yourmemory 1.2.1__tar.gz → 1.2.2__tar.gz

{yourmemory-1.2.1/yourmemory.egg-info → yourmemory-1.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: yourmemory
-Version: 1.2.1
+Version: 1.2.2
 Summary: Persistent memory for Claude — Ebbinghaus forgetting curve, semantic deduplication, MCP-native
 Author-email: Sachit Misra <mishrasachit1@gmail.com>
 License:                                  Apache License
@@ -163,8 +163,8 @@ License:                                  Apache License
            See the License for the specific language governing permissions and
            limitations under the License.
         
-Project-URL: Homepage, https://github.com/sachitrafa/cognitive-ai-memory
-Project-URL: Repository, https://github.com/sachitrafa/cognitive-ai-memory
+Project-URL: Homepage, https://github.com/sachitrafa/YourMemory
+Project-URL: Repository, https://github.com/sachitrafa/YourMemory
 Keywords: mcp,claude,memory,ebbinghaus,ai,sqlite,postgresql
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.11
@@ -246,6 +246,8 @@ Importance additionally modulates the decay rate within each category. Memories
 
 **Zero infrastructure required** — uses DuckDB out of the box. Two commands and you're done.
 
+Supports **Python 3.11, 3.12, 3.13, and 3.14**.
+
 ### 1. Install
 
 ```bash
@@ -286,9 +288,13 @@ Reload Claude Code (`Cmd+Shift+P` → `Developer: Reload Window`).
 
 #### Cline (VS Code)
 
-VS Code doesn't inherit your shell PATH, so use the **full path** from `yourmemory-path`.
+VS Code doesn't inherit your shell PATH. Run this in terminal to get the exact config to paste:
+
+```bash
+yourmemory-path
+```
 
-In Cline → **MCP Servers** → **Edit MCP Settings**:
+Then in Cline → **MCP Servers** → **Edit MCP Settings**, paste the output. It looks like:
 
 ```json
 {
@@ -305,7 +311,7 @@ In Cline → **MCP Servers** → **Edit MCP Settings**:
 }
 ```
 
-Run `yourmemory-path` in terminal — it prints the exact config to paste.
+Restart Cline after saving.
 
 #### Cursor
 
@@ -439,6 +445,7 @@ Runs automatically every 24 hours on startup — no cron needed. Memories below
 
 - **DuckDB** — default backend, zero setup, native vector similarity (same quality as pgvector)
 - **sentence-transformers** — local embeddings (`all-mpnet-base-v2`, 768 dims, no external service needed)
+- **spaCy 3.8.13+** — local NLP for deduplication and categorization (Python 3.11–3.14 compatible)
 - **APScheduler** — automatic 24h decay job
 - **MCP** — Claude integration via Model Context Protocol
 - **PostgreSQL + pgvector** — optional, for teams / large datasets

{yourmemory-1.2.1 → yourmemory-1.2.2}/README.md

@@ -51,6 +51,8 @@ Importance additionally modulates the decay rate within each category. Memories
 
 **Zero infrastructure required** — uses DuckDB out of the box. Two commands and you're done.
 
+Supports **Python 3.11, 3.12, 3.13, and 3.14**.
+
 ### 1. Install
 
 ```bash
@@ -91,9 +93,13 @@ Reload Claude Code (`Cmd+Shift+P` → `Developer: Reload Window`).
 
 #### Cline (VS Code)
 
-VS Code doesn't inherit your shell PATH, so use the **full path** from `yourmemory-path`.
+VS Code doesn't inherit your shell PATH. Run this in terminal to get the exact config to paste:
+
+```bash
+yourmemory-path
+```
 
-In Cline → **MCP Servers** → **Edit MCP Settings**:
+Then in Cline → **MCP Servers** → **Edit MCP Settings**, paste the output. It looks like:
 
 ```json
 {
@@ -110,7 +116,7 @@ In Cline → **MCP Servers** → **Edit MCP Settings**:
 }
 ```
 
-Run `yourmemory-path` in terminal — it prints the exact config to paste.
+Restart Cline after saving.
 
