occa-opencode-provider

OCCA OpenCode Provider Plugin

Automatically reads occa.json and registers multiple providers (OpenAI / Claude / Gemini) into OpenCode with model lists fetched from the API.

Features

• Auto-detect config — scans current directory, then fallback to ~/.config/opencode/occa.json
• Multi-provider — one config file, unlimited providers
• Auto model fetch — pulls model lists from each API on startup
• Config validation — clear error messages for misconfiguration
• Hot reload — watches config file, auto-reloads on changes
• Model filtering — include/exclude models per provider with glob patterns
• Token masking — API keys masked in logs (sk-***abc)
• Model caching — caches model lists with configurable TTL
• Custom headers — add extra headers per provider
• Per-provider timeout — configurable request timeout

Installation

npm install -g occa-opencode-provider

Add to opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["occa-opencode-provider"]
}

Configuration

The plugin looks for occa.json in this order:

• Environment variable: OCCA_CONFIG_PATH (full path to config file)
• Current directory: ./occa.json (where you run opencode)
• Default location: ~/.config/opencode/occa.json

Example: Run opencode in /workspace/ with /workspace/occa.json — plugin auto-detects it.

Minimal

{
  "provider": {
    "openai": {
      "baseurl": "https://api.openai.com/v1",
      "key": "sk-xxx",
      "type": "openai"
    }
  }
}

Full

{
  "settings": {
    "cache_ttl": 1800,
    "hot_reload": true
  },
  "provider": {
    "openai": {
      "baseurl": "https://api.openai.com/v1",
      "key": "sk-xxx",
      "type": "openai",
      "timeout": 15000,
      "headers": {
        "X-Custom-Header": "value"
      },
      "models": {
        "include": ["gpt-4*", "o3*"],
        "exclude": ["*vision*", "*audio*"]
      }
    },
    "claude": {
      "baseurl": "https://api.anthropic.com",
      "key": "sk-ant-xxx",
      "type": "claude"
    },
    "gemini": {
      "baseurl": "https://generativelanguage.googleapis.com/v1beta",
      "key": "AIza-xxx",
      "type": "gemini"
    },
    "openrouter": {
      "baseurl": "https://openrouter.ai/api/v1",
      "key": "sk-or-xxx",
      "type": "openai",
      "models": {
        "include": ["anthropic/*", "google/*", "openai/*"]
      }
    },
    "custom-api": {
      "baseurl": "https://api.example.com/v1",
      "key": "your-api-key",
      "type": "openai",
      "headers": {
        "auth_header": {
          "key": "X-API-Key",
          "value": "your-api-key"
        }
      }
    }
  }
}

Config Reference

settings

Field	Type	Default	Description
cache_ttl	number	1800	Model cache TTL in seconds (0 to disable)
hot_reload	boolean	true	Watch config file for changes and auto-reload

provider.<name>

Field	Type	Required	Description
baseurl	string	Yes	API base URL
key	string	Yes	API key / token
type	string	No	API type: openai | claude | gemini (default: openai)
timeout	number	No	Request timeout in ms (default: 15000)
headers	object	No	Custom headers to include in API requests
headers.auth_header	object	No	Custom authentication header format (e.g. {key: "X-API-Key", value: "your-key"})
models.include	string[]	No	Glob patterns — only include matching models
models.exclude	string[]	No	Glob patterns — exclude matching models

Supported API Types

Type	SDK	Models Endpoint	Providers
openai	@ai-sdk/openai-compatible	GET {baseurl}/models	OpenAI, DeepSeek, Groq, OpenRouter, Together, Fireworks, Ollama, xAI, etc.
claude	@ai-sdk/anthropic	GET {baseurl}/v1/models	Anthropic Claude
gemini	@ai-sdk/google	GET {baseurl}/models?key={key}	Google Gemini

Example Providers

OpenAI

{
  "baseurl": "https://api.openai.com/v1",
  "key": "sk-proj-xxx",
  "type": "openai"
}

Claude (Anthropic)

{
  "baseurl": "https://api.anthropic.com",
  "key": "sk-ant-api03-xxx",
  "type": "claude"
}

Gemini (Google)

{
  "baseurl": "https://generativelanguage.googleapis.com/v1beta",
  "key": "AIzaSy-xxx",
  "type": "gemini"
}

OpenRouter

{
  "baseurl": "https://openrouter.ai/api/v1",
  "key": "sk-or-v1-xxx",
  "type": "openai",
  "models": {
    "include": ["anthropic/*", "google/*", "openai/*", "meta-llama/*"]
  }
}

DeepSeek

{
  "baseurl": "https://api.deepseek.com/v1",
  "key": "sk-xxx",
  "type": "openai"
}

Groq

{
  "baseurl": "https://api.groq.com/openai/v1",
  "key": "gsk_xxx",
  "type": "openai"
}

Ollama (local)

{
  "baseurl": "http://localhost:11434/v1",
  "key": "ollama",
  "type": "openai"
}

Logging & Debugging

Logs are written to ~/.cache/opencode/occa-plugin/:

File	Content
debug.log	All activity (startup, config, model fetches, cache hits)
error.log	Errors only (failed fetches, parse errors, validation)
models-cache.json	Cached model lists

# Watch all logs
tail -f ~/.cache/opencode/occa-plugin/debug.log

# Watch errors only
tail -f ~/.cache/opencode/occa-plugin/error.log

License

MIT