Smart Router
Live routing rules · model performance · change LLM per task without a deploy
LLM Providers — Live Status
Task Routing Rules
Per-task LLM assignment — click Edit to override the model for any task type
SARAH Platform Routing
Per social-platform routing — Twitter, LinkedIn, Instagram, TikTok and more
Model Performance Stats
Per-Agent Model Config
Priority 2 — overrides generic task rulesClick Edit on any row to override a model at runtime (saved to Redis instantly — no restart needed).
Provider API Keys
— set in .env file, restart API to apply| Provider | Env Variable | Models Available | Status | Priority |
|---|---|---|---|---|
| OpenRouter | OPENROUTER_API_KEY | All 22 models (gateway) | ○ Not configured | — |
| OpenAI | OPENAI_API_KEY | GPT-4o, GPT-4o-mini, GPT-4.1, GPT-5... | ○ Not configured | — |
| xAI (Grok) | XAI_API_KEY | Grok-3, Grok-4.1-fast | ○ Not configured | — |
| Anthropic | ANTHROPIC_API_KEY | Claude 3.5, 3.7, Sonnet 4 | ○ Not configured | — |
| Google AI | GOOGLE_AI_API_KEY | Gemini 2.0, 2.5, 3.1... | ○ Not configured | — |
| DeepSeek | DEEPSEEK_API_KEY | DeepSeek V3, V3.2 | ○ Not configured | — |
| Mistral | MISTRAL_API_KEY | Mistral Large, Medium... | ○ Not configured | — |
| Groq/Llama | GROQ_API_KEY | Llama 3.3 70B, 8B... | ○ Not configured | — |
To activate a provider: add its API key to morezo/.env → restart API (pm2 restart morezo-api) → provider appears in LLM Control automatically. Set USE_DIRECT_PROVIDERS=true to bypass OpenRouter entirely and use direct APIs.
How Smart Router Works
Explicit Model
Agent requests a specific model → used directly, bypassing routing.
Task Type
taskType set → router picks the best model per task routing rules above.
Platform Rule
SARAH social task → platform-specific routing (Twitter=Grok, LinkedIn=GPT-4o-mini).
Fallback Chain
Primary fails → fallback model → highest-priority available provider.