 #### Cursor
 
@@ -244,6 +250,7 @@ Runs automatically every 24 hours on startup — no cron needed. Memories below
 
 - **DuckDB** — default backend, zero setup, native vector similarity (same quality as pgvector)
 - **sentence-transformers** — local embeddings (`all-mpnet-base-v2`, 768 dims, no external service needed)
+- **spaCy 3.8.13+** — local NLP for deduplication and categorization (Python 3.11–3.14 compatible)
 - **APScheduler** — automatic 24h decay job
 - **MCP** — Claude integration via Model Context Protocol
 - **PostgreSQL + pgvector** — optional, for teams / large datasets

{yourmemory-1.2.1 → yourmemory-1.2.2}/memory_mcp.py

@@ -563,6 +563,40 @@ def print_path():
     print("Paste this into your Cline MCP settings:\n")
     print(_json.dumps(config, indent=2))
 
+def setup():
+    """Run once after pip install to download the spaCy model."""
+    import subprocess
+    print("YourMemory setup — installing spaCy language model...")
+    result = subprocess.run(
+        [sys.executable, "-m", "spacy", "download", "en_core_web_sm"],
+        check=False,
+    )
+    if result.returncode == 0:
+        print("✓ spaCy model installed successfully.")
+    else:
+        # Fallback: install via direct wheel URL
+        print("Direct download fallback...")
+        result2 = subprocess.run(
+            [sys.executable, "-m", "pip", "install",
+             "https://github.com/explosion/spacy-models/releases/download/"
+             "en_core_web_sm-3.8.0/en_core_web_sm-3.8.0-py3-none-any.whl",
+             "--break-system-packages"],
+            check=False,
+        )
+        if result2.returncode == 0:
+            print("✓ spaCy model installed successfully.")
+        else:
+            print("✗ Could not install spaCy model automatically.")
+            print("  Run manually: python -m spacy download en_core_web_sm")
+            print("  YourMemory will still work using the built-in regex fallback.")
+
+    # Also run DB migration
+    from src.db.migrate import migrate
+    migrate()
+    print("✓ Database initialised.")
+    print("\nSetup complete. Run yourmemory-path to get your MCP config.")
+
+
 def run():
     from src.db.migrate import migrate
     migrate()

{yourmemory-1.2.1 → yourmemory-1.2.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "yourmemory"
-version = "1.2.1"
+version = "1.2.2"
 description = "Persistent memory for Claude — Ebbinghaus forgetting curve, semantic deduplication, MCP-native"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -39,10 +39,11 @@ all = ["yourmemory[postgres,sse]"]
 [project.scripts]
 yourmemory = "memory_mcp:run"
 yourmemory-path = "memory_mcp:print_path"
+yourmemory-setup = "memory_mcp:setup"
 
 [project.urls]
-Homepage = "https://github.com/sachitrafa/cognitive-ai-memory"
-Repository = "https://github.com/sachitrafa/cognitive-ai-memory"
+Homepage = "https://github.com/sachitrafa/YourMemory"
+Repository = "https://github.com/sachitrafa/YourMemory"
 
 [tool.setuptools]
 py-modules = ["memory_mcp"]

yourmemory-1.2.2/src/services/extract.py

@@ -0,0 +1,52 @@
+import re
+import sys
+
+_QUESTION_WORDS = {"what", "who", "where", "when", "why", "how", "which", "whose", "whom"}
+
+_IMPERATIVE_PATTERNS = [
+    r'^(please|use|try|do|don\'t|make|create|add|remove|delete|update)',
+    r'^(convert|transform|change|modify|fix|help|show|tell)',
+    r'^(install|run|execute|start|stop|restart|configure)',
+]
+
+# Load spaCy if available — falls back to regex if model not installed yet
+# Run `yourmemory-setup` once after pip install to download the model
+_nlp = None
+try:
+    import spacy
+    _nlp = spacy.load("en_core_web_sm")
+except OSError:
+    print(
+        "YourMemory: spaCy model not found. Run `yourmemory-setup` once to install it.\n"
+        "  Falling back to built-in regex categorization.",
+        file=sys.stderr,
+    )
+except Exception:
+    pass
+
+
+def is_question(text: str) -> bool:
+    """Return True if the text is a question — questions are not stored as memories."""
+    stripped = text.strip()
+    if stripped.endswith("?"):
+        return True
+    first_word = re.split(r"\s+", stripped.lower())[0]
+    return first_word in _QUESTION_WORDS
+
+
+def categorize(text: str) -> str:
+    """
+    Classify text as fact or assumption.
+    Uses spaCy dependency parse when available, regex heuristics otherwise.
+    Run `yourmemory-setup` to enable spaCy.
+    """
+    if _nlp is not None:
+        doc = _nlp(text)
+        has_subject = any(tok.dep_ in ("nsubj", "nsubjpass") for tok in doc)
+        return "fact" if has_subject else "assumption"
+
+    text_lower = text.lower().strip()
+    for pattern in _IMPERATIVE_PATTERNS:
+        if re.match(pattern, text_lower):
+            return "assumption"
+    return "fact"

