Actors
An Actor is a YAML-configured LLM graph that CleverThis deploys and executes as if it were a single model. From a caller's perspective, sending a request to an Actor looks identical to calling gpt-4o — the same POST /v1/chat/completions endpoint, the same request shape, the same streaming response. Internally, an Actor may route through several models, call tools, apply prompt templates, run conditional logic, or fan out to parallel sub-agents before assembling a final response.
Actors are the primary extension point in CleverThis. They let you encode product-specific behavior — retrieval-augmented generation pipelines, multi-step reasoning chains, tool-using agents, and model routing strategies — into a declarative YAML file that can be versioned, tested, packaged, and deployed without touching application code.
What's in this section
- Actor YAML Format — high-level overview of the YAML structure
- Creating an Actor — step-by-step: write YAML, upload, deploy
- Testing Actors — using the Chat interface and the test harness to validate actor behavior
When to use Actors
Consider building an Actor when you want to:
- Reuse prompt logic across multiple API consumers without duplicating it in client code
- Route dynamically between models based on request content, token count, or cost
- Chain steps — e.g. classify → retrieve → generate — as a single model call
- Inject guardrails at the gateway layer rather than in application code
- Package and version a complete AI behavior for promotion through dev/staging/prod
The Actor YAML Configuration Standard reference is the authoritative spec for all YAML fields: Actor Configuration Standard.