All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Update Anthropic Sonnet version to v2
- Require collection id when deleting a document from search index
- Update Google Gemini models
- Added openai model costs
- Record model id from API response
- Added gpt-4o-mini model
- Fix notion execution
- Added model costs for openai.gpt-4o and anthropic.claude-3-5-sonnet-20240620-v1:0
- Record model usage with model id from response
- Updated system prompt
- Replace PostgresDB with MongoDB
- Implement tool calls for better retrieval of data
- Updated multiple protobuf schemas
- Added support for Claude 3.5 Sonnet
- Improved source quoting
- Fix a bug that prevented the insertion of embedding vectors into the database
- Fix a bug that prevented to get the current notion api key
- Added GPT-4o
- Trim suffix from notion filenames
- Fix a problem api key caching
- Added notion api key management
- Update system prompt
- Added notion service
- Removed Anthropic package
- Fixed Gemini model costs
- Added gpt-4-turbo model
- Update Gemini Pro 1.5 model pricing
- Fix model selection for Gemini Models
- Update model Gemini prices
- Added Gemini Pro 1.5
- Added chat batch processing
- Removed support for Mistral AI
- Sort models by name
- Added support for Claude 3 Haiku
- Removed support Pinecone
- Fixed user tracking for Claude Opus
- Added pricing for Claude 3 Opus
- Add support for Claude 3 Opus
- Fix alternating roles bug for Claude 3
- Add cost support for Claude 3
- Added support for Claude 3
- Added Mistral AI models
- Added crashlytics service
- Fixed balance sheet
- Remove case sensitivity from document search
- Fixed document search bug
- Add support for website indexing
- Added support for references in chat messages
- Changed protobuf schemas
- Enable support for latex in text generation
- Added support for more bedrock models
- Changed LLM model names
- Added support for AWS Bedrock
- Pool embedding usage tracking to reduce database entries
- Changed collection proto specs
- Added GetCollection function
- Fix cost calculation
- Prevent cloud deployment of insecure auth service
- Fix qdrant document deletion
- Add support for follow-up questions
- Retry embedding generation on error
- Fix bug that prevented insertions of documents in the database
- Add support for Qdrant Vector Search
- Switched to OpenAI's text-embedding-3-large embedding model
- Implement DeleteChatMessage function
- Fix page numbers bug
- Rename package to github.com/pzierahn/chatbot_services
- Fix none existing local credentials bug
- Support for Google Gemini Pro
- Changed protobuf schema
- Use Google Cloud instead of Supabase
- Fix openai api issue
- Added GTP-4 Turbo model
- Fix Markdown code block issue with text generation
- Added founding check for users