Changelog

All notable changes to Libby D. Bot will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

0.9.0 - 2026-03-06

Added

• Integrated Ollama server directly in Docker container
• Pre-installed mxbai-embed-large embedding model in Docker image
• Added docker/start.sh startup script for automatic Ollama initialization
• Added ollama-models volume for persisting downloaded models
• Added zstd package for Ollama installation support

Changed

• Updated Dockerfile to include Ollama server installation
• Updated docker-compose.yml to use internal Ollama instead of host connection
• Increased health check start period to 60s for Ollama initialization
• Updated documentation with new Docker deployment instructions
• Improved Docker build process with model pre-loading

Fixed

• Fixed Docker connectivity issues by embedding Ollama server in container
• Fixed port conflicts by not exposing Ollama port externally (11434)

0.8.0 - 2026-03-03

Added

• REST API server with FastAPI for programmatic access
• Docker support with Dockerfile and docker-compose configuration
• Re-embedding functionality to update embeddings with new models
• libby-server CLI command to start the API server
• API endpoints for text embedding, file upload, document retrieval, and health checks
• .env.example file for environment configuration

Changed

• Bumped version from 0.7.0 to 0.8.0
• Made server more robust with better error handling
• Updated documentation with API usage examples
• Improved configuration management

Fixed

• Fixed various bugs in the API server
• Fixed server stability issues

0.7.0 - 2026-02-13

Added

• RAG (Retrieval Augmented Generation) capabilities
• Document embedder class for managing embeddings
• Library search tool for agents
• Support for DuckDB with HNSW indexes
• New database files for embedding and memory storage
• PostgreSQL backend tests

Changed

• Refined embedding logic and improved tests
• Updated dependencies
• Locked project dependencies using uv

Fixed

• Corrected DuckDB table creation (replaced SERIAL with nextval)
• Fixed DuckDB connection initialization
• Fixed connection handling to properly use self.engine
• Corrected AUTOINCREMENT syntax for DuckDB
• Improved DuckDB interaction using direct SQL queries

Removed

• Removed EmbeddingDuckdb class, simplified table creation

0.6.0 - 2025-12-08

Added

• Full DuckDB support with HNSW vector indexes
• Hybrid search combining vector and full-text search (FTS)
• FTS index creation for DuckDB

Changed

• Migrated from SQLAlchemy to direct SQL for DuckDB
• Improved database connection management
• Updated code structure and fixed multiple bugs

Fixed

• Fixed many bugs related to database connections
• Corrected table creation and query execution
• Fixed SQLite fallback in DocEmbedder

0.5.0 - 2025-08-17

Added

• SQLite support with sqlite-vec extension for embeddings
• Method to retrieve embedded documents (get_embedded_documents)
• Connection property to manage database connections
• Multi-thread mode for SQLite

Changed

• Migrated embedding storage from SQLAlchemy to direct SQLite
• Improved database configuration and embedding handling
• Changed default embedding model to Gemini
• Refined tests and improved test coverage

Fixed

• Fixed SQLite3 database closing issues
• Corrected table creation and queries for SQLite with vec0
• Fixed fallback behavior for SQLite in DocEmbedder

0.4.0 - 2025-08-08

Added

• Gemini embedding model support
• Google as default AI provider
• Support for multiple embedding models (Gemini, Ollama)

Changed

• Updated settings configuration to be more robust
• Improved provider handling and configuration
• Bumped version from 0.3.x to 0.4.0

Fixed

• Fixed bugs in embedding generation
• Corrected settings loading

0.3.11 - 2025-06-25

Fixed

• Adapted ask function to remove <think\> content from reasoning models' responses

0.3.10 - 2025-05-10

Changed

• Updated dependencies
• Updated documentation

0.3.9 - 2025-03-02

Added

• API reference documentation
• Comprehensive tutorial documentation

Changed

• Migrated configuration from YAML to Pydantic settings
• Updated documentation structure
• Removed PyYAML dependency

Removed

• Removed YAML configuration files

0.3.8 - 2025-01-04

Added

• Method to embed a path (directory of documents)
• Initial support for SQLite and DuckDB as backends for DocEmbedder
• Flexible database URL support

Changed

• Updated package dependencies
• Improved database URL handling

0.3.7 - 2024-11-20

Added

• Qwen model support
• Support for reading prompts from input text files
• Context support for the generate command
• Option to specify LLM model on CLI
• Model selection from config with validation

Changed

• Increased number of retrieved documents from 15 to 100
• Updated README with comprehensive usage instructions
• Bumped version to 0.3.x series

Fixed

• Fixed bug in calling generate with a prompt file

0.3.0 - 2024-07-31

Added

• Ingest module with PDF processing capabilities
• Tests for CLI functionality
• Support for collections of embeddings
• Filter by collection when retrieving documents

Changed

• Improved documentation
• Refactored CLI for better testability
• Bumped version from 0.2.x to 0.3.0

0.2.0 - 2024-07-05

Added

• Memory functionality for Libby
• Chat history storage and retrieval
• Support for multiple chat sessions

Changed

• Improved memory implementation
• Enhanced conversation context handling

0.1.0 - 2024-04-09

Added

• Initial release of Libby D. Bot
• Basic CLI interface with Fire
• Support for multiple AI models (Llama3, Gemma, ChatGPT)
• PDF document processing and embedding
• Question answering with context
• Content generation capabilities
• Poetry configuration and dependencies
• Basic project structure

Features

• Multiple language support (English and Portuguese)
• Various AI models available
• PDF document processing
• Question answering with document context
• Content generation