Workflow Stability and Detection Improvements
This release adds targeted safeguards for workflow execution, starting with anomaly detection for isolated branches in workflow graphs. This alerts users to disconnected workflow segments that would otherwise fail during runtime. The workflow scheduling code was also optimized to eliminate deep recursion, preventing stack overflow errors in complex, nested workflow runs. Additionally, workflow recursion detection was improved with grouped checks on recursive paths, supporting a broader set of connection patterns than prior versions, reducing false negative detections and catching invalid recursive workflows earlier in the build process.
Embedding Vector Normalization Updates
The embedding processing pipeline received targeted performance and behavior updates. A clear, conditional normalization rule set was implemented for truncated embedding vectors:
| Truncated Vector Dimension | Normalization Behavior |
|---|---|
| Greater than 1536 | Enforced normalization |
| 1536 or fewer | Config-controlled only |
This change eliminates unnecessary automatic normalization computations for vectors under 1537 dimensions, reducing computational overhead while preserving full configuration flexibility for standard use cases.
LLM and Plugin SDK Refinements
Two key updates were made to LLM and plugin infrastructure. First, model provider configuration was moved entirely into the plugin SDK, centralizing provider-specific settings and simplifying cross-provider maintenance. Second, LLM call functions were encapsulated to standardize and simplify LLM requests and tool call integrations. This encapsulation reduces redundant code across the codebase, making it easier to implement and maintain LLM-powered workflow steps and tool interactions.
Source: FastGPT official source
Applicability and version scope
Use this page for the documented Deployment and upgrades scenario. Confirm the FastGPT, dependency, API, and deployment versions in the official source before applying a change.
Safety guardrails
Use [REDACTED_CREDENTIAL] for credentials and private data. Confirm the documented environment and version before review.
Rollback guidance
Restore the prior technical-content authority snapshot. Restore saved configuration and data snapshots, then repeat the smallest verification scenario.