SRSafetyRouter Talk to us
Tested · v1
Check your version first

Run opencode --version. If it begins with 2., follow the separate OpenCode v2 guide. The configurations are not interchangeable.

1. Create an API-key identity

In the tenant portal, create a key named for the person and client—for example Steele-OpenCode-Mac. Add project tags separately.

2. Export the key

export SAFETYROUTER_API_KEY='sr_live_...'

3. Add the v1 provider

Add this provider to opencode.json. Replace the hostname and model with values allowed by your tenant.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "safetyrouter": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "SafetyRouter",
      "options": {
        "baseURL": "https://your-company.safetyrouter.io/v1",
        "apiKey": "{env:SAFETYROUTER_API_KEY}",
        "timeout": 900000,
        "headerTimeout": 900000,
        "chunkTimeout": 900000
      },
      "models": {
        "provider/model-name": { "name": "Model via SafetyRouter" }
      }
    }
  },
  "model": "safetyrouter/provider/model-name"
}

4. Start OpenCode v1

opencode
Buffered-safe streaming

OpenCode may pause while SafetyRouter generates and inspects the full answer. Approved content is then emitted as a compatible stream.

Buffered-safety timeout

Allow up to 15 minutes

SafetyRouter intentionally waits for input review, model generation, and output review before releasing model content. Where your client exposes request, response-header, socket-read, or stream-idle timeouts, set them to at least 900000 ms or 900 seconds. Keep streaming enabled. A longer timeout preserves the safety boundary; it does not bypass inspection.

This page preserves the OpenCode 1.x configuration format. New installations should use the OpenCode v2 guide.