Integrations

How to deploy Buddy agents and connect them to model providers, vector stores, and chat platforms.

Deploy as a REST API (FastAPI)

FastAPIApp wraps agents, teams, and workflows in a FastAPI application. Build the app with get_app(), then serve() it.

from buddy import Agent
from buddy.models.openai import OpenAIChat
from buddy.app.fastapi import FastAPIApp

agent = Agent(
    name="api_agent",
    model=OpenAIChat(id="gpt-4o-mini"),
    instructions="You are a helpful assistant.",
    markdown=True,
)

app = FastAPIApp(agents=[agent])
fastapi_app = app.get_app()  # returns a FastAPI instance

if __name__ == "__main__":
    app.serve(fastapi_app, host="0.0.0.0", port=7777)

FastAPIApp(agents=..., teams=..., workflows=...) accepts any combination (at least one is required). serve() is a method on the app and requires the FastAPI app object as its first argument.

`serve()` parameter	Default	Description
`app`	— (required)	The FastAPI app (from `get_app()`) or an import string.
`host`	`"localhost"`	Bind address.
`port`	`7777`	Port.
`reload`	`False`	Auto-reload (dev).
`workers`	`None`	Number of worker processes.

The generated API exposes GET /status and POST /runs. Select the target with the agent_id (or team_id / workflow_id) query parameter:

curl -X POST "http://localhost:7777/runs?agent_id=<AGENT_ID>" \
  -F "message=What is 2 + 2?"

Common deployment pitfalls

FastAPIApp has no port argument in its constructor — set the port on serve(). And serve() is not a zero-argument call: pass the app from get_app().

Chat platforms

Buddy ships adapters for several messaging platforms under buddy.app.*:

Platform	Import	Class
Slack	`from buddy.app.slack import SlackAPI`	`SlackAPI`
WhatsApp	`from buddy.app.whatsapp import WhatsappAPI`	`WhatsappAPI`
Discord	`from buddy.app.discord import DiscordClient`	`DiscordClient`

Each wraps an agent or team behind that platform's webhook/client. Configure the platform's credentials (tokens, signing secrets) via environment variables as required by the provider's SDK.

Model providers

Providers live under buddy.models.*. Install the matching extra and set the provider's API key.

Provider	Import	Extra
OpenAI	`from buddy.models.openai import OpenAIChat`	`buddy-ai[openai]`
Anthropic	`from buddy.models.anthropic import Claude`	`buddy-ai[anthropic]`
Google	`from buddy.models.google import Gemini`	`buddy-ai[google]`
AWS Bedrock	`from buddy.models.aws import AwsBedrock`	`buddy-ai[aws]`
Groq	`from buddy.models.groq import Groq`	—
Ollama (local)	`from buddy.models.ollama import Ollama`	—

Buddy includes 30+ provider packages under buddy.models.* — see Model Providers for the full list.

from buddy import Agent
from buddy.models.ollama import Ollama

# Run fully locally with Ollama — no API key required
agent = Agent(model=Ollama(id="llama3"))
agent.print_response("Hello from a local model!")

Vector databases

Knowledge bases can be backed by any vector store under buddy.vectordb.*, for example ChromaDb:

from buddy.knowledge.pdf import PDFKnowledgeBase
from buddy.vectordb.chroma import ChromaDb

kb = PDFKnowledgeBase(
    path="docs/",
    vector_db=ChromaDb(collection="docs", persistent_client=True, path="tmp/chromadb"),
)
kb.load()

See the Knowledge API for the full knowledge-base and vector-DB reference.

PULSE — virtual employee

For the higher-level "virtual employee" integration (onboarding, knowledge transfer, professional memory), see PULSE — Virtual Employee. Availability can be checked at runtime with check_feature("pulse").