This article records configuration and changes from a historical FastGPT release. Use it to understand that release, and consult the release notes for your installed and target versions before applying dependencies, configuration, or migration steps.
New Core Features
This update delivers targeted functional enhancements:
- Plugin inventory now includes clear indicators for deprecated and deleted plugins
- Chat logs support source categorization, title-based search, and full export capabilities
- Global variables can be reordered via intuitive drag-and-drop actions
- New official model presets: Doubao 1.5 and Alibaba Embedding v3
- Vector models now support normalization configuration for unnormalized models such as Doubao embeddings
- AI chat nodes can output and share thinking process results with downstream workflow nodes
- Embedded chat widget automatically adapts to window positioning
- LLM models now support three additional configurable parameters
Configurable LLM Model Parameters
The following newly supported parameters are available for LLM model configuration:
| Parameter Name | Functional Purpose |
|---|---|
top_p | Nucleus sampling probability threshold for model output control |
response_format | Defines the desired output format for model responses |
json_schema | Enforces JSON schema validation for structured model outputs |
Usability & Privacy Improvements
Several quality-of-life and privacy-focused updates are included:
- Improved error messaging for unconfigured LLM or embedding models to streamline troubleshooting
- Added support for non-streaming mode thinking process output
- Implemented null pointer protection for unconfigured TTS voice settings to prevent crashes
- Switched markdown link parsing to strict matching mode to reduce false positive detections
- Reduced data fetch scope for unauthenticated users to enhance system privacy
Resolved Bug Fixes
This update addresses the following stability and functionality issues:
- Fixed Simple Mode: switching to a non-vision model now correctly disables image recognition
- Resolved o1/o3 model field mapping failure during testing, which previously caused runtime errors
- Fixed null pointer exception in WeChat Official Account chat integration
- Corrected incorrect display of multiple audio and video file attachments
- Fixed share link authentication error that caused infinite redirect loops