RecoVerse 🌌 — A Multi-Agent Recommendation System

RecoVerse is an intelligent, modular, and extensible multi-agent recommendation system built on the Camel framework. By integrating collaborative agents with OceanBase's hybrid transactional and vector database capabilities, it delivers personalized, explainable, and adaptive recommendations across diverse domains.

🚀 Overview

RecoVerse transforms traditional recommendation pipelines by introducing Multi-Agent Collaboration, where each agent is responsible for a specific cognitive task — from understanding user intent to reasoning over historical data and generating tailored suggestions.

✨ Key Features

• 🤖 Multi-Agent Architecture: Decentralized agents handle user modeling, content analysis, filtering, ranking, and explanation tasks.
• 🧠 Intent-Aware: Agents jointly infer both explicit and implicit user needs based on behavior, preferences, and feedback.
• 🔄 Pluggable Workflows: Easily extend or replace agents for domain-specific customization.
• 📊 Hybrid Recommendations: Supports collaborative filtering, content-based methods, and rule-augmented heuristics.
• 🧩 Explainability: Provide reasoning traces from agents for every recommendation.

---

🧭 System Workflow

---

📦 Installation

1. Clone this repository:

git clone https://github.com/Ol1ver0413/RecoVerse.git
cd RecoVerse

2. Install dependencies:

pip install -r requirements.txt

⚙️ Configuration

Prior to running the system, make sure to import the datasets into the database using the following two files:
recoverse\database\sqldb_store.py and recoverse\database\vectordb_store.py.

Before running the system, ensure your database and embedding model are correctly configured. Update main.py or your own script with the following:

db_config = {
    'host': '127.0.0.1',
    'port': 2881,
    'user': 'lyz',
    'password': '123qwe',
    'database': 'Yelp'
}

from recosystem.embedding import SentenceTransformerEncoder
embed_model = SentenceTransformerEncoder(model_name="/home/lyz/Rag/models/bge-m3")

Alternatively, test a single recommendation in an interactive script:

from recosystem.core import UserBusinessRecommender
from recosystem.demo import examples_demo

recommender = UserBusinessRecommender(
    db_config=db_config,
    api_key='',  # Optional, for external services if needed
    embed_model=embed_model
)

i = 0
result = recommender.recommend(username=examples_demo[i][1], query=examples_demo[i][0])
print("[推荐结果]：")
print(result)

🧪 Running the System

You can run the main pipeline by executing:

python main.py

---

🖼️ Example Results

Below are some example results generated by RecoVerse:

示例 1：基于用户隐式意图的个性化推荐

示例 2：结合历史行为的相似服务推荐

示例 3：可解释性推荐结果输出

---

🏆 Awards & Recognition

RecoVerse has been recognized in national-level competitions for its innovative architecture and multi-agent intelligence:

🏅 OceanBase × Hackathon 全国八强

🌟 OceanBase × Hackathon 全国八强

---

💡 Want to Contribute?

If you're interested in upgrading or extending RecoVerse — whether it's improving agent capabilities, optimizing retrieval, or expanding to new domains — feel free to open an Issue. I’ll be happy to respond promptly and collaborate on the future of intelligent recommendation systems!