nexithium-agent-platform

Nexithium Agent Platform Documentation

Welcome to the official documentation for Nexithium Agent Platform, your go‑to framework for building, deploying, and extending AI agents.

Overview
Getting Started
Architecture
Using the CLI
Telegram Bot Interface
REST API
Writing New Agents
Registering Custom Tools
Memory Management
Contributing
License
Whitepaper

Overview

Nexithium Agent Platform is a modular Python framework that lets you:

Create LLM‑driven agents with custom system prompts
Integrate external APIs (CoinGecko, Google Search, Tavily, etc.) as “tools”
Manage conversation memory (short‑term or persistent)
Expose agents via CLI, Telegram, or REST API (FastAPI)

Use it for crypto analysis, customer support, research assistants, and more.

Getting Started

Prerequisites

Python 3.10+
Git
API keys for OpenAI and any desired tools (CoinGecko doesn’t require a key)

Installation

git clone https://github.com/nexithium/nexithium-agent-platform.git
cd nexithium-agent-platform
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Configuration

Copy the example and fill in your environment variables:

cp config.env.example .env

Edit .env:

OPENAI_API_KEY=sk-...
TELEGRAM_BOT_TOKEN=12345:abcde
NEXITHIUM_API_KEY=your-secret-key
SERPER_API_KEY=your-serper-api-key
# REDIS_URL=redis://:pass@host:6379/0

Architecture

Core Components

Agent (core/agent.py): runs the LLM, manages messages, and calls tools
Memory (core/memory.py): short‑term buffer and persistent JSON/Redis storage
Tools (core/tools.py): registry for API wrappers and custom functions

Tools

Located in tools/, each tool is a simple function registered via @register_tool:

coingecko.get_price — fetches USD price
token_trend — analyzes 2‑day trend
token_description — retrieves project description
token_market_data — market cap & volume
google_search — web search via Serper.dev
tavily_news — news summary fallback

Interfaces

CLI (interfaces/cli.py) — interactive terminal interface
Telegram (interfaces/telegram_bot.py) — bot with inline menu & commands
REST API (interfaces/fastapi_server.py) — /chat and /tools endpoints

Using the CLI

python interfaces/cli.py --memory short --model gpt-4

Commands:

help — show commands
tools — list available tools
use <tool> [args] — invoke a specific tool
free text — chat with the AI agent
exit/quit — stop

Telegram Bot Interface

Start the bot:

python interfaces/telegram_bot.py

Commands:

/start — show menu
/help — usage guide
/tools — list tools
price <symbol>, trend <symbol>, analyze <symbol>, forecast <symbol>
free chat — general questions

REST API

Run server:

uvicorn interfaces.fastapi_server:app --reload --port 8000

Endpoints:

GET / — health check
GET /tools — list tools (requires X-API-Key)
POST /chat — agent interaction (requires X-API-Key)

Use Swagger UI at /docs.

Writing New Agents

Create an agent in agents/:

from core.agent import Agent
from core.tools import load_all_tools
from core.memory import ShortTermMemory

my_agent = Agent(
    name="MyCustomAgent",
    system_prompt="You are...",
    tools=list(load_all_tools().values()),
    memory=ShortTermMemory().get()
)

Registering Custom Tools

In tools/, create a new file with a function:

# tools/my_tool.py
def my_tool(arg1: str, arg2: int) -> str:
    return f"Received {arg1} and {arg2}"

Then register in core/tools.py:

from tools.my_tool import my_tool
register_tool("my_tool")(my_tool)

Memory Management

ShortTermMemory: retains last N exchanges in memory buffer
PersistentMemory: saves full conversation per user in memory_logs/
Access via MemoryManager.get_memory(user_id)

Contributing

See CONTRIBUTING.md for guidelines on coding standards, testing, and pull requests.

License

This project is licensed under the MIT License. See LICENSE for details.

Whitepaper

Read the full project whitepaper here.

Documentation generated by Nexithium Agent Platform.

This site is open source. Improve this page.