AthenaAgent

Crafting LLM with User-Customizable, Evolvable AI Agents

AthenaAgent is a TypeScript port of GeneralAgent by CosmosShadow, which aims to build an intuitive framework designed for dynamic adaptation. Unlike others, GeneralAgent specializes in equipping users with intelligent agents that are:

• capable of self-evolving through interpretable skills;
• with customizable and dynamic interfaces.

It's more than a tool, it's a companion that grows with users' needs, making AI's advanced capabilities accessible and operational for everyone.

Refer to GeneralAgent/README for more details about the agent framework.

Features

• Simple, fast, stable: support Azure OpenAI models: GPT3.5-turbo, GPT4, GPT4-turbo
• Build-in interpreters: JavaScript(built-in vm context)

Coming Soon

• functions and json_mode support for more stable and consistent response
• Interpreters: File, Shell, AppleScript, Retrieve Embedding, etc.
• Skill Set: txt-to-img(DALLE-3), txt-to-voice
• Memory: File Memory, Compressed Memory, Link Memory, etc.
• UI: Command Line Interface, Dynamic UI (a cross-platform native UI), Web UI

Status ?‍

This project is now under active development with daily updates.

Installation

yarn install

The following environment variables are required for LLM inferrence:

### reference: https://learn.microsoft.com/en-us/azure/ai-services/openai/reference
# GPT4 (for LLM ChatCompletion inference
GPT4_URL=https://{resource-name}.openai.azure.com/openai/deployments/{deployment-name}/chat/completions?api-version={api-version}
GPT4_APIKEY={api-key}

# GPT4-Turbo (for LLM ChatCompletion inference. It's not stable due to heavy traffic, not recommended. encountering too many empty responses)
GPT4_T_URL=https://{resource-name}.openai.azure.com/openai/deployments/{deployment-name}/chat/completions?api-version={api-version}
GPT4_T_APIKEY={api-key}

# GPT3.5-Turbo (for LLM ChatCompletion inference)
GPT3_5_URL=https://{resource-name}.openai.azure.com/openai/deployments/{deployment-name}/chat/completions?api-version={api-version}
GPT3_5_APIKEY={api-key}

# GPT3.5-Turbo-Instruct (for LLM Completion inference)
GPT3_5_I_APIKEY={api-key}
GPT3_5_I_URL=https://{resource-name}.openai.azure.com/openai/deployments/{deployment-name}/completions?api-version={api-version}