Loading repository data…
CyrilFrancois / repository
Using an Agentic Microservices Architecture, this is a showcase of a simple LLM (chatGPT) + RAG (NeoJ) + MiddleBack/ETL (Python) about requesting in natural langage hospitals data.
A transparent discovery signal based on current public GitHub metadata.
This score does not audit code, security, maintainers, documentation quality, or suitability. Verify the repository and its current documentation before adoption.
Hosto-bot is an advanced Agentic Microservices demonstration combining Large Language Models (LLMs) with a Graph Database (Neo4j) to deliver a powerful Hybrid RAG (Retrieval-Augmented Generation) system.
Traditional RAG systems often struggle with structured aggregation and relationship-heavy queries. Hosto-bot addresses this limitation using a Router Agent that dynamically selects the best tool:
Healthcare data is inherently fragmented:
Hosto-bot provides a unified natural language interface allowing users (administrators, patients) to query across all data types without needing SQL, Cypher, or Python.
The system follows a decoupled microservices architecture.
/hospital_neo4j_etl)Responsible for building the data foundation:
(:Physician)-[:TREATS]->(:Visit))Review nodes/chatbot_api)Core intelligence built with FastAPI + LangChain.
Router Agent:
create_openai_functions_agentSelected from shared topics, language and repository description—not editorial ratings.
assafelovic /
An autonomous agent that conducts deep research on any data using any LLM providers
97/100 healthdeepnote /
Deepnote is a drop-in replacement for Jupyter with an AI-first design, sleek UI, new blocks, and native data integrations. Use Python, R, and SQL locally in your favorite IDE, then scale to Deepnote cloud for real-time collaboration, Deepnote agent, and deployable data apps. https://deepnote.com/
92/100 healthHybrid Querying:
Error Handling:
/chatbot_frontend)Built with Streamlit.
LLM_RAG_ETL_Showcase-Hostobot/
├── .env # Critical environment configurations
├── docker-compose.yml # Orchestration for all 3 services
├── requirements.txt # Root dependencies (mostly for local dev)
│
├── img/ # Images to display in the Readme
│ ├── interface.png
│ ├── dbscheme.png
│
├── hospital_neo4j_etl/ # DATA LAYER
│ ├── src/
│ │ ├── hospital_bulk_csv_write.py # Bulk loader & relationship mapper
│ │ └── entrypoint.sh # Execution script
│ └── Dockerfile
│
├── chatbot_api/ # LOGIC LAYER (FastAPI)
│ ├── src/
│ │ ├── main.py # API entry point & routes
│ │ ├── agents/
│ │ │ └── hospital_rag_agent.py # ReAct Agent & Tool definitions
│ │ ├── chains/
│ │ │ ├── hospital_cypher_chain.py # Cypher generation logic
│ │ │ └── hospital_review_chain.py # Vector search logic
│ │ ├── tools/
│ │ │ └── wait_times.py # Simulated real-time tool
│ │ ├── models/
│ │ │ └── hospital_rag_query.py # Pydantic schemas
│ │ └── utils/
│ │ └── async_utils.py # Retry decorators
│ └── Dockerfile
│
└── chatbot_frontend/ # PRESENTATION LAYER
├── src/
│ └── main.py # Streamlit UI logic
└── Dockerfile
The Neo4j database is modeled to handle multi-hop questions.
{id, name, state_name}{id, name}{id, name, dob, school, salary}{id, name, sex, blood_type}{id, room_number, admission_type, status, diagnosis}{id, text, patient_name, physician_name}(Patient)-[:HAS]->(Visit)(Physician)-[:TREATS]->(Visit)(Visit)-[:AT]->(Hospital)(Visit)-[:COVERED_BY]->(Payer)(Visit)-[:WRITES]->(Review)(Hospital)-[:EMPLOYS]->(Physician)Create a .env file in the root directory:
# API Keys
OPENAI_API_KEY=sk-...
# Neo4j Database (AuraDB or Local)
NEO4J_URI=neo4j+s://...
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD=...
NEO4J_DATABASE=neo4j
# Data Source URLs
HOSPITALS_CSV_PATH=https://.../hospitals.csv
# ... (Fill in paths for Payers, Physicians, Patients, Visits, Reviews)
# Model Settings
HOSPITAL_AGENT_MODEL=gpt-3.5-turbo-1106
HOSPITAL_CYPHER_MODEL=gpt-3.5-turbo-1106
HOSPITAL_QA_MODEL=gpt-3.5-turbo-0125
# Internal Networking
CHATBOT_URL=http://chatbot_api:8000/hospital-rag-agent
### 2. Launch with Docker
Ensure Docker Desktop is running, then execute:
```bash
docker-compose up --build
This command builds the images and starts the containers in the correct order:
| Type | Query |
|---|---|
| Quantitative | "What was the total amount billed to each payer during 2023?" |
| Qualitative | "How are patients describing the nursing team at Castaneda-Hardy?" |
| Real-time | "What’s the current wait time at Wallace-Hamilton Hospital?" |
| Comparative | "Which doctor has the shortest average visit duration?" |
arkaloscom /
Arkalos is an easy-to-use Python framework for data analysis, building data apps, warehouses, AI agents, and training LLMs with elegant syntax. It just works.
67/100 healthbrightdata /
AI-powered agent that scrapes leads with Bright Data, qualifies them using OpenAI, and delivers outreach-ready results via an easy Streamlit UI.
73/100 healthdeepeshmittal /
Implemented various AI algorithms in Pac-Man projects developed by UC Berkeley. Implemented informed/blind state-space search using search algorithms like BFS, DFS, UCS and A* algorithm with heuristic calculation. Designed an algorithm for reflex agent, minimax and alpha-beta pruning. Reinforcement Learning using MDP (Value Iteration and Policy Evaluation) - Implemented Markov Decision process using Value iteration in the Pacman World. Designed perceptron classifier for data classification. Developed applications using python. http://ai.berkeley.edu/project_overview.html
55/100 healthjohnsonhk88 /
Use LLM/AI agent for Web scraping (collection data) and analysis data with deep research
59/100 health