Bee Agent Framework

Open-source framework for building, deploying, and serving powerful agentic workflows at scale.

The Bee Agent Framework makes it easy to build scalable agent-based workflows with your model of choice. Our default agent, Bee, is designed to perform robustly with Llama 3.1, and we're actively working on optimizing its performance with other popular LLMs.

Our goal is to empower developers to adopt the latest open-source and proprietary models with minimal changes to their current agent implementation.

Key Features

• 🤖 AI agents: Use our powerful Bee agent or build your own.
• 🛠️ Tools: Use our built-in tools or create your own in Javascript/Python.
• 👩‍💻 Code interpreter: Run code safely in a sandbox container.
• 💾 Memory: Multiple strategies to optimize token spend.
• ⏸️ Serialization Handle complex agentic workflows and easily pause/resume them without losing state.
• 🔍 Traceability: Get full visibility of your agent’s inner workings, log all running events, and use our MLflow integration (coming soon) to debug performance.
• 🎛️ Production-level control with caching and error handling.
• 🚧 (Coming soon) API: Configure and deploy your agents with a production-hardened API.
• 🚧 (Coming soon) Chat UI: Serve your agent to users in a delightful GUI with built-in transparency, explainability, and user controls.
• ... more on our Roadmap

Get started with Bee

Installation

npm install bee-agent-framework

or

yarn add bee-agent-framework

Example

import { BeeAgent } from "bee-agent-framework/agents/bee/agent";
import { OllamaChatLLM } from "bee-agent-framework/adapters/ollama/chat";
import { TokenMemory } from "bee-agent-framework/memory/tokenMemory";
import { DuckDuckGoSearchTool } from "bee-agent-framework/tools/search/duckDuckGoSearch";
import { OpenMeteoTool } from "bee-agent-framework/tools/weather/openMeteo";

const llm = new OllamaChatLLM(); // default is llama3.1 (8B), it is recommended to use 70B model

const agent = new BeeAgent({
  llm, // for more explore 'bee-agent-framework/adapters'
  memory: new TokenMemory({ llm }), // for more explore 'bee-agent-framework/memory'
  tools: [new DuckDuckGoSearchTool(), new OpenMeteoTool()], // for more explore 'bee-agent-framework/tools'
});

const response = await agent
  .run({ prompt: "What's the current weather in Las Vegas?" })
  .observe((emitter) => {
    emitter.on("update", async ({ data, update, meta }) => {
      console.log(`Agent (${update.key}) 🤖 : `, update.value);
    });
  });

console.log(`Agent 🤖 : `, response.result.text);

To run this example, be sure that you have installed ollama with the llama3.1 model downloaded.

➡️ See a more advanced example.

➡️ All examples can be found in the examples directory.

➡️ To run an arbitrary example, use the following command yarn start -- examples/agents/bee.ts (just pass the appropriate path to the desired example).

Local Installation (Python Interpreter + Interactive CLI)

Note: yarn should be installed via Corepack (tutorial)

Note: To make any asset available to a local code interpreter place them the following directory: ./examples/tmp/local

Note: Docker distribution with support for compose is required, the following are supported:

• Rancher - recommended
• Docker
• Podman - requires compose and rootful machine (if your current machine is rootless, please create a new one)

1. Clone the repository git clone git@github.com:i-am-bee/bee-agent-framework.
2. Install dependencies yarn install.
3. Create .env (from .env.template) and fill in missing values (if any).
4. Start the code interpreter yarn run infra:start-code-interpreter.
5. Start the agent yarn run start:bee (it runs ./examples/agents/bee.ts file).

🛠️ Tools

Name	Description
PythonTool	Run arbitrary Python code in the remote environment.
WikipediaTool	Search for data on Wikipedia.
GoogleSearchTool	Search for data on Google using Custom Search Engine.
DuckDuckGoTool	Search for data on DuckDuckGo.
SQLTool	Execute SQL queries against relational databases. Instructions.
CustomTool	Run your own Python function in the remote environment.
LLMTool	Use an LLM to process input data.
DynamicTool	Construct to create dynamic tools.
ArXivTool	Retrieve research articles published on arXiv.
WebCrawlerTool	Retrieve content of an arbitrary website.
CustomTool	Run your own Python function in the remote environment.
OpenMeteoTool	Retrieve current, previous, or upcoming weather for a given destination.
➕ Request

🔌️ Adapters (LLM - Inference providers)

Name	Description
Ollama	LLM + ChatLLM support (example)
OpenAI	LLM + ChatLLM support (example)
LangChain	Use any LLM that LangChain supports (example)
WatsonX	LLM + ChatLLM support (example)
Groq	ChatLLM support (example)
BAM (Internal)	LLM + ChatLLM support (example)
➕ Request

📦 Modules

The source directory (src) provides numerous modules that one can use.

Name	Description
agents	Base classes defining the common interface for agent.
llms	Base classes defining the common interface for text inference (standard or chat).
template	Prompt Templating system based on Mustache with various improvements_.
memory	Various types of memories to use with agent.
tools	Tools that an agent can use.
cache	Preset of different caching approaches that can be used together with tools.
errors	Base framework error classes used by each module.
adapters	Concrete implementations of given modules for different environments.
logger	Core component for logging all actions within the framework.
serializer	Core component for the ability to serialize/deserialize modules into the serialized format.
version	Constants representing the framework (e.g., latest version)
internals	Modules used by other modules within the framework.

To see more in-depth explanation see docs.

Tutorials

🚧 Coming soon 🚧

Roadmap

• MLFlow integration for trace observability

• Improvements to Bee agent and performance optimization with Llama3.1 and Granite model suites

• API

• Python SDK

• Structured outputs

• Chat Client (GUI)

• 🚧 TBD 🚧

Contribution guidelines

The Bee Agent Framework is an open-source project and we ❤️ contributions.

If you'd like to contribute to Bee, please take a look at our contribution guidelines.

Bugs

We are using GitHub Issues to manage our public bugs. We keep a close eye on this, so before filing a new issue, please check to make sure it hasn't already been logged.

Code of conduct

This project and everyone participating in it are governed by the Code of Conduct. By participating, you are expected to uphold this code. Please read the full text so that you can read which actions may or may not be tolerated.

Legal notice

All content in these repositories including code has been provided by IBM under the associated open source software license and IBM is under no obligation to provide enhancements, updates, or support. IBM developers produced this code as an open source project (not as an IBM product), and IBM makes no assertions as to the level of quality nor security, and will not be maintaining this code going forward.

Contributors

Special thanks to our contributors for helping us improve Bee Agent Framework.