mem-llm 1.3.1__tar.gz → 1.3.2__tar.gz

{mem_llm-1.3.1 → mem_llm-1.3.2}/CHANGELOG.md

@@ -5,6 +5,68 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.3.2] - 2025-11-02
+
+### 🎉 Major Features
+
+- 📊 **Response Metrics & Quality Analytics** (v1.3.1+)
+  - `ChatResponse` dataclass: Comprehensive response tracking
+  - `ResponseMetricsAnalyzer`: Aggregate analytics and monitoring
+  - Confidence scoring: Based on KB usage, memory, temperature, and length
+  - Real-time latency tracking: Monitor response performance
+  - Quality labels: High/Medium/Low classification
+  - Export metrics: JSON and summary formats for dashboards
+  - Production monitoring: Health checks and SLA tracking
+
+- 🔍 **Vector Search & Semantic Knowledge Base** (v1.3.2+)
+  - ChromaDB integration: Semantic search with embeddings
+  - Sentence-transformers support: `all-MiniLM-L6-v2` default model
+  - Cross-lingual search: Understands meaning across languages
+  - Hybrid search: Vector + keyword search combination
+  - Better relevancy: Semantic understanding vs keyword matching
+  - Optional feature: Install with `pip install chromadb sentence-transformers`
+
+### 🆕 New Components
+
+- `response_metrics.py`: `ChatResponse`, `ResponseMetricsAnalyzer`, `calculate_confidence`
+- `vector_store.py`: `VectorStore`, `ChromaVectorStore`, `create_vector_store`
+- Enhanced `SQLMemoryManager`: Vector search integration
+- Enhanced `MemAgent`: Response metrics and vector search support
+
+### 🔄 Enhanced Features
+
+- **MemAgent.chat()**: New `return_metrics` parameter for detailed response analysis
+- **Memory Metadata**: Automatic saving of response metrics in conversations
+- **User Profile**: Improved preferences and summary extraction/parsing
+- **Knowledge Base Search**: Optional vector search with `use_vector_search=True`
+- **ChromaDB Sync**: `sync_all_kb_to_vector_store()` method for existing KB entries
+
+### 📚 New Examples
+
+- `15_response_metrics.py`: Response quality metrics and analytics
+- `16_vector_search.py`: Semantic/vector search demonstration
+
+### 🐛 Bug Fixes
+
+- Fixed metadata not being saved in conversation history
+- Fixed preferences parsing from JSON string to dict
+- Fixed summary generation for existing users
+- Fixed `get_user_profile()` SQL/JSON memory detection logic
+- Fixed ChromaDB embedding function compatibility
+
+### 📝 Documentation
+
+- Updated all examples: Simplified and more readable
+- Enhanced README with new features
+- Vector search usage guide
+
+### ⚡ Improved
+
+- Better error handling for ChromaDB initialization
+- Fallback mechanism for embedding function loading
+- Enhanced similarity score calculation for vector search
+- Improved conversation metadata tracking
+
 ## [1.3.1] - 2025-10-31
 
 ### 📝 Documentation

{mem_llm-1.3.1/mem_llm.egg-info → mem_llm-1.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: mem-llm
-Version: 1.3.1
+Version: 1.3.2
 Summary: Memory-enabled AI assistant with multi-backend LLM support (Ollama, LM Studio, Gemini) - Local and cloud ready
 Author-email: "C. Emre Karataş" <karatasqemre@gmail.com>
 License: MIT
@@ -70,7 +70,14 @@ Mem-LLM is a powerful Python library that brings persistent memory capabilities
 - **Issues**: https://github.com/emredeveloper/Mem-LLM/issues
 - **Documentation**: See examples/ directory
 
-## 🆕 What's New in v1.3.0
+## 🆕 What's New in v1.3.2
+
+- 📊 **Response Metrics** (v1.3.1+) – Track confidence, latency, KB usage, and quality analytics
+- 🔍 **Vector Search** (v1.3.2+) – Semantic search with ChromaDB, cross-lingual support
+- 🎯 **Quality Monitoring** – Production-ready metrics for response quality
+- 🌐 **Semantic Understanding** – Understands meaning, not just keywords
+
+## What's New in v1.3.0
 
 - 🔌 **Multi-Backend Support**: Choose between Ollama (local), LM Studio (local), or Google Gemini (cloud)
 - 🏗️ **Factory Pattern**: Clean, extensible architecture for easy backend switching
@@ -79,12 +86,14 @@ Mem-LLM is a powerful Python library that brings persistent memory capabilities
 - 📚 **New Examples**: 4 additional examples showing multi-backend usage
 - 🎯 **Backward Compatible**: All v1.2.0 code still works without changes
 
-[See full changelog](CHANGELOG.md#130---2025-10-31)
+[See full changelog](CHANGELOG.md)
 
