Articles tagged "Model monitoring"

DeepSeek Harness Model Sync Plugin: Monitor and Update Provider Models

2026-10-10 DeepSeek Harness plugin DeepSeek DSH plugin Model synchronization Model monitoring Configuration Management

The DeepSeek Harness (DSH) Model Sync plugin is designed to address the issue of missing new models when adding them manually. It supports fetching the latest model lists from various providers and offers features such as read-only inspection, selective adding, and search/filtering. It can also automatically set up multi-protocol routing and backup configurations. Users need to install and restart DSH, then sync models via the UI or the command line (`/model-sync`); all write operations are validated for safety and backed up.

dsh-balance-quota: Balance, Quota, and Model Health Monitoring Plugin

2026-10-04 DeepSeek Harness plugin DeepSeek DSH plugin dsh-balance-quota Model monitoring Balance Management

`dsh-balance-quota` is a model inference visualization plugin for DeepSeek Harness Web, designed to provide unified management of balance and health status across multiple API providers. Integrated below the chat interface, it supports official presets, custom HTTP endpoints, and JSON data extraction, and can monitor model availability, response time, and error rates. It offers flexible configuration options, including currency handling, status value mapping, full-model preview, and security mechanisms. The plugin must be installed via the DSH CLI, is released under the MIT License, and is suitable for developers who need to efficiently monitor the operational status of multiple models.

dsh-percentile: DeepSeek Harness Percentile Plugin

2026-09-22 DeepSeek Harness plugin DeepSeek DSH dsh-percentile Model monitoring Inferential analysis

`dsh-percentile` is a plugin for DeepSeek Harness that locally computes percentiles for inference metrics, addressing the limitation that averages cannot fully reveal details of data distribution. Maintained by developer `uckkk`, it is implemented in pure Node.js, has no network dependencies, and offers high execution efficiency. Installation requires `@deepseek-ai/cordis` ^4.0.1 and `@deepseek-ai/dsh-tools` >= 0.1.0-rc.6. It is suitable for developers and operations personnel who need in-depth analysis of distributions for metrics such as latency and token count, ensuring stable performance in offline environments.