Skip to main content

SAMO AI Assistant - Release Notes v1.1.3

Release Date: November 2025
Version: 1.1.3

🎉 What's New in This Release​

Multi-Model Agent Architecture​

  • Introduced separate configurable models for different agent tasks (ReAct, Intent, Synthesis)
  • Enhanced flexibility for performance optimization and cost management

Performance Optimizations​

  • Metadata Caching: Faster metadata retrieval for frequently accessed feature types
  • Connection Pooling: Performance improvement for database queries
  • Configurable Cache Size: Tune cache size based on workload (default: 16 entries)

🚀 Current Functionalities​

Multi-Intent AI Agent Architecture​

Intent Analysis

  • Automatically classifies queries into three categories:
    • Data Query: Natural language questions requiring database access
    • Documentation: Questions about uploaded documents using RAG
    • Conversation: General assistant interactions and follow-up questions
  • Context-aware processing determines if queries reference previous conversation history
  • Configurable dedicated LLM model for intent analysis

ReAct Agent Framework

  • Iterative reasoning and action loops for complex queries
  • Tool integration: fuzzy search of featuretypes, metadata retrieval, database queries, document search
  • Real-time streaming of reasoning process
  • Configurable model with adjustable parameters

Database Query Capabilities​

Natural Language to SQL Translation

  • Multi-vendor support: Oracle, PostgreSQL
  • Two-stage query process:
    1. Feature Type Retrieval Agent identifies relevant database entities
    2. SQL Generation Agent constructs and executes queries
  • Intelligent caching and connection pooling for optimal performance

Document Intelligence & RAG​

Document Processing

  • Supported formats: PDF and DOCX
  • Azure Document Intelligence for advanced text extraction
  • Semantic chunking for optimal retrieval
  • OpenAI text-embedding-3-small for vector embeddings

ChromaDB Vector Store

  • Local persistent storage or remote ChromaDB server
  • Collection management: create, delete, list
  • Per-collection file management with metadata

RAG-Based Question Answering

  • Dedicated DocumentationAgent using ReAct framework
  • Context-aware retrieval using conversation history

Conversation Management​

Session-Based History

  • Persistent conversation storage
  • Last 5 interactions included for context
  • History-aware responses understand references to previous questions
  • Unique session IDs for multi-turn conversations

Conversational AI

  • Natural language understanding for greetings, clarifications, follow-ups
  • Dedicated synthesis model for formatting answers
  • Configurable temperature for response creativity

Real-Time Streaming Responses​

Server-Sent Events (SSE)

  • Token-by-token streaming for immediate feedback
  • Progress indicators for workflow stages:
    • Intent analysis
    • Feature type retrieval
    • SQL generation
    • Documentation search
    • Response synthesis
  • Transparent reasoning display
  • Metadata: SQL queries, execution times, reasoning paths

Endpoints

  • /ai/api/collections/<name>/question/stream - SSE streaming
  • /ai/api/collections/<name>/question - Standard JSON response

Web-Based Admin Console​

Interactive Chat Interface

  • Collection selection dropdown
  • Streaming toggle for real-time responses
  • Markdown rendering with:
    • Syntax highlighting for code
    • Tables and lists
    • Headers and formatting
  • Collapsible reasoning sections
  • Conversation reset functionality

Security & Authentication​

Dual Authentication Methods

  1. Basic Authentication: For development/testing
  2. JWT Token Authentication: For production with external User Service validation

Authorization

  • Protected endpoints for file operations
  • Flexible authentication method selection
  • Request header validation

REST API​

Collections Management

  • GET /ai/api/collections - List all collections
  • GET /ai/api/collections/<name> - Get collection details
  • PUT /ai/api/collections/<name>/create - Create collection
  • DELETE /ai/api/collections/<name>/delete - Delete collection

File Management

  • PUT /ai/api/collections/<name>/files/<filename>/upload - Upload file
  • DELETE /ai/api/collections/<name>/files/<filename>/delete - Delete file

Query Endpoints

  • POST /ai/api/collections/<name>/question - Standard query
  • POST /ai/api/collections/<name>/question/stream - Streaming query

⚠️ Known Limitations​

  1. Reasoning Model Constraints: OpenAI reasoning models (o1, o3, gpt-5) require:

    • Temperature = 1.0 (cannot be adjusted)
    • max_tokens >= 16000
    • Application validates these constraints at startup

  2. ChromaDB Configuration: Cannot use both local and remote ChromaDB simultaneously

  3. File Format Support: Currently limited to PDF and DOCX files

👥 Contributors​

Development1 Team and Development3 Team - SAMO AI Assistant Project - SAMO AI Assistant Project

Last Updated: November 2025