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.
Core New Features
This release adds targeted functionality for workflow expansion, commercial deployment, and analytics:
- Simple Mode tool calls now support array-type plugins, enabling structured data handling for integrated third-party tools.
- Workflow auto-save on abnormal exit prevents permanent loss of unsaved workflow configurations during unexpected crashes or interruptions.
- Commercial edition administrators can configure the official template marketplace directly through the admin panel, simplifying template distribution across teams.
- Commercial edition admins may set custom workflow variables to support business system authentication integration, enabling secure cross-system workflow connectivity.
- The search test API now supports query rewriting, improving search result relevance for natural language user queries.
- Workflow token tracking splits input and output tokens separately for granular cost monitoring, with a critical fix for billing systems that failed to record output tokens for some prior requests.
LLM Parameter & UI Configuration Updates
This section includes revised model parameter controls and usability improvements: LLM model parameters now support disabling both max_tokens and temperature fields, giving deployments full control over exposed model settings. A reference table for the updated parameter behavior is below:
| Parameter | Previous Behavior | Current Behavior |
|---|---|---|
max_tokens | Mandatory fixed input field | Optional field, can be fully disabled |
temperature | Mandatory fixed input field | Optional field, can be fully disabled |
Additional improvements include:
- Markdown size check updated: content exceeding 200K characters no longer uses the Markdown component, eliminating application crashes from large markdown content.
- Dataset search parameters: slider controls now support direct manual input, allowing for more precise parameter adjustment beyond standard slider snap points.
- The available models display UI has been revised for improved readability and faster load times.
- MongoDB queries now include virtual fields, expanding data query flexibility for custom dataset integrations.
Resolved Bug Fixes
Four critical and minor bugs are resolved in this release:
- Fixed an issue where the File response API omitted the
Content-Lengthheader, causing Alibaba vision models to fail image recognition when uploading files from cross-origin sources. - Removed hidden leading and trailing line breaks from condition node strings, fixing condition evaluation failures caused by unintended whitespace.
- Fixed a variable update node bug where non-string data types could not be automatically converted when manually entering update content.
- Resolved an issue where Doubao models were unable to execute tool calls properly.