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.
This document details all resolved bugs included in the FastGPT 4822 upgrade, spanning core chat functionality, API workflows, deployment compatibility, and team management tools.
Core Chat & Streaming Fixes
This category addresses issues with chat response generation and streaming behavior:
- Thinking content generated by models is no longer excluded from output token count calculations
- Streaming thinking chain content now appears in correct sequential order alongside main chat content, eliminating out-of-order display
- Vector search results now include deduplication to avoid returning duplicate entries
- User selection nodes in workflows now establish proper connections without configuration failures
- Chat record source metadata is saved correctly, preserving original interaction context
API & Workflow Fixes
Updates to API and workflow functionality resolve critical operational errors:
- API workflow calls now include error detection to prevent incorrect filtering of images that do not support HEAD detection, resolving failed image processing for non-compliant media
- Template marketplace templates now load and deploy without embedded errors
- Workflow base nodes now complete upgrade processes successfully, eliminating stuck deployment states for new workflows
- List API calls no longer trigger null pointer exceptions when joining member data, stabilizing administrative API responses
Deployment & Team Management Fixes
Fixes improve compatibility and administrative usability:
- Guest chat windows correctly detect whether language recognition is enabled, ensuring proper activation of auto-detection features
- Chat log exports now work correctly for sub-path deployments, resolving file path errors during export
- Member lists refresh automatically when switching between teams, removing the need for manual page reloads
Validation Step-by-Step Checklist
Use this checklist to confirm all resolved bugs are fixed post-upgrade:
- Confirm thinking token counting: Initiate a chat with a model supporting thinking chains, verify thinking content is included in output token metrics.
- Validate streaming order: Start a streaming chat, confirm thinking chain content displays in the correct sequence relative to main response text.
- Test non-HEAD image processing: Call an API workflow with images that do not support HEAD detection, confirm no filtering errors and successful processing.
- Check team member list refresh: Switch between teams in the admin interface, confirm the member list updates automatically.
- Verify sub-path export compatibility: Deploy FastGPT to a sub-path, export chat logs, confirm the exported file loads without path errors.
- Test template marketplace import: Import a template from the official marketplace, confirm deployment without embedded errors.
- Validate user selection node connectivity: Add a user selection node to a workflow, confirm successful connection to adjacent nodes.
- Check vector search deduplication: Run repeated vector search queries, confirm only unique results are returned.
- Confirm guest language detection: Enable guest language recognition, access the guest chat window to confirm auto-detection activates.
- Verify chat record source saving: Complete a chat, confirm the saved record preserves original source metadata.