# Multi-Provider Setup Reference

## Ray's Provider Configuration (Updated 2026-05-10)

### Goal: Local model as default, API providers available for switching

**Default model:** Local (`Qwen3.6-35B-A3B-UD-Q4_K_M.gguf` at `localhost:8081`)

### Providers Added

| Provider | Env Var | Base URL | Models Available |
|----------|---------|----------|-----------------|
| OpenRouter | `OPENROUTER_API_KEY` | `https://openrouter.ai/api/v1` | minimax-m2.5, minimax/minimax-m2.7, qwen/qwen3-235b-a22b, qwen/qwen3-coder:free |
| DeepSeek | `DEEPSEEK_API_KEY` | `https://api.deepseek.com/v1` | DeepSeek models |

### config.yaml Snippet

```yaml
model:
  default: Qwen3.6-35B-A3B-UD-Q4_K_M.gguf
  provider: custom:Qwen3.6-35B
  base_url: http://127.0.0.1:8081/v1

providers:
  openrouter:
    base_url: https://openrouter.ai/api/v1
    api_key_env: OPENROUTER_API_KEY
  deepseek:
    base_url: https://api.deepseek.com/v1
    api_key_env: DEEPSEEK_API_KEY

credential_pool_strategies:
  openrouter: fill_first
  deepseek: fill_first
```

### Switching Models

```bash
# In-session slash commands
/model openrouter/minimax-m2.5
/model openrouter/minimax/minimax-m2.7
/model openrouter/qwen/qwen3-235b-a22b
/model openrouter/qwen/qwen3-coder:free

# CLI
hermes model
```

### OAuth Setup

ChatGPT (OpenAI Codex): `hermes login --provider openai-codex`