Deployment and upgradesOfficial documentation6 min readDeployment and upgrades

FastGPT 4.14.14 Performance and Compatibility Optimizations

This document covers the targeted optimization and compatibility improvements released in FastGPT 4.14.14 for self-hosted deployments.

This document covers the targeted optimization and compatibility improvements released in FastGPT 4.14.14 for self-hosted deployments. All changes are designed to reduce operational bottlenecks and improve reliability across core workflows.

WeChat Publishing Channel Polling Optimization

The personal WeChat publishing channel’s polling strategy has been fully optimized by decoupling message pull operations from reply handling. This structural change prevents system blocking during periods of high message volume, ensuring consistent performance even when the channel is processing large numbers of concurrent messages.

A new environment variable has been added to control the concurrency of WeChat channel poll workers:

Environment Variable NameDefault ValueRecommended Setting
WECHATCHANNELCONCURRENCY1000Set to a value ≥ peak online channel count

This variable allows administrators to fine-tune the number of concurrent poll workers to match their deployment’s active channel capacity, eliminating under-provisioning or over-provisioning of worker resources.

Internal Network Address Detection Improvements

The internal network address detection logic has been significantly improved. This update refines how the system identifies and validates internal network addresses, reducing configuration errors and improving compatibility with complex network topologies commonly found in self-hosted FastGPT deployments. No additional user configuration is required to enable this improvement, as it is applied automatically during deployment.

DeepSeek Tool Calling and Thinking Mode Compatibility

A critical API compatibility issue has been resolved for users integrating DeepSeek models. Prior to this update, combining DeepSeek tool calling functionality with the thinking mode feature would trigger 400 errors from the DeepSeek API. This release adds full compatibility for this combined workflow, ensuring that users can leverage both tool calling and thinking mode without encountering API errors, with no manual adjustments needed to their existing model configurations.

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.