API Overview

This section is the reference for the Buddy AI Python library — the classes and functions you import from the buddy package to build agents, teams, knowledge bases, and workflows. It is not a REST API reference (though Buddy can also serve agents over HTTP — see Serving over HTTP below and Integrations).

How to read these pages

Each page documents the constructor parameters and the runtime methods of one core class, grounded in the source under buddy/. Parameters are grouped into tables; defaults match the code. Optional features (planning, multimodal, reasoning, security, PULSE) import lazily, so they only appear if their extra dependencies are installed — check with get_available_features().

The public surface

Everything you need for the common cases is importable directly from the top-level package:

from buddy import (
    Agent,            # a single LLM-powered agent
    Team,             # a group of agents that collaborate
    Model,            # base class for model providers
    Function,         # a tool wrapping a callable
    Toolkit,          # a collection of related tools
    AgentMemory,      # conversation/session memory
    AgentKnowledge,   # retrieval / knowledge base base class
    get_version_info,
    get_available_features,
    check_feature,
)

Core classes

Class	Import	Reference
Agent	from buddy import Agent	Agent API
Team	from buddy import Team	Team API
AgentKnowledge	from buddy import AgentKnowledge	Knowledge API
AgentMemory	from buddy import AgentMemory	Memory
Workflow	from buddy.workflow import Workflow	Workflow API
Model	from buddy.models.base import Model	Models
Toolkit / Function / @tool	from buddy.tools import Toolkit, Function, tool	Tools

Provider classes live under buddy.models.* (e.g. from buddy.models.openai import OpenAIChat), tools under buddy.tools.*, knowledge bases under buddy.knowledge.*, and vector stores under buddy.vectordb.*.

Feature detection

The package degrades gracefully when optional extras are missing. Use the helper functions to introspect what is available at runtime:

from buddy import get_version_info, get_available_features, check_feature

get_version_info()        # {"version": ..., "features": {...}, ...}
get_available_features()  # e.g. ["planning", "reasoning", "pulse", "core"]
check_feature("pulse")    # True / False

The feature keys are planning, multimodal, evolution, reasoning, personality, security, pulse, and core (always True).

Serving over HTTP

Agents, teams, and workflows can be exposed as a FastAPI application via FastAPIApp (from buddy.app.fastapi import FastAPIApp). The generated app exposes a small REST surface:

Method	Path	Purpose
GET	/status	Health check — returns {"status": "available"}
POST	/runs	Run an agent, team, or workflow

POST /runs accepts form fields (message, stream, monitor, session_id, user_id, workflow_input, files) and selects the target via the agent_id, team_id, or workflow_id query parameter. See Integrations for a deployment walkthrough.

Prefer the Python API for embedding

If you are embedding Buddy inside your own service, call Agent.run() / Team.run() directly rather than going through HTTP — you get typed RunResponse objects instead of JSON.