README.md

BeeAI Framework Examples

This repository contains examples demonstrating the usage of the BeeAI Framework, a toolkit for building AI agents and applications.

Table of Contents

1. Agents
2. Workflows
3. Cache
4. Errors
5. Helpers
6. LLMs (Language Models)
7. Logger
8. Memory
9. Serialization
10. Templates
11. Tools

Agents

• react.ts: Basic ReAct Agent implementation
• react_advanced.ts: Advanced ReAct Agent with custom configurations
• react_reusable.ts: Demonstration of serializing and reusing ReAct Agents
• custom_agent.ts: Example of creating a custom agent
• granite_react.ts: Basic ReAct Agent using an IBM Granite LLM
• granite_wiki_react.ts: Advanced ReAct Agent using an IBM Granite LLM with wikipedia retrieval
• simple.ts: Simple agent implementation
• sql.ts: Agent for SQL-related tasks

Workflows

• simple.ts: Introduction to workflows
• nesting.ts: How to nest workflows
• agent.ts: Using workflows to interconnect two agents with a critique step.
• multiAgents.ts: Multi-step sequential agentic workflow.
• contentCreator.ts: Multi-step workflow for writing blog posts.

Cache

• cacheFn.ts: Function caching example
• custom.ts: Custom cache implementation
• decoratorCache.ts: Cache decorator usage
• decoratorCacheComplex.ts: Complex cache decorator example
• fileCache.ts: File-based caching
• fileCacheCustomProvider.ts: Custom provider for file cache
• slidingCache.ts: Sliding window cache implementation
• toolCache.ts: Caching for tools
• unconstrainedCache.ts: Unconstrained cache example
• unconstrainedCacheFunction.ts: Function using unconstrained cache

Errors

• base.ts: Basic error handling
• cast.ts: Error casting example
• tool.ts: Tool-specific error handling

Helpers

• io.ts: Input/Output helpers
• setup.ts: Setup utilities

LLMs (Language Models)

• chat.ts: Chat-based language model usage
• chatCallback.ts: Callbacks for chat models
• structured.ts: Structured output from language models

LLM Providers

• groq.ts: Groq language model integration
• langchain.ts: LangChain integration
• ollama.ts: Ollama model usage
• openai.ts: OpenAI integration
• watsonx.ts: Watsonx integration
• anthropic.ts: Anthropic integration

Logger

• agent.ts: Agent-specific logging
• base.ts: Basic logging setup
• pino.ts: Pino logger integration

Memory

• agentMemory.ts: Memory management for agents
• custom.ts: Custom memory implementation
• llmMemory.ts: Memory for language models
• slidingMemory.ts: Sliding window memory
• summarizeMemory.ts: Memory with summarization
• tokenMemory.ts: Token-based memory
• unconstrainedMemory.ts: Unconstrained memory example

Serialization

• base.ts: Basic serialization
• context.ts: Context serialization
• customExternal.ts: Custom external serialization
• customInternal.ts: Custom internal serialization
• memory.ts: Memory serialization

Templates

• arrays.ts: Array-based templates
• forking.ts: Template forking
• functions.ts: Function-based templates
• objects.ts: Object-based templates
• primitives.ts: Primitive data type templates

Tools

• advanced.ts: Advanced tool usage
• agent.ts: Agent-specific tools
• base.ts: Basic tool implementation
• mcp.ts: MCP tool usage

Custom Tools

• base.ts: Custom tool base implementation

• dynamic.ts: Dynamic tool creation

• openLibrary.ts: OpenLibrary API tool

• python.ts: Python-based custom tool

• langchain.ts: LangChain tool integration

Usage

To run these examples, make sure you have the BeeAI Framework cloned and properly configured. Each file demonstrates a specific feature or use case of the framework. You can run individual examples using Node.js with TypeScript support.

1. Clone the repository:

   git clone git@github.com:i-am-bee/beeai-framework

2. Install dependencies:

   yarn install --immutable && yarn prepare

3. Create .env file (from .env.template) and fill in missing values (if any).

4. Run an arbitrary example, use the following command:

   yarn start examples/path/to/example.ts

For more detailed information on the BeeAI Framework, please refer to the documentation.

Tip

To run examples that use Ollama, be sure that you have installed Ollama with the llama3.1 model downloaded.