Skip to content

CHANGELOG.md

Latest commit

 

History

History
148 lines (118 loc) · 4.42 KB

CHANGELOG.md

File metadata and controls

148 lines (118 loc) · 4.42 KB

v0.4.0 (2025-10-27)

Changed

  • BREAKING: Upgraded Chainlit from 1.3.1 to 2.8.3
  • BREAKING: Configuration file structure changed (removed deprecated theme and audio sections)
  • BREAKING: Python 3.8 support dropped (minimum Python 3.10)
  • Upgraded LangGraph from 0.2.45 to 1.0.1 (backward compatible)
  • Updated LangChain packages to latest compatible versions (langchain-openai 0.3.35, langchain-anthropic 0.3.0)
  • Upgraded Docker base image from Python 3.10 to 3.11

Added

  • Implemented missing BaseStorageClient methods for Chainlit 2.x compatibility
    • get_read_url() for presigned URL generation
    • delete_file() for file deletion from MinIO
    • close() for cleanup operations
  • Comprehensive migration documentation (MIGRATION_PLAN.md, UPGRADE_SUMMARY.md, UPGRADE_CHECKLIST.md)
  • Better security with secure cookie-based authentication
  • Improved UI with Shadcn/Tailwind rewrite from Chainlit 2.x
  • Access to LangGraph 1.0 features (node caching, deferred nodes, commands)

Fixed

  • Updated import path for BaseStorageClient from chainlit.data.base to chainlit.data.storage_clients.base
  • Removed deprecated chainlit.telemetry.trace decorator from auth.py
  • Fixed CVE-2024-47874 via FastAPI/Starlette security updates
  • Fixed UI text display issue by updating translation files to Chainlit 2.8.3 format
    • Updated en-US.json to new flat structure (common, auth, chat hierarchy)
    • Removed 18 legacy translation files that used old 1.x nested structure
    • UI now properly displays buttons, labels, and dialog text

Migration

  • See MIGRATION_PLAN.md for detailed upgrade instructions
  • Configuration cleanup required: remove [features.audio] and [UI.theme] sections from .chainlit/config.toml
  • Update [meta] generated_by to "2.8.3"
  • Database backup recommended before upgrading

v0.3.1 (2024-11-14)

Feat

  • Google model provider

v0.3.0 (2024-11-05)

Feat

  • image capability
  • model capability

Fix

  • wrong capability of haiku
  • missing llm_factory

v0.2.0 (2024-11-05)

Feat

  • groq support
  • grok support

Refactor

  • model display naming
  • llm factory pattern

v0.1.1 (2024-10-25)

Feat

  • dynamic parameter parsing for Ollama
  • parsing model specific params from ollama api response
  • update claude 3.5 sonnet
  • add starters for all workflows
  • add password auth to simplify the setup
  • simple resume chat
  • docker compose support
  • compose and dockerfile
  • chat persistence
  • google oauth support
  • update claude model list
  • get available model automatically
  • auto module discovery
  • extra header to enable the prompt caching
  • new resume module
  • Tavily search engine integration
  • tool selection node
  • search before answer
  • llm module to support multiple llm api vendors
  • simple chat demo

Fix

  • issues when parsing list of str
  • minor issue
  • remove the extra operation of creating session in the on_chat_start
  • the messages serialize/deserialize issue
  • final state is incorrectly updated
  • wrong .chainlit folder issue
  • error when OAuth provider is not given
  • error when TAVILY_API_KEY is not given in .env
  • remove redundant function def
  • update .env example
  • exclude .files created by chainlit app cache
  • minio invalid bucket issue
  • resume chat settings correctly
  • error handling when openai key is not set
  • temporary fix for randomly missing thread id in the data store.
  • correctly resume chat_profile
  • save/load graph states to db
  • app not reloading when making changes
  • duplicate import
  • setting simple chat as default workflow
  • host machine ollama api access
  • remove the fixed cache request header
  • setting menu not updating when switching to other workflows
  • duplicate tool name issue
  • llama tool calling issue
  • langsmith setting in .env
  • missing example in .env
  • noisy console print
  • conversation history issues

Refactor

  • llm module
  • change ollama model naming rule to prioritize it
  • workflow factory
  • move the tool routing logic to the base class
  • better logging
  • move get_chat_settings to BaseWorkflow
  • move chat_settings to BaseWorkflow and add workflow-wise ChatProfile
  • remove hard code output chat model name
  • dynamic workflow loading and use langchain native tool calling
  • separate name and value fields of the tool results
  • move to search module
  • setting module
  • move graph and state to graphs module

Perf

  • better performance
  • async tools binding
  • change tool selection model options
  • modify tool selection prompt