deepseek-harness-model-config

When connecting multiple model services through DeepSeek Harness, users often need to manage endpoints, credentials, protocols, model lists, and capacity settings. This plugin adds an advanced model configuration page to the DeepSeek Harness Web UI, making it easier to create custom endpoints and manage candidate models visually.

It supports endpoint name, URL, API key, protocol, and custom request headers, with compatibility for openai-completions, openai-responses, and anthropic-messages. Users can fetch and select candidate models through a unified flow, search by model ID or display name, manually add model IDs, and adjust parameters such as default capacity, input modalities, reasoning compatibility, caching, transport, timeout, and retry behavior. It can also fill in missing context window, maximum output, input modality, and reasoning information from models.dev.

This plugin is a good fit for teams or individual users who connect DeepSeek Harness to self-hosted proxies, third-party compatible APIs, or multiple model providers, and who want to review model configurations before saving, keep API keys out of plain configuration files, and tune capabilities for each model.