yourmemory-1.2.2/src/services/resolve_fallback.py

@@ -0,0 +1,192 @@
+"""
+Semantic deduplication for POST /memories - Fallback version without spaCy.
+
+Detects near-duplicate memories via cosine similarity and applies one of:
+  - reinforce : sim ≥ 0.85  — paraphrase, bump recall_count only
+  - replace   : 0.65–0.85 + contradiction detected — overwrite with incoming
+  - merge     : 0.65–0.85 + no contradiction — entity-append to existing
+  - new       : sim < 0.65  — genuinely distinct, plain INSERT
+"""
+
+import json
+import math
+import re
+from src.db.connection import get_backend
+
+DEDUP_THRESHOLD     = 0.65   # below → always new memory
+REINFORCE_THRESHOLD = 0.85   # at or above → reinforce (near-identical paraphrase)
+
+# Simple contradiction detection patterns (fallback)
+_CONTRADICTION_PATTERNS = [
+    (r'\b(love|like|prefer|enjoy)\b', r'\b(hate|dislike|avoid)\b'),
+    (r'\b(start|begin|use)\b', r'\b(stop|quit|avoid)\b'),
+    (r'\b(want|need)\b', r'\b(refuse|reject)\b'),
+    (r'\b(good|great|excellent)\b', r'\b(bad|terrible|awful)\b'),
+    (r'\b(yes|true|correct)\b', r'\b(no|false|wrong)\b'),
+]
+
+
+def _cosine(a: list, b: list) -> float:
+    import numpy as np
+    va, vb = np.array(a, dtype=float), np.array(b, dtype=float)
+    denom = np.linalg.norm(va) * np.linalg.norm(vb)
+    return float(np.dot(va, vb) / denom) if denom else 0.0
+
+
+def find_near_duplicate(user_id: str, embedding: list, conn) -> dict | None:
+    """
+    Return the closest existing memory if cosine similarity >= DEDUP_THRESHOLD,
+    else None. Uses the caller's open connection.
+    """
+    backend = get_backend()
+
+    if backend == "postgres":
+        embedding_str = f"[{','.join(str(x) for x in embedding)}]"
+        cur = conn.cursor()
+        cur.execute("""
+            SELECT id, content, category, importance, recall_count,
+                   1 - (embedding <=> %s::vector) AS similarity
+            FROM memories
+            WHERE user_id = %s
+            ORDER BY embedding <=> %s::vector
+            LIMIT 1
+        """, (embedding_str, user_id, embedding_str))
+        row = cur.fetchone()
+        cur.close()
+        if row is None:
+            return None
+        sim = row[5]
+        if sim < DEDUP_THRESHOLD:
+            return None
+        return {"id": row[0], "content": row[1], "category": row[2],
+                "importance": row[3], "recall_count": row[4], "similarity": sim}
+
+    if backend == "duckdb":
+        from src.db.connection import duckdb_row
+        cur = conn.execute("""
+            SELECT id, content, category, importance, recall_count,
+                   array_cosine_similarity(embedding, ?::FLOAT[768]) AS similarity
+            FROM memories
+            WHERE user_id = ?
+            ORDER BY similarity DESC
+            LIMIT 1
+        """, [embedding, user_id])
+        row = duckdb_row(cur)
+        if row is None or row["similarity"] < DEDUP_THRESHOLD:
+            return None
+        return row
+
+    # SQLite: numpy cosine over all user memories
+    cur = conn.cursor()
+    cur.execute("""
+        SELECT id, content, category, importance, recall_count, embedding
+        FROM memories WHERE user_id = ?
+    """, (user_id,))
+    rows = cur.fetchall()
+    cur.close()
+
+    best, sim = None, -1.0
+    for row in rows:
+        raw = row[5] if isinstance(row, tuple) else row["embedding"]
+        if raw is None:
+            continue
+        s = _cosine(embedding, json.loads(raw))
+        if s > sim:
+            sim, best = s, row
+    if best is None or sim < DEDUP_THRESHOLD:
+        return None
+    return {"id": best[0], "content": best[1], "category": best[2],
