DeepSeek Harness (DSH) Plugin Introduction: dsh-llm-opencode-go
The DeepSeek Harness (DSH) plugin `dsh-llm-opencode-go` natively integrates OpenCode Go with DSH. It supports both the OpenAI and Anthropic protocols, implementing full tool calling, reasoning mapping, session association, and streaming output. The plugin provides a hybrid in-memory and disk-based caching mechanism and stable error codes. Users must configure an API key reference in `cordis.patch.yml`, set the key via environment variables or the DSH credential seam, and view quotas and usage in the web client. It requires Node.js ≥ 20 and takes effect only after restarting the profile.
Read Moredsh-llm-opencode-go: OpenCode Go Model Inference Plugin for DeepSeek Harness
`dsh-llm-opencode-go` is a DeepSeek Harness plugin for integrating the OpenCode Go model into DSH environments. It provides hybrid support through a bridging protocol, uses native endpoints for model discovery, and supports visual and reasoning capabilities. Developers can configure the API key via environment variables and manage model resources through the Web interface. DSH version compatibility must be ensured during installation. Note that the plugin does not rely on `registerConfigurableProviders` and has limitations such as CodexHub lock file conflicts and an unsupported `stop` parameter.
Read Moredsh-llm-opencode-go: OpenCode Go Model Provider
This plugin serves as the OpenCode Go model provider for DeepSeek Harness, implementing session identity mapping and real-time model catalog refresh through Wire protocol routing. Its core features include adding session IDs to request headers, dynamically fetching new models, and providing a web management card. Users can configure API keys, endpoints, and refresh intervals in the web interface, with support for model overrides, model disabling, and input type checks. It requires DeepSeek Harness 0.2.0-rc.2 or later, and configuration is persisted via `cordis.patch.yml`.
Read More