This project is an intelligent document management and retrieval-augmented generation (RAG) solution that combines ChromaDB’s efficient embedding storage with FastAPI’s high-performance capabilities. It supports document types such as PDF, DOCX, and TXT. The server allows users to ingest documents, extract their contents, generate embeddings, and query the stored documents efficiently.
Python 3.9 or higher FastAPI ChromaDB Sentence-Transformers Uvicorn (for running the server)
- Clone the repository:
git clone https://github.com/rajsingh16/ChromaServe-with-fastAPI- Install the dependences:
pip install -r requirements.txt
- Run the server: Start the FastAPI server with the following command:
uvicorn main:app --reloadThe server will run on http://127.0.0.1:8000.
POST /ingest
Ingest a document (PDF, DOCX, or TXT) and store its text and embedding in ChromaDB.
Request Content-Type: multipart/form-data File: The document file to ingest (PDF, DOCX, or TXT).
Example Request:
curl -X 'POST' \
'http://127.0.0.1:8000/ingest' \
-F '[email protected]'
Here’s a sample README.md for your FastAPI project that implements Retrieval-Augmented Generation (RAG) using ChromaDB and Hugging Face's SentenceTransformer:
FastAPI RAG Server with ChromaDB and Sentence-Transformers This project implements a lightweight FastAPI server for Retrieval-Augmented Generation (RAG) using ChