Skip to content

opencode-go.md

Latest commit

 

History

History
120 lines (83 loc) · 3.4 KB

opencode-go.md

File metadata and controls

120 lines (83 loc) · 3.4 KB

OpenCode Go Backend

The opencode-go backend provides a single public backend name for OpenCode Go while routing requests internally across two upstream protocol flavors:

  • OpenAI-compatible chat completions for most supported models
  • Anthropic-compatible messages for the Minimax models that require it

This keeps client-side routing simple. You select opencode-go, and the proxy chooses the correct upstream protocol based on the model id.

Upstream Service Details

OpenCode Go service details and upstream usage documentation:

Configuration

Environment Variables

Configure the backend with your OpenCode API key:

export OPENCODE_GO_API_KEY="..."

You can also configure numbered variants to create conventional backend instances automatically:

export OPENCODE_GO_API_KEY_1="..."
export OPENCODE_GO_API_KEY_2="..."

When numbered variants are present, the proxy discovers backend instances such as opencode-go.1 and opencode-go.2.

YAML Example

backends:
  opencode-go:
    api_key: "${OPENCODE_GO_API_KEY}"
    api_url: "https://opencode.ai/zen/go/v1"
    extra:
      model_protocol_overrides:
        # Optional escape hatch for future upstream protocol changes.
        # custom-model: anthropic

Model Naming

Use the backend with canonical model identifiers in the form:

  • opencode-go/opencode-model-id for advertised model names
  • opencode-go:<model-id> when selecting the backend explicitly in request model strings

Examples:

  • opencode-go/glm-5
  • opencode-go/glm-5.1
  • opencode-go/kimi-k2.5
  • opencode-go/mimo-v2-pro
  • opencode-go/mimo-v2-omni
  • opencode-go/minimax-m2.5
  • opencode-go/minimax-m2.7

Protocol Routing

The proxy currently routes these models through the OpenAI-compatible upstream path:

  • glm-5
  • glm-5.1
  • kimi-k2.5
  • mimo-v2-pro
  • mimo-v2-omni

The proxy currently routes these models through the Anthropic-compatible upstream path:

  • minimax-m2.5
  • minimax-m2.7

If OpenCode changes protocol assignments later, you can override the built-in mapping with backends.opencode-go.extra.model_protocol_overrides.

Multiple API Keys and Instances

If you provide numbered environment variables such as OPENCODE_GO_API_KEY_1 and OPENCODE_GO_API_KEY_2, the proxy creates conventional backend instances:

  • opencode-go.1
  • opencode-go.2

This matches the project's standard backend-instance discovery behavior for multi-key backends. You can then target those specific instances with selectors like opencode-go.1:glm-5.1.

Usage Example

Example request through the OpenAI Chat Completions frontend:

curl -X POST http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "opencode-go:glm-5.1",
    "messages": [{"role": "user", "content": "Explain this code."}],
    "stream": true
  }'

Example request targeting a model that the proxy routes through the Anthropic-compatible upstream path:

curl -X POST http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "opencode-go:minimax-m2.7",
    "messages": [{"role": "user", "content": "Summarize the design tradeoffs."}]
  }'

Related Documentation