BafCode Framework Documentation

Overview 🌐

BafCode is a powerful Flask framework designed for the swift development of AI agents. Its robust features and simple setup make it an essential tool for developers looking to dive into AI solution development without the fuss.

Features 🌟

Auto Project Setup: Get started quickly with an organized project structure.

Docker Option: Use Docker for a more streamlined development environment.

BafCode CLI: Speed up agent tools creation with our dedicated command-line interface.

Focus on Agent Tools: Spend time creating the tools, not setting up the framework.

Automated Processes: Auto-generates and imports tool APIs and prompt files, so you can focus on agent tool logic.

Why Choose BafCode? 🤖

With BafCode, you can solely concentrate on agent tool development. The framework takes care of generating and importing essential components like prompt and API for you, leading to a more efficient development process.

Requirements 📋

Python Environment

Optional: Docker for containerization

Command-Line Interface (for BafCode CLI commands, install it using pip install bafcode)

Core Components ⚙️

Manager:

  1. Generates a Task list.
  2. Sends a single task to the agent.
  3. Receives task results.
  4. Saves the task result in a session with status (either Completed or Failed).
  5. Removes completed task from the task list and sends the next task with "pending" status to the agent.

Agent:

  1. Receives a task.
  2. Decides which tool to use based on the received task.
  3. Uses the tool to process the task.
  4. Sends the tool results to the manager.

Responder:

  1. Generates a good response based on the tool results.
  2. Returns the response to the application.
  3. Receives the final message from the manager.

Tools:

  1. Tools are used by the Agent to perform a specific Task.
  2. You can generally place any action function inside the tool, and it will be executed once the Agent decides to use it.
  3. The tool can utilize external APIs by coding the API logic and invoking the imported API process function in any tool file you produce.

Prompts:

  1. Prompts are cues used by LLMS.
  2. You can use prompt files wherever LLMS is utilized.
  3. You can modify default prompts used by the main components of the framework.
  4. You can also tailor default prompts employed by the central elements of the framework.
  5. It is possible to create a unique prompt for every tool or API.

LLMS:

  1. You can use Multiple LLMs, each LLM for a specific case.
  2. Can be used in the APIs to generate a like humane text. e.g to be saved in a database or posted in X tweet...
  3. Can be used by the Responder to generate a clean response based on the task list results provided by the Manager Component.
  4. Can be used by the agent to decide which tool to use based on the task provided by the Manager Component.
  5. Can be used by Manager to generate Task List based on the user requirements.
  6. Used By Other Components to generate a Humanistic response.

How to Build an AI Agent using BafCode 🔧

Initial Setup 🚀 Use the BafCode CLI to kickstart your project.

pip install bafcode
bafcode setup <project_name>
cd <project_name>/src

Develop Your Tool 🛠 Begin by crafting your desired tool. Initialize a file within the tools directory. For larger agents, categorize related tools into subdirectories.

bafcode make tool <your_tool_name>

Once your tool is created, related APIs and prompts are auto-generated and imported.

Setup LLM 🧠: Set up the brain behind your agent and responder with ease using BafCode CLI.

By default, the agent and the responder components are using openai_llm, you just need to add your OPENAI KEYS in .env

If you are interested to use any other llm, you can make one using our BafCode CLI

bafcode make llm <your_llm_name>

The LLM file will be created in the llms directory. You're welcome to incorporate your desired LLM logic into it.

Agent Running 🚀 With all components ready, launch the agent.

Docker Configuration (Optional) 🐳 For Docker enthusiasts, use Docker and docker-compose for easier framework deployment.

docker-compose up -d

Manual Initialization 📂 Prefer a hands-on approach? Navigate to the src directory and start the framework.

cd src
bafcode start or
python app.py

For testing, employ Postman to send a POST request to the /generate endpoint with your message.

Request Examples 📬 To interact with your AI agent, you can send a POST request to the /generate endpoint. Here are some examples:

Single Mode In single mode, the agent processes one task at a time. Here's an example of a request in single mode:

{  "message":"Your message here",    "mode": "single",}

Multiple Mode In Multiple Mode, the manager creates a list of tasks and begins to dispatch them individually to the agent. Below is an example of a request in multiple mode:

{  "message":"Your message here",    "mode": "multiple",}

Additional Notes 📝 When developing tools, often you won't need to alter the tool logic. The bafcode make command takes care of importing and generating essentials like prompts and APIs.

LLMs currently need manual configuration. Visit config/llms_config.py and replace the DEFAULT_LLM with your LLM Class name.

Congratulations, you are now empowered to efficiently craft AI agents using the BafCode Framework!

Contributing to BafCode

We welcome contributions to the BafCode framework! Here's a quick guide:

  1. Understand the Philosophy: BafCode emphasizes enabling developers to focus on agent functionalities, not just the components.
  2. Fork & Branch: Fork the repository and create a new branch for your changes.
  3. Commit: Make sure to write clear and meaningful commit messages.
  4. Pull Request: Submit a PR from your forked repository to the main BafCode repository.
  5. Review & Feedback: Await review from maintainers and be open to feedback.

For detailed guidelines, please see our Contribution Guide.

Your contributions help improve BafCode for all users. Thank you! 🚀🌐

Table of Contents