Chainlit

Latest version: v1.1.301

Safety actively analyzes 638699 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 6 of 9

0.7.2

Not secure
Added

- Logo is displayed in the UI header (works with custom logo)
- Azure AD single tenant is now supported
- `collapsed` attribute on the `Action` class
- Latency improvements when data persistence is enabled

Changed

- Chat history has been entirely reworked
- Chat messages redesign
- `config.ui.base_url` becomes `CHAINLIT_URL` env variable

Fixed

- File watcher (-w) is now working with nested module imports
- Unsupported character during OAuth authentication

0.7.1

Not secure
Added

- Pydantic v2 support
- Okta auth provider
- Auth0 auth provider
- Prompt playground support for mix of template/formatted prompts
- `cl.on_chat_end` decorator
- Textual comments to user feedback

Fixed

- Langchain errors are now correctly indented
- Langchain nested chains prompts are now correctly displayed
- Langchain error TypeError: 'NoneType' object is not a mapping.
- Actions are now displayed on mobile
- Custom logo is now working as intended

0.7.0

Not secure
Changed

- Authentication is now unopinionated:
1. `cl.password_auth_callback` for login/password auth
2. `cl.oauth_callback` for oAuth auth
3. `cl.header_auth_callback` for header auth
- Data persistence is now enabled through `CHAINLIT_API_KEY` env variable

Removed

- `cl.auth_client_factory` (see new authentication)
- `cl.db_client_factory` (see new data persistence)

Added

- `disable_human_feedback` parameter on `cl.Message`
- Configurable logo
- Configurable favicon
- Custom CSS injection
- GCP Vertex AI LLM provider
- Long message collpasing feature flag
- Enable Prompt Playground feature flag

Fixed

- History page filters now work properly
- History page does not show empty conversations anymore
- Langchain callback handler Message errors

0.6.4

Not secure
Added

- `cl.on_file_upload` to enable spontaneous file uploads
- `LangchainGenericProvider` to add any Langchain LLM in the Prompt Playground
- `cl.Message` content now support dict (previously only supported string)
- Long messages are now collapsed by default

Fixed

- Deadlock in the Llama Index callback handler
- Langchain MessagesPlaceholder and FunctionMessage are now correctly supported

0.6.3

Not secure
Added

- Complete rework of the Prompt playground. Now supports custom LLMs, templates, variables and more
- Enhanced Langchain final answer streaming
- `remove_actions` method on the `Message` class
- Button to clear message history

Fixed

- Chainlit CLI performance issue
- Llama Index v0.8+ callback handler. Now supports messages prompts
- Tasklist display, persistence and `.remove()`
- Custom headers growing infinitely large
- Action callback can now handle multiple actions
- Langflow integration load_flow_from_json
- Video and audio elements on Safari

0.6.2

Not secure
Added

- Make the chat experience configurable with Chat Settings
- Authenticate users based on custom headers with the Custom Auth client

Fixed

- Author rename now works with all kinds of messages
- Create message error with chainlit cloud (chenjuneking)

Page 6 of 9

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.