get-claudia 1.55.14 → 1.55.16

package/CHANGELOG.md

@@ -2,6 +2,24 @@
 
 All notable changes to Claudia will be documented in this file.
 
+## 1.55.16 (2026-03-18)
+
+### Reliability Fixes
+
+Three fixes for issues surfaced from daemon logs. All backward-compatible, no schema changes.
+
+- **Overnight jobs now fire after sleep** -- APScheduler's `BackgroundScheduler` now has `misfire_grace_time=14400` (4 hours) and `coalesce=True`. Previously, the default 1-second grace time meant every scheduled job (decay, backup, consolidation, vault sync) was silently skipped when a Mac slept through the 2am-3:15am window. Now jobs fire immediately on wake if missed within the last 4 hours, with multiple missed runs collapsed into one execution.
+- **Reduced log noise from summary memories** -- The content length warning threshold was raised from 500 to 800 chars. Legitimate summary-type memories (550-850 chars) no longer trigger "Long content" warnings. Hard truncation at 1000 chars is unchanged.
+- **Fuzzy entity dedup on write** -- `_ensure_entity()` and `_find_or_create_entity()` now perform a fuzzy pre-check (SequenceMatcher > 0.90) before creating new entities. Name variants like "Kris Krisko" vs "Kris Krisco" (ratio ~0.92) match the existing entity instead of creating a duplicate. Only compares entities of the same type, skips deleted entities.
+- **Expanded STOP_WORDS** -- Added ~55 common English words that spaCy misidentifies as entities ("drawn", "overall", "recently", "several", etc.). Prevents ghost entities from cluttering the graph.
+- **Person entities require 2+ words** -- Regex-extracted person entities must have at least two words (e.g., "First Last"). Single-word extractions like "Metal" or "Drawn" are rejected. spaCy-identified entities are unaffected.
+- 637 tests pass, 0 regressions, 22 new tests across 4 new test files.
+
+## 1.55.15 (2026-03-18)
+
+- **Fix mixed-timezone datetime crash** -- The memory daemon could crash with `can't subtract offset-naive and offset-aware datetimes` when recall or consolidation queries hit records with timezone suffixes (e.g., `+00:00` from email or transcript timestamps). Added a shared `parse_naive()` utility that strips timezone info on parse, applied across 14 locations in 5 files (recall.py, consolidate.py, server.py, vault_sync.py, canvas_generator.py). Replaces the older `[:19]` string truncation workaround. 615 tests pass.
+- **License updated to PolyForm Noncommercial 1.0.0** -- README, package.json, and ARCHITECTURE.md now reflect the license change from Apache 2.0 to PolyForm NC. Free for personal, research, educational, and nonprofit use. Commercial licensing available via mail@kbanc.com.
+
 ## 1.55.14 (2026-03-16)
 
 - **LaunchAgent no longer bakes in --project-dir** -- The standalone background daemon now starts without a `--project-dir` argument. This forces a plist content change for all existing installs, which triggers an automatic LaunchAgent reload on next `claudia setup`, picking up the current Python daemon code. Previously, the plist could be identical across updates, leaving old daemon code running indefinitely even after `pip install --upgrade`.

package/README.md

@@ -11,7 +11,7 @@ Remembers your people. Catches your commitments. Learns how you work.
 <p align="center">
   <a href="https://github.com/kbanc85/claudia/stargazers"><img src="https://img.shields.io/github/stars/kbanc85/claudia?style=flat-square" alt="GitHub stars"></a>
   <a href="https://www.npmjs.com/package/get-claudia"><img src="https://img.shields.io/npm/v/get-claudia?style=flat-square" alt="npm version"></a>
-  <a href="https://github.com/kbanc85/claudia/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License"></a>
+  <a href="https://github.com/kbanc85/claudia/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-PolyForm%20NC%201.0.0-purple?style=flat-square" alt="License"></a>
 </p>
 
 <p align="center">
@@ -479,7 +479,7 @@ This updates daemon code, skills, and rules while preserving your databases and
 
 ## Contributing
 
-Claudia is open source under the Apache 2.0 License.
+Claudia is source-available under the PolyForm Noncommercial License 1.0.0.
 
