Built-in Tools (Large Language Models)

Large Language Model Tools

LLM capabilities are built into monkeys-server and are used by workflows, chat views, agents, and OpenAI-compatible APIs. Local or private deployments can point Monkeys to hosted providers or to self-hosted OpenAI-compatible model servers such as vLLM.

Configuration

Model configuration is deployment-specific. A typical config.yaml model entry looks like this:

models:
  - model: gpt-4o-mini
    baseURL: https://api.openai.com/v1
    apiKey: sk-xxxxxxxx
    type:
      - chat_completions
  - model: local-qwen
    baseURL: http://127.0.0.1:8000/v1
    apiKey: token-abc123
    type:
      - chat_completions
    defaultParams:
      temperature: 0.7

Parameter	Description
`model`	Model name shown to workflows and API callers.
`baseURL`	OpenAI-compatible API base URL.
`apiKey`	Provider API key. Leave empty only if the provider does not require one.
`type`	Supported invocation type, usually `chat_completions` or `completions`.
`defaultParams`	Default request parameters passed to the provider.

Usage

Use LLMs in three common ways:

Add a model node to a workflow in Studio.
Use an Agent backed by monkeys-agent-server.
Call a published workflow through an OpenAI-compatible API.

Tool Calling

Tool calling depends on both the model provider and workflow configuration. Not every model supports OpenAI-style tools parameters. When reliability matters, make tool inputs explicit in the workflow and validate outputs before passing them to downstream steps.

Private Data

To ground model responses in private data, combine an LLM node with text data, data assets, or private data search. See Private Data Search.