dsh-llm-opencode-go

Web configuration

Open Settings → LLM Providers → OpenCode Go. The provider-management RPC returns only decoded settings, revision, and value-free credential status; API keys are write-only and never echoed or logged. The Connection RPC authenticates through the Web trust fence, while durable settings writes require a loopback settings scope; use SSH forwarding when the browser is remote.

The card saves the public base URL and model catalog together as one revision-fenced llm-opencode-go settings mutation. Fetch available models opens the picker immediately. The Host reads GET /zen/go/v1/models (OpenAI-shaped ids only) and fills name, context, vision, thinking, and protocol from a local snapshot, then from a live models.dev opencode-go overlay so newly published ids such as omen-alpha are not blank after Fetch.

When a key is stored, the card loads account quota while collapsed (and again on Fetch/Refresh). The Host reads GET <baseURL>/usage; remaining 5-hour, weekly, and monthly windows paint from a local cache on first open, then refresh in the background. The credential never crosses to the browser.

The model catalog starts collapsed and lists one row per model: a drag handle reorders rows (the order persists with the catalog), the chevron opens that row's context and capability flags, and the trash button removes it.

Plugin configuration

The Models page lists saved opencode-go models and can select them. Current Harness releases do not expose a third-party editor slot inside that page, so this package owns its editor under LLM Providers.

Plugin configuration

The Models page lists saved opencode-go models and can select them. Current Harness releases do not expose a third-party editor slot inside that page, so this package owns its editor under LLM Providers.