Deployment and upgradesOfficial documentation5 min read版本解读

Key FastGPT Self-Hosted Upgrade Performance Improvements

This article records configuration and changes from a historical FastGPT release.

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.

Document and Database Performance Fixes

This upgrade includes targeted optimizations for document parsing and database query operations. The Doc2x document parsing module now features enhanced error message capture, which provides detailed failure information for failed parsing jobs to streamline troubleshooting. Additionally, the module’s timeout duration has been extended to support processing of larger document files without premature job termination. For PostgreSQL vector database deployments, the vector query logic has been adjusted to force usage of preconfigured vector indexes, preventing unindexed full-table scans and improving search efficiency for vector-based retrieval tasks.

Timing and Data Retrieval Updates

Two critical timing and retrieval improvements have been implemented: First, conversation time statistics now accurately return the total end-to-end workflow execution time, replacing the previous partial segment timing reporting that undercounted total processing length. Second, audio parsing duration data is now retrieved directly from the ai_proxy service, aligning duration tracking with proxy-side processing metrics to ensure consistency across distributed deployments. The chat log list API has also been refactored to handle scenarios with very large numbers of messages in a single conversation, reducing latency and preventing request timeouts for long chat histories.

Token Count Calculation Optimizations

The AI model token count calculation logic has been updated to prioritize accuracy by using official API-provided usage values when available. If API usage values are not accessible, the system falls back to GPT-3.5 estimation for token count calculations. A quick reference for the token count workflow is below:

Trigger ConditionToken Count Source
API usage values presentOfficial API-reported token counts
No API usage values availableGPT-3.5 estimation model

Source: FastGPT official documentation and source