeddings, cognitive tools run on your machine. No external APIs for storage.
 - **No external actions without approval.** Every email, calendar event, external action requires your explicit "yes."
-- **Your data in two formats.** SQLite database (`~/.claudia/memory/`) for fast semantic search, plus plain markdown files in your Obsidian vault for reading and graph navigation. Two independent copies you own forever.
+- **Your data in two formats.** SQLite database (`~/.claudia/memory/`) for fast semantic search, plus a PARA-organized Obsidian vault for reading and graph navigation. Two independent copies you own forever.
 - **Delete anything, anytime.** Full control over your data. No lock-in, no cloud dependency.
 
 ---

package/bin/index.js

@@ -376,7 +376,7 @@ async function main() {
         stdio: 'inherit',
         env: {
           ...process.env,
-          CLAUDIA_PROJECT_PATH: isUpgrade ? targetPath : '',
+          CLAUDIA_PROJECT_PATH: targetPath,
           CLAUDIA_NONINTERACTIVE: '1'
         }
       });

package/memory-daemon/claudia_memory/daemon/health.py

@@ -160,6 +160,31 @@ class HealthCheckHandler(BaseHTTPRequestHandler):
         else:
             self.send_error(404, "Not Found")
 
+    def do_POST(self):
+        """Handle POST requests"""
+        if self.path == "/backup":
+            self._send_backup_response()
+        else:
+            self.send_error(405, "Method Not Allowed")
+
+    def _send_backup_response(self):
+        """Trigger a database backup and return the path."""
+        try:
+            db = get_db()
+            path = db.backup()
+            self._send_json({"status": "ok", "path": str(path)})
+        except Exception as e:
+            logger.exception("Error triggering backup")
+            self._send_json({"status": "error", "message": str(e)}, code=500)
+
+    def _send_json(self, data: dict, code: int = 200):
+        """Helper to send a JSON response."""
+        body = json.dumps(data).encode()
+        self.send_response(code)
+        self.send_header("Content-Type", "application/json")
+        self.end_headers()
+        self.wfile.write(body)
+
     def _send_health_response(self):
         """Send basic health check response"""
         health = {

package/memory-daemon/claudia_memory/mcp/server.py

@@ -722,6 +722,19 @@ async def list_tools() -> ListToolsResult:
                 "properties": {},
             },
         ),
+        Tool(
+            name="memory.backup",
+            title="Trigger Database Backup",
+            description=(
+                "Trigger an immediate backup of the memory database. Returns the path "
+                "of the newly created backup file. Backups use a timestamp suffix and "
+                "older backups are pruned automatically per the retention policy."
+            ),
+            inputSchema={
+                "type": "object",
+                "properties": {},
+            },
+        ),
         Tool(
             name="memory.project_health",
             title="Project Health Check",
@@ -2587,8 +2600,14 @@ async def call_tool(name: str, arguments: Dict[str, Any]) -> CallToolResult:
                 )
 
             elif name == "memory.system_health":
-                from ..daemon.health import build_status_report
-                report = build_status_report()
+                import urllib.request, urllib.error
+                report = None
+                try:
+                    with urllib.request.urlopen("http://localhost:3848/status", timeout=2) as resp:
+                        report = json.loads(resp.read().decode())
+                except (urllib.error.URLError, OSError):
+                    from ..daemon.health import build_status_report
+                    report = build_status_report()
                 embedding_svc = get_embedding_service()
                 if hasattr(embedding_svc, '_model_mismatch') and embedding_svc._model_mismatch:
                     if "components" not in report:
@@ -2603,6 +2622,28 @@ async def call_tool(name: str, arguments: Dict[str, Any]) -> CallToolResult:
                     ]
                 )
 
+            elif name == "memory.backup":
+                import urllib.request, urllib.error
+                result = None
+                try:
+                    req = urllib.request.Request(
+                        "http://localhost:3848/backup", method="POST", data=b""
+                    )
+                    with urllib.request.urlopen(req, timeout=10) as resp:
+                        result = json.loads(resp.read().decode())
+                except (urllib.error.URLError, OSError):
+                    # Daemon not running — trigger backup directly
+                    backup_path = get_db().backup()
+                    result = {"status": "ok", "path": str(backup_path)}
+                return CallToolResult(
+                    content=[
+                        TextContent(
+                            type="text",
+                            text=json.dumps(result, indent=2),
+                        )
+                    ]
+                )
+
             elif name == "memory.sync_vault":
                 from ..config import _project_id
                 from ..services.vault_sync import run_vault_sync

package/memory-daemon/claudia_memory/services/consolidate.py

@@ -2010,6 +2010,49 @@ class ConsolidateService:
 
         logger.debug(f"Merged reflection {duplicate['id']} into {primary['id']}")
 
