AI Capabilities

Structure of ai/

The ai/ folder is Python. Each file is a module that can be called from TypeScript through the RPC boundary.

ai/
├── __init__.py
├── chat.py          # LLM chat functions
├── agents/          # Agent orchestration
│   ├── __init__.py
│   └── researcher.py
├── rag/             # Retrieval-augmented generation
│   ├── __init__.py
│   └── ingest.py
└── tools/           # Function calling tools
    ├── __init__.py
    └── search.py

Providers

Each AI provider is a plugin package. The plugin hands you the official SDK client — you don't learn a BionicJS-specific API.

Plugin	Provider	Default Model
@bionicjs/openai	OpenAI	gpt-4o
@bionicjs/anthropic	Anthropic	claude-sonnet-4-5
@bionicjs/google	Google Gemini	gemini-2.5-pro
@bionicjs/ollama	Ollama (local)	llama3.2

Calling from the web

Python functions are called from TypeScript through the generated RPC boundary. The types flow automatically — Python defines them, TypeScript consumes them.

import { chat } from "@bionicjs/core/server/ai";

export default defineEventHandler(async (event) => {
  const { message } = await readBody(event);
  const reply = await chat(message);
  return { reply };
});

No manual fetch. No serialization. Just a typed function call.