Models
In CleverThis, a "model" is anything addressable by the /v1/chat/completions endpoint. This includes two distinct kinds of things:
Provider models are upstream LLM models from providers like OpenAI, Anthropic, Google, Mistral, and others that have been wired into your CleverThis instance. They are identified by a provider/model-name slug (e.g. openai/gpt-4o, anthropic/claude-3-5-sonnet-20241022). Calling them proxies the request through CleverThis to the upstream provider, applying rate limiting, usage tracking, and any namespace-level policies.
Actor endpoints are custom LLM graphs that you or your team have defined in YAML and deployed to a namespace. From the API's perspective they look identical to provider models — they have a model ID, appear in /v1/models, and accept the same chat/completions request shape. Internally they may fan requests across multiple models, apply system prompts, invoke tools, or run routing logic.
What's in this section
- Browsing Models — how to use the Models page to find and filter available models
- Deploying a Model — how to add a model to a namespace endpoint so it is addressable under a specific route
Model identifiers
All models, whether provider or actor, are referenced by a string model ID. Provider model IDs follow the provider/name convention. Actor model IDs are assigned when the actor is deployed to a namespace and typically follow the pattern namespace/actor-name.
A searchable model catalog with context windows, pricing per million tokens, and capability flags is being added to this section.