 - **Template (skills, rules, identity):** `template-v2/`
 - **Memory daemon (Python):** `memory-daemon/` (tests: `cd memory-daemon && pytest tests/`)
@@ -491,9 +491,9 @@ Claudia is open source under the Apache 2.0 License.
 
 ## License
 
-[Apache License 2.0](LICENSE)
+[PolyForm Noncommercial 1.0.0](LICENSE)
 
-Open source. Free for personal and commercial use. Attribution required.
+Free for personal, research, educational, and nonprofit use. Commercial licensing: mail@kbanc.com
 
 ---
 

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

@@ -27,7 +27,9 @@ class MemoryScheduler:
     """Manages scheduled memory maintenance tasks"""
 
     def __init__(self):
-        self.scheduler = BackgroundScheduler()
+        self.scheduler = BackgroundScheduler(
+            job_defaults={"misfire_grace_time": 14400, "coalesce": True}
+        )
         self.config = get_config()
         self._started = False
 

package/memory-daemon/claudia_memory/extraction/entity_extractor.py

@@ -188,50 +188,33 @@ class EntityExtractor:
 
     # Common non-entity words to filter out
     STOP_WORDS = {
-        "monday",
-        "tuesday",
-        "wednesday",
-        "thursday",
-        "friday",
-        "saturday",
-        "sunday",
-        "january",
-        "february",
-        "march",
-        "april",
-        "may",
-        "june",
-        "july",
-        "august",
-        "september",
-        "october",
-        "november",
-        "december",
-        "today",
-        "tomorrow",
-        "yesterday",
-        "morning",
-        "afternoon",
-        "evening",
-        "night",
-        "the",
-        "this",
-        "that",
-        "these",
-        "those",
-        "here",
-        "there",
-        "where",
-        "when",
-        "what",
-        "which",
-        "who",
-        "how",
-        "just",
-        "only",
-        "also",
-        "even",
-        "still",
+        # Days and months
+        "monday", "tuesday", "wednesday", "thursday", "friday",
+        "saturday", "sunday",
+        "january", "february", "march", "april", "may", "june",
+        "july", "august", "september", "october", "november", "december",
+        # Temporal
+        "today", "tomorrow", "yesterday",
+        "morning", "afternoon", "evening", "night",
+        # Pronouns and determiners
+        "the", "this", "that", "these", "those",
+        "here", "there", "where", "when", "what", "which", "who", "how",
+        # Adverbs
+        "just", "only", "also", "even", "still",
+        "recently", "nearly", "almost", "already", "rather",
+        "somewhat", "perhaps", "quite", "likely", "enough",
+        # Quantifiers and adjectives
+        "several", "various", "another", "certain",
+        "much", "many", "some", "most", "both",
+        "each", "every", "other", "such", "same",
+        "new", "old", "big", "long", "last", "next",
+        "good", "well", "nice", "overall", "drawn",
+        # Common verbs (past tense / short forms spaCy misidentifies)
+        "done", "made", "said", "went", "got",
+        "set", "put", "run", "let", "get",
+        # Common nouns too generic to be entities
+        "work", "part", "plan", "team", "data",
+        "note", "time", "home", "call", "open",
     }
 
     def __init__(self):
@@ -296,12 +279,15 @@ class EntityExtractor:
         """Extract entities using regex patterns"""
         entities = []
 
-        # Extract persons
+        # Extract persons (require at least 2 words to avoid ghost entities)
         for pattern in self.PERSON_PATTERNS:
             for match in re.finditer(pattern, text):
                 name = match.group(1) if match.lastindex else match.group(0)
                 canonical = self.canonical_name(name)
-                if canonical and len(canonical) > 1 and canonical not in self.STOP_WORDS:
+                if (canonical
+                        and len(canonical) > 1
+                        and canonical not in self.STOP_WORDS
+                        and len(canonical.split()) >= 2):
                     entities.append(
                         ExtractedEntity(
                             name=name,