+            "importance": best[3], "recall_count": best[4], "similarity": sim}
+
+
+def detect_contradiction(existing_text: str, incoming_text: str) -> bool:
+    """
+    Fallback contradiction detection using regex patterns.
+    Return True if the incoming text contradicts the existing one.
+    """
+    existing_lower = existing_text.lower()
+    incoming_lower = incoming_text.lower()
+    
+    for positive_pattern, negative_pattern in _CONTRADICTION_PATTERNS:
+        # Check if existing has positive and incoming has negative
+        if re.search(positive_pattern, existing_lower) and re.search(negative_pattern, incoming_lower):
+            return True
+        # Check if existing has negative and incoming has positive  
+        if re.search(negative_pattern, existing_lower) and re.search(positive_pattern, incoming_lower):
+            return True
+    
+    return False
+
+
+def merge_entities(existing_text: str, incoming_text: str) -> str:
+    """
+    Fallback entity merging using simple heuristics.
+    Append capitalized words and quoted strings from incoming that are absent from existing.
+    Returns the merged string, or existing_text unchanged if nothing new found.
+    """
+    existing_lower = existing_text.lower()
+    
+    # Extract potential entities using simple patterns
+    candidates = []
+    
+    # Capitalized words (potential proper nouns)
+    capitalized_words = re.findall(r'\b[A-Z][a-zA-Z]{2,}\b', incoming_text)
+    candidates.extend(capitalized_words)
+    
+    # Quoted strings
+    quoted_strings = re.findall(r'"([^"]+)"', incoming_text)
+    quoted_strings.extend(re.findall(r"'([^']+)'", incoming_text))
+    candidates.extend(quoted_strings)
+    
+    # Technical terms (words with numbers, dots, underscores)
+    tech_terms = re.findall(r'\b[a-zA-Z][a-zA-Z0-9._-]*[a-zA-Z0-9]\b', incoming_text)
+    candidates.extend([t for t in tech_terms if '.' in t or '_' in t or any(c.isdigit() for c in t)])
+
+    # Filter out terms already present in existing text
+    new_terms = [t for t in candidates if t.lower() not in existing_lower and len(t.strip()) > 2]
+
+    # Deduplicate while preserving order
+    seen, deduped = set(), []
+    for t in new_terms:
+        if t.lower() not in seen:
+            seen.add(t.lower())
+            deduped.append(t)
+
+    if not deduped:
+        return existing_text
+    if len(deduped) == 1:
+        return f"{existing_text} with {deduped[0]}"
+    return f"{existing_text} with {', '.join(deduped[:-1])} and {deduped[-1]}"
+
+
+def resolve(user_id: str, content: str, embedding: list, conn) -> dict:
+    """
+    Facade: decide what to do with an incoming memory.
+
+    Returns:
+        {
+          "action":   "new" | "reinforce" | "replace" | "merge",
+          "content":  str,          # final content to store/update
+          "existing": dict | None,  # matched row if any
+        }
+    """
+    match = find_near_duplicate(user_id, embedding, conn)
+
+    if match is None:
+        return {"action": "new", "content": content, "existing": None}
+
+    sim = match["similarity"]
+
+    if sim >= REINFORCE_THRESHOLD:
+        return {"action": "reinforce", "content": match["content"], "existing": match}
+
+    # DEDUP_THRESHOLD ≤ sim < REINFORCE_THRESHOLD
+    if detect_contradiction(match["content"], content):
+        return {"action": "replace", "content": content, "existing": match}
+
+    merged = merge_entities(match["content"], content)
+    if merged == match["content"]:
+        # No new entities found — treat as paraphrase
+        return {"action": "reinforce", "content": match["content"], "existing": match}
+
+    return {"action": "merge", "content": merged, "existing": match}