+    def close_stale_episodes(self) -> int:
+        """Auto-close orphan episodes that have no end_session call.
+
+        Single-turn sessions and interrupted sessions leave episodes with
+        ``ended_at IS NULL``.  After 24 hours these will never be closed
+        naturally, so this pass marks them as summarized with a synthetic
+        summary to prevent them from appearing as false positives in health
+        reports.
+
+        Returns the number of episodes closed.
+        """
+        cutoff = (datetime.utcnow() - timedelta(hours=24)).isoformat()
+        try:
+            # Find stale open episodes and close them.  ended_at is set to the
+            # timestamp of the latest buffered turn if one exists, otherwise to
+            # started_at itself (single-turn / empty sessions).
+            self.db.execute(
+                """
+                UPDATE episodes
+                SET
+                    ended_at = COALESCE(
+                        (
+                            SELECT MAX(created_at) FROM turn_buffer
+                            WHERE turn_buffer.session_id = episodes.session_id
+                        ),
+                        started_at
+                    ),
+                    is_summarized = 1,
+                    summary = 'Auto-closed: session ended without explicit end_session call'
+                WHERE ended_at IS NULL
+                  AND started_at < ?
+                """,
+                (cutoff,),
+            )
+            rows = self.db.execute("SELECT changes()", fetch=True)
+            count = rows[0][0] if rows else 0
+            if count:
+                logger.info(f"Auto-closed {count} stale open episode(s)")
+            return count
+        except Exception as e:
+            logger.warning(f"close_stale_episodes failed: {e}")
+            return 0
+
     def run_retention_cleanup(self) -> Dict[str, int]:
         """Clean up old data per retention policies.
 
@@ -2018,6 +2061,7 @@ class ConsolidateService:
         - Expired predictions past retention window
         - Archived turn_buffer from old episodes
         - Old metrics rows
+        - Auto-closes stale open episodes (no end_session after 24 h)
         """
         results = {}
         now = datetime.utcnow()
@@ -2074,6 +2118,9 @@ class ConsolidateService:
             logger.warning(f"Metrics cleanup failed: {e}")
             results["metrics_deleted"] = 0
 
+        # Auto-close orphan episodes (no end_session after 24 h)
+        results["stale_episodes_closed"] = self.close_stale_episodes()
+
         logger.info(f"Retention cleanup: {results}")
         return results
 

package/memory-daemon/scripts/install.ps1

@@ -366,9 +366,13 @@ try {
     }
 
     # Create the scheduled task action
+    $taskArgs = "-m claudia_memory --standalone"
+    if ($env:CLAUDIA_PROJECT_PATH) {
+        $taskArgs += " --project-dir `"$($env:CLAUDIA_PROJECT_PATH)`""
+    }
     $action = New-ScheduledTaskAction `
         -Execute $VENV_PYTHON `
-        -Argument "-m claudia_memory --standalone" `
+        -Argument $taskArgs `
         -WorkingDirectory $DAEMON_DIR
 
     # Trigger: at logon for current user

package/memory-daemon/scripts/install.sh

@@ -518,7 +518,7 @@ if [[ "$OSTYPE" == "darwin"* ]]; then
         <string>$VENV_DIR/bin/python</string>
         <string>-m</string>
         <string>claudia_memory</string>
-        <string>--standalone</string>
+        <string>--standalone</string>$(if [ -n "$CLAUDIA_PROJECT_PATH" ]; then printf '\n        <string>--project-dir</string>\n        <string>%s</string>' "$CLAUDIA_PROJECT_PATH"; fi)
     </array>
     <key>WorkingDirectory</key>
     <string>$DAEMON_DIR</string>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "get-claudia",
-  "version": "1.42.1",
+  "version": "1.42.3",
   "description": "An AI assistant who learns how you work.",
   "keywords": [
     "claudia",