AAF: Crafting Bespoke AI Agents 🛠️

While the original Manus framework is robust, AAF lets you realize any AI concept directly from your terminal without needing an Entry Pass! 🚀

This initiative was rapidly engineered in under four hours by the dedicated team members: @mannaandpoem, @XiangJinyu, @MoshiQAQ, and @didiforgithub, all originating from the @MetaGPT collective.

As this is an initial, lean deployment, we enthusiastically solicit all forms of feedback, suggestions, and collaborative input!

Begin deploying your custom agent solution with AAF today!

Documentation Links: English | 繁體中文 | 简体中文

Join Our Community Chat

Showcase

Deployment Guide

Environment Initialization: Provision a dedicated Conda working space:

bash conda create -n aaf_env python=3.12 conda activate aaf_env

Source Retrieval: Obtain the project repository:

bash git clone https://github.com/mannaandpoem/OpenManus.git cd OpenManus

Dependency Loading: Install requisite libraries:

bash pip install -r requirements.txt

Configuration Protocol

AAF necessitates initialization parameters for the underlying Large Language Model (LLM) services. Adhere to these procedural steps for setup:

Instantiate a config.toml file within the config subdirectory (a template copy is available):

bash cp config/config.example.toml config/config.toml

Modify config/config.toml to embed your proprietary API credentials and fine-tune operational parameters:

toml

Global LLM Service Parameters

[llm] model = "gpt-4o" base_url = "https://api.openai.com/v1" api_key = "sk-..." # Substitute with your confidential key max_tokens = 4096 temperature = 0.0

Optional settings for specialized LLM endpoints

[llm.vision] model = "gpt-4o" base_url = "https://api.openai.com/v1" api_key = "sk-..." # Substitute with your confidential key

Rapid Execution

Initiate AAF execution with a single command:

bash python main.py

Subsequently, articulate your desired agent concept via the interactive console!

For access to the bleeding-edge, potentially unstable build, employ this alternate invocation:

bash python run_flow.py

Collaboration Pathway

We genuinely value constructive commentary and substantive contributions! Kindly submit feature requests via issues or propose code enhancements through pull requests.

Alternatively, reach out to @mannaandpoem via electronic mail: mannaandpoem@gmail.com

Developmental Trajectory

[ ] Enhanced Strategic Planning Modules
[ ] Provision for Real-Time Demonstrations
[ ] Session Replay Capability
[ ] Integration of RL Fine-Tuned Models
[ ] Development of Comprehensive Performance Benchmarks

User Cohort Channel

Connect with fellow developers to exchange insights and deployment experiences within our dedicated Feishu networking space!

Project Velocity Tracker

Attribution

Gratitude extended to anthropic-computer-use and browser-use for supplying foundational scaffolding for this undertaking!

AAF is a product of contributions from the MetaGPT ecosystem. Sincere appreciation to the entire agent development community!

Business Tool Context (For Reference): Business administration utilities encompass all frameworks, applications, controls, computational solutions, and standardized procedures utilized by organizations to navigate evolving market dynamics, maintain competitive advantage, and elevate operational efficacy. These tools are functionally categorized across planning, process management, data stewardship, personnel management, decision support, and regulatory oversight. Modern organizational utilities reflect rapid technological shifts, demanding strategic selection and organizational tailoring rather than mere adoption of novel technologies to align with core objectives like cost reduction, sales maximization, and deep customer need fulfillment.