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

@@ -22,6 +22,7 @@ from mcp.types import (
 )
 
 from ..database import get_db
+from ..utils import parse_naive
 from ..services.consolidate import (
     get_consolidate_service,
     get_predictions,
@@ -3190,9 +3191,9 @@ def _build_briefing() -> str:
                     "SELECT updated_at FROM _meta WHERE key = 'unified_db'", fetch=True
                 )
                 if ts_row and ts_row[0]["updated_at"]:
-                    from datetime import datetime as _dt, timedelta as _td
-                    consolidated_at = _dt.fromisoformat(ts_row[0]["updated_at"][:19])
-                    if (_dt.utcnow() - consolidated_at) < _td(minutes=5):
+                    from datetime import timedelta as _td
+                    consolidated_at = parse_naive(ts_row[0]["updated_at"])
+                    if (datetime.utcnow() - consolidated_at) < _td(minutes=5):
                         # Just consolidated, include stats
                         mem_row = db.execute("SELECT COUNT(*) as c FROM memories", fetch=True)
                         ent_row = db.execute("SELECT COUNT(*) as c FROM entities WHERE deleted_at IS NULL", fetch=True)
@@ -3660,7 +3661,7 @@ def _build_morning_context() -> str:
         if stale:
             sections.append(f"## Stale Commitments ({len(stale)})\n")
             for c in stale:
-                days_old = (datetime.utcnow() - datetime.fromisoformat(c["created_at"])).days
+                days_old = (datetime.utcnow() - parse_naive(c["created_at"])).days
                 entities = c["entity_names"] or ""
                 prefix = f"[{entities}] " if entities else ""
                 sections.append(f"- {prefix}{c['content'][:100]} ({days_old}d old, importance: {c['importance']:.1f})")

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

@@ -28,6 +28,7 @@ from pathlib import Path
 from typing import Any, Dict, List, Optional, Tuple
 
 from ..database import get_db
+from ..utils import parse_naive
 
 logger = logging.getLogger(__name__)
 
@@ -418,7 +419,7 @@ class CanvasGenerator:
                 last = r["last_contact_at"]
                 if last:
                     try:
-                        days_ago = (datetime.utcnow() - datetime.fromisoformat(last[:19])).days
+                        days_ago = (datetime.utcnow() - parse_naive(last)).days
                         reconnect_lines.append(f"- [[{r['name']}]] ({trend}, {days_ago}d ago)")
                     except (ValueError, TypeError):
                         reconnect_lines.append(f"- [[{r['name']}]] ({trend})")

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

@@ -14,6 +14,7 @@ from typing import Any, Dict, List, Optional, Tuple
 
 from ..config import get_config
 from ..database import get_db
+from ..utils import parse_naive
 
 logger = logging.getLogger(__name__)
 
@@ -366,7 +367,7 @@ class ConsolidateService:
             timestamps = []
             for r in rows:
                 try:
-                    timestamps.append(datetime.fromisoformat(r["created_at"]))
+                    timestamps.append(parse_naive(r["created_at"]))
                 except (ValueError, TypeError):
                     continue
 
@@ -538,7 +539,7 @@ class ConsolidateService:
             days_since = 0
             if entity["last_contact_at"]:
                 try:
-                    last_dt = datetime.fromisoformat(entity["last_contact_at"])
+                    last_dt = parse_naive(entity["last_contact_at"])
                     days_since = int((now - last_dt).total_seconds() / 86400)
                 except (ValueError, TypeError):
                     pass
@@ -671,7 +672,7 @@ class ConsolidateService:
         for row in rows:
             days_since = None
             if row["last_mention"]:
-                last_dt = datetime.fromisoformat(row["last_mention"])
+                last_dt = parse_naive(row["last_mention"])
                 days_since = (datetime.utcnow() - last_dt).days
 
             severity = "warning" if days_since and days_since > 60 else "observation"
@@ -1377,7 +1378,7 @@ class ConsolidateService:
         )
 
         for commitment in old_commitments:
-            created = datetime.fromisoformat(commitment["created_at"])
+            created = parse_naive(commitment["created_at"])
             days_old = (datetime.utcnow() - created).days
 
             if days_old > 3:
@@ -2302,7 +2303,7 @@ class ConsolidateService:
                 velocity_parts.append(f"tier: {entity['attention_tier']}")
             if entity["last_contact_at"]:
                 try:
-                    last_dt = datetime.fromisoformat(entity["last_contact_at"])
+                    last_dt = parse_naive(entity["last_contact_at"])
                     days_since = (datetime.utcnow() - last_dt).days
                     velocity_parts.append(f"last contact: {days_since} days ago")
                 except (ValueError, TypeError):

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

@@ -42,7 +42,7 @@ def validate_memory(
     Validate a memory before storage.
 