{yourmemory-1.2.1 → yourmemory-1.2.2/yourmemory.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: yourmemory
-Version: 1.2.1
+Version: 1.2.2
 Summary: Persistent memory for Claude — Ebbinghaus forgetting curve, semantic deduplication, MCP-native
 Author-email: Sachit Misra <mishrasachit1@gmail.com>
 License:                                  Apache License
@@ -163,8 +163,8 @@ License:                                  Apache License
            See the License for the specific language governing permissions and
            limitations under the License.
         
-Project-URL: Homepage, https://github.com/sachitrafa/cognitive-ai-memory
-Project-URL: Repository, https://github.com/sachitrafa/cognitive-ai-memory
+Project-URL: Homepage, https://github.com/sachitrafa/YourMemory
+Project-URL: Repository, https://github.com/sachitrafa/YourMemory
 Keywords: mcp,claude,memory,ebbinghaus,ai,sqlite,postgresql
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.11
@@ -246,6 +246,8 @@ Importance additionally modulates the decay rate within each category. Memories
 
 **Zero infrastructure required** — uses DuckDB out of the box. Two commands and you're done.
 
+Supports **Python 3.11, 3.12, 3.13, and 3.14**.
+
 ### 1. Install
 
 ```bash
@@ -286,9 +288,13 @@ Reload Claude Code (`Cmd+Shift+P` → `Developer: Reload Window`).
 
 #### Cline (VS Code)
 
-VS Code doesn't inherit your shell PATH, so use the **full path** from `yourmemory-path`.
+VS Code doesn't inherit your shell PATH. Run this in terminal to get the exact config to paste:
+
+```bash
+yourmemory-path
+```
 
-In Cline → **MCP Servers** → **Edit MCP Settings**:
+Then in Cline → **MCP Servers** → **Edit MCP Settings**, paste the output. It looks like:
 
 ```json
 {
@@ -305,7 +311,7 @@ In Cline → **MCP Servers** → **Edit MCP Settings**:
 }
 ```
 
-Run `yourmemory-path` in terminal — it prints the exact config to paste.
+Restart Cline after saving.
 
 #### Cursor
 
@@ -439,6 +445,7 @@ Runs automatically every 24 hours on startup — no cron needed. Memories below
 
 - **DuckDB** — default backend, zero setup, native vector similarity (same quality as pgvector)
 - **sentence-transformers** — local embeddings (`all-mpnet-base-v2`, 768 dims, no external service needed)
+- **spaCy 3.8.13+** — local NLP for deduplication and categorization (Python 3.11–3.14 compatible)
 - **APScheduler** — automatic 24h decay job
 - **MCP** — Claude integration via Model Context Protocol
 - **PostgreSQL + pgvector** — optional, for teams / large datasets

{yourmemory-1.2.1 → yourmemory-1.2.2}/yourmemory.egg-info/SOURCES.txt

@@ -22,6 +22,7 @@ src/services/embed.py
 src/services/extract.py
 src/services/extract_fallback.py
 src/services/resolve.py
+src/services/resolve_fallback.py
 src/services/retrieve.py
 yourmemory.egg-info/PKG-INFO
 yourmemory.egg-info/SOURCES.txt

{yourmemory-1.2.1 → yourmemory-1.2.2}/yourmemory.egg-info/entry_points.txt

@@ -1,3 +1,4 @@
 [console_scripts]
 yourmemory = memory_mcp:run
 yourmemory-path = memory_mcp:print_path
+yourmemory-setup = memory_mcp:setup

yourmemory-1.2.1/src/services/extract.py

@@ -1,36 +0,0 @@
-import re
-import spacy
-
-try:
-    _nlp = spacy.load("en_core_web_sm")
-except OSError:
-    import subprocess, sys
-    subprocess.run(
-        [sys.executable, "-m", "pip", "install",
-         "https://github.com/explosion/spacy-models/releases/download/"
-         "en_core_web_sm-3.8.0/en_core_web_sm-3.8.0-py3-none-any.whl"],
-        check=True,
-    )
-    _nlp = spacy.load("en_core_web_sm")
-
-_QUESTION_WORDS = {"what", "who", "where", "when", "why", "how", "which", "whose", "whom"}
-
-
-def is_question(text: str) -> bool:
-    """Return True if the text is a question — questions are not stored as memories."""
-    stripped = text.strip()
-    if stripped.endswith("?"):
-        return True
-    first_word = re.split(r"\s+", stripped.lower())[0]
-    return first_word in _QUESTION_WORDS
-
-
-def categorize(text: str) -> str:
-    """
-    Use spaCy dependency parse to classify:
-      fact       — declarative sentence with an explicit subject
-      assumption — imperative sentence with no subject (command/instruction)
-    """
-    doc = _nlp(text)
-    has_subject = any(tok.dep_ in ("nsubj", "nsubjpass") for tok in doc)
-    return "fact" if has_subject else "assumption"