This release fixes issues for SREs watching stability and regressions.
✓ No known CVEs patched in this version
Topics
+3 more
ReleasePort's take
Light signalRelease v4.14.22 of FastGPT fixes several workflow issues: model selection now propagates form values, auto‑save no longer risks edge loss, admin notification dialogs work correctly, and mixed model contexts avoid reasoning field errors.
Why it matters: Addresses bugs that caused mismatched models, potential data loss during auto‑save, notification UI failures, and context errors in workflows; version v4.14.22 released 2026-05-22.
Summary
AI summaryFixed workflow model mismatch, auto-save loss risk, admin notification error, and independent reason field causing model call errors.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
Adds safeguard to avoid standalone reasoning messages in FastGPT workflows. Adds safeguard to avoid standalone reasoning messages in FastGPT workflows. Source: llm_adapter@2026-05-22 Confidence: low |
— |
| Bugfix | Medium |
Fixes issue where workflow default model selection did not propagate form values, causing mismatched models. Fixes issue where workflow default model selection did not propagate form values, causing mismatched models. Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Bugfix | Medium |
Resolves error when admin updates system notification dialog. Resolves error when admin updates system notification dialog. Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Bugfix | Medium |
Prevents reasoning field context errors when workflows mix thinking and non-thinking models. Prevents reasoning field context errors when workflows mix thinking and non-thinking models. Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Bugfix | Medium |
Prevents edge case data loss when workflows auto-save, reducing risk of missing edges. Prevents edge case data loss when workflows auto-save, reducing risk of missing edges. Source: llm_adapter@2026-05-22 Confidence: low |
— |
| Bugfix | Medium |
Prevents edge loss risk during automatic workflow saving. Prevents edge loss risk during automatic workflow saving. Source: granite4.1:30b@2026-05-22-audit Confidence: low |
— |
Full changelog
🐛 修复
- 工作流默认选中模型未回传表单值,导致看到的模型和实际运行模型不一致。
- 工作流自动保存时,存在边丢失风险。
- admin 修改系统通知弹窗时候报错。
- 工作流混用思考/非思考模型,可能出现独立 reason 字段上下文,导致模型调用报错。
What's Changed
- fix: avoid standalone reasoning messages by @c121914yu in https://github.com/labring/FastGPT/pull/6960
Full Changelog: https://github.com/labring/FastGPT/compare/v4.14.21...v4.14.22
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About FastGPT
All releases →Related context
Related tools
Beta — feedback welcome: [email protected]