 ## ✨ Key Features
 
 - 🔌 **Multi-Backend Support** (v1.3.0+) - Choose Ollama, LM Studio, or Gemini with unified API
 - 🔍 **Auto-Detection** (v1.3.0+) - Automatically find and use available LLM services
+- 📊 **Response Metrics** (v1.3.1+) - Track confidence, latency, KB usage, and quality analytics
+- 🔍 **Vector Search** (v1.3.2+) - Semantic search with ChromaDB, cross-lingual support
 - 🧠 **Persistent Memory** - Remembers conversations across sessions
 - 🤖 **Universal Model Support** - Works with 100+ Ollama models, LM Studio models, and Gemini
 - 💾 **Dual Storage Modes** - JSON (simple) or SQLite (advanced) memory backends

{mem_llm-1.3.1 → mem_llm-1.3.2}/README.md

@@ -15,7 +15,14 @@ Mem-LLM is a powerful Python library that brings persistent memory capabilities
 - **Issues**: https://github.com/emredeveloper/Mem-LLM/issues
 - **Documentation**: See examples/ directory
 
-## 🆕 What's New in v1.3.0
+## 🆕 What's New in v1.3.2
+
+- 📊 **Response Metrics** (v1.3.1+) – Track confidence, latency, KB usage, and quality analytics
+- 🔍 **Vector Search** (v1.3.2+) – Semantic search with ChromaDB, cross-lingual support
+- 🎯 **Quality Monitoring** – Production-ready metrics for response quality
+- 🌐 **Semantic Understanding** – Understands meaning, not just keywords
+
+## What's New in v1.3.0
 
 - 🔌 **Multi-Backend Support**: Choose between Ollama (local), LM Studio (local), or Google Gemini (cloud)
 - 🏗️ **Factory Pattern**: Clean, extensible architecture for easy backend switching
@@ -24,12 +31,14 @@ Mem-LLM is a powerful Python library that brings persistent memory capabilities
 - 📚 **New Examples**: 4 additional examples showing multi-backend usage
 - 🎯 **Backward Compatible**: All v1.2.0 code still works without changes
 
-[See full changelog](CHANGELOG.md#130---2025-10-31)
+[See full changelog](CHANGELOG.md)
 
 ## ✨ Key Features
 
 - 🔌 **Multi-Backend Support** (v1.3.0+) - Choose Ollama, LM Studio, or Gemini with unified API
 - 🔍 **Auto-Detection** (v1.3.0+) - Automatically find and use available LLM services
+- 📊 **Response Metrics** (v1.3.1+) - Track confidence, latency, KB usage, and quality analytics
+- 🔍 **Vector Search** (v1.3.2+) - Semantic search with ChromaDB, cross-lingual support
 - 🧠 **Persistent Memory** - Remembers conversations across sessions
 - 🤖 **Universal Model Support** - Works with 100+ Ollama models, LM Studio models, and Gemini
 - 💾 **Dual Storage Modes** - JSON (simple) or SQLite (advanced) memory backends

{mem_llm-1.3.1 → mem_llm-1.3.2}/mem_llm/__init__.py

@@ -63,7 +63,14 @@ try:
 except ImportError:
     __all_export_import__ = []
 
-__version__ = "1.3.1"
+# Response Metrics (v1.3.1+)
+try:
+    from .response_metrics import ChatResponse, ResponseMetricsAnalyzer, calculate_confidence
+    __all_metrics__ = ["ChatResponse", "ResponseMetricsAnalyzer", "calculate_confidence"]
+except ImportError:
+    __all_metrics__ = []
+
+__version__ = "1.3.2"
 __author__ = "C. Emre Karataş"
 
 # Multi-backend LLM support (v1.3.0+)
@@ -80,4 +87,4 @@ __all__ = [
     "MemAgent",
     "MemoryManager", 
     "OllamaClient",
-] + __all_llm_backends__ + __all_tools__ + __all_pro__ + __all_cli__ + __all_security__ + __all_enhanced__ + __all_summarizer__ + __all_export_import__
+] + __all_llm_backends__ + __all_tools__ + __all_pro__ + __all_cli__ + __all_security__ + __all_enhanced__ + __all_summarizer__ + __all_export_import__ + __all_metrics__

{mem_llm-1.3.1 → mem_llm-1.3.2}/mem_llm/config_manager.py

@@ -62,7 +62,9 @@ class ConfigManager:
                 "default_kb": "ecommerce",
                 "custom_kb_file": None,
                 "search_limit": 5,
-                "min_relevance_score": 0.3
+                "min_relevance_score": 0.3,
+                "enable_vector_search": False,  # v1.3.2+ - Optional semantic search
+                "embedding_model": "all-MiniLM-L6-v2"  # Sentence transformers model
             },
             "response": {
                 "use_knowledge_base": True,