     Checks:
-    - Content length (warn >500, truncate >1000)
+    - Content length (warn >800, truncate >1000)
     - Commitment deadline detection via regex
     - Importance clamped to [0, 1]
     """
@@ -52,7 +52,7 @@ def validate_memory(
     if len(content) > 1000:
         result.warnings.append(f"Content truncated from {len(content)} to 1000 characters")
         result.adjustments["content"] = content[:1000]
-    elif len(content) > 500:
+    elif len(content) > 800:
         result.warnings.append(f"Long content ({len(content)} chars) -- consider breaking into multiple memories")
 
     # Importance clamping

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

@@ -18,6 +18,7 @@ from typing import Any, Dict, List, Optional, Tuple
 from ..config import get_config
 from ..database import get_db
 from ..embeddings import embed_sync, get_embedding_service
+from ..utils import parse_naive
 from ..extraction.entity_extractor import get_extractor
 
 logger = logging.getLogger(__name__)
@@ -240,7 +241,7 @@ class RecallService:
                 row = vector_rows.get(mid)
                 if row:
                     try:
-                        created = datetime.fromisoformat(row["created_at"])
+                        created = parse_naive(row["created_at"])
                         recency_data[mid] = (now - created).total_seconds()
                     except (ValueError, TypeError):
                         recency_data[mid] = float("inf")
@@ -333,7 +334,7 @@ class RecallService:
         importance_score = row["importance"]
 
         # Recency score (configurable half-life decay)
-        created = datetime.fromisoformat(row["created_at"])
+        created = parse_naive(row["created_at"])
         days_old = (now - created).days
         recency_score = math.exp(-days_old / self.config.recency_half_life_days)
 
@@ -2122,8 +2123,8 @@ class RecallService:
             results = []
             now = datetime.utcnow()
             for row in rows:
-                source_last = datetime.fromisoformat(row["source_last_memory"])
-                target_last = datetime.fromisoformat(row["target_last_memory"])
+                source_last = parse_naive(row["source_last_memory"])
+                target_last = parse_naive(row["target_last_memory"])
                 most_recent = max(source_last, target_last)
                 days_dormant = (now - most_recent).days
 
@@ -2513,7 +2514,7 @@ class RecallService:
             urgency = "later"
             if deadline_str:
                 try:
-                    deadline_dt = datetime.fromisoformat(deadline_str)
+                    deadline_dt = parse_naive(deadline_str)
                     if deadline_dt < now:
                         urgency = "overdue"
                     elif deadline_dt < now + timedelta(days=1):
@@ -2705,12 +2706,9 @@ class RecallService:
             }
 
         try:
-            last_dt = datetime.strptime(last_contact, "%Y-%m-%d %H:%M:%S")
-        except (ValueError, TypeError):
-            try:
-                last_dt = datetime.fromisoformat(last_contact.replace("Z", "+00:00")).replace(tzinfo=None)
-            except Exception:
-                return {"entity": entity["name"], "status": "parse_error"}
+            last_dt = parse_naive(last_contact.replace("Z", "+00:00"))
+        except Exception:
+            return {"entity": entity["name"], "status": "parse_error"}
 
         now = datetime.utcnow()
         days_since = (now - last_dt).days

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

@@ -1697,6 +1697,11 @@ class RememberService:
         if alias_match:
             return alias_match["entity_id"]
 
+        # Fuzzy pre-check: find near-matches of the same type
+        fuzzy_match = self._fuzzy_find_entity(extracted.canonical_name, extracted.type)
+        if fuzzy_match:
+            return fuzzy_match
+
         # Create new entity
         return self.remember_entity(
             name=extracted.name,
@@ -1725,9 +1730,44 @@ class RememberService:
         if alias_match:
             return alias_match["entity_id"]
 
+        # Fuzzy pre-check: find near-matches of the same type
+        fuzzy_match = self._fuzzy_find_entity(canonical, entity_type)
+        if fuzzy_match:
+            return fuzzy_match
+
         # Create new
         return self.remember_entity(name=name, entity_type=entity_type)
 
+    def _fuzzy_find_entity(self, canonical: str, entity_type: str) -> Optional[int]:
+        """Find a near-match entity of the same type using fuzzy string matching.
+
+        Queries entities of the given type and returns the ID of the best match
+        if similarity > 0.90 (SequenceMatcher ratio). Returns None if no match.
+        """
+        from difflib import SequenceMatcher
+
+        candidates = self.db.execute(
+            "SELECT id, canonical_name FROM entities WHERE type = ? AND deleted_at IS NULL",
+            (entity_type,),
+            fetch=True,
+        ) or []
+
+        best_id = None
+        best_ratio = 0.0
+        for row in candidates:
+            ratio = SequenceMatcher(None, canonical, row["canonical_name"]).ratio()
+            if ratio > 0.90 and ratio > best_ratio:
+                best_ratio = ratio
+                best_id = row["id"]
+
+        if best_id is not None:
+            logger.info(
+                f"Fuzzy entity match: '{canonical}' matched existing entity id={best_id} "
+                f"(type={entity_type}, similarity={best_ratio:.2f})"
+            )
+
+        return best_id
+
     def _get_or_create_episode(self, source: Optional[str] = None) -> int:
         """Get current episode or create a new one"""
         # For now, create a new episode each time

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

@@ -42,6 +42,7 @@ from typing import Any, Dict, List, Optional, Tuple
 
 from ..config import get_config
 from ..database import get_db
+from ..utils import parse_naive
 
 logger = logging.getLogger(__name__)
 
@@ -1234,7 +1235,7 @@ class VaultSyncService:
                 last = w["last_contact_at"]
                 if last:
                     try:
-                        days_ago = (datetime.utcnow() - datetime.fromisoformat(last[:19])).days
+                        days_ago = (datetime.utcnow() - parse_naive(last)).days
                         lines.append(f"- [[{w['name']}]] - {trend} ({days_ago}d)")
                     except (ValueError, TypeError):
                         lines.append(f"- [[{w['name']}]] - {trend}")
@@ -1599,7 +1600,7 @@ class VaultSyncService:
                 last_contact = p["last_contact_at"]
                 if last_contact:
                     try:
-                        dt = datetime.fromisoformat(last_contact[:19])
+                        dt = parse_naive(last_contact)
                         days_ago = (now - dt).days
                         last_str = f"{days_ago}d ago"
                     except (ValueError, TypeError):

package/memory-daemon/claudia_memory/utils.py

@@ -0,0 +1,22 @@
+"""
+Shared utilities for Claudia Memory System.
+"""
+
+from datetime import datetime
+
+
+def parse_naive(dt_string: str) -> datetime:
+    """Parse an ISO datetime string and strip timezone info.
+
+    The database stores a mix of naive and offset-aware datetimes.
+    External sources (emails, transcripts, calendar events) often include
+    timezone suffixes like +00:00 or Z. Since all timestamps are treated
+    as UTC internally, we strip tzinfo to avoid:
+
+        TypeError: can't subtract offset-naive and offset-aware datetimes
+
+    This is used everywhere a parsed timestamp participates in arithmetic
+    with datetime.utcnow() (which returns a naive datetime).
+    """
+    dt = datetime.fromisoformat(dt_string)
+    return dt.replace(tzinfo=None) if dt.tzinfo else dt

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "get-claudia",
-  "version": "1.55.14",
+  "version": "1.55.16",
   "description": "An AI assistant who learns how you work.",
   "keywords": [
     "claudia",
@@ -16,7 +16,7 @@
     "adaptive"
   ],
   "author": "Kamil Banc",
-  "license": "Apache-2.0",
+  "license": "SEE LICENSE IN LICENSE",
   "repository": {
     "type": "git",
     "url": "git+https://github.com/kbanc85/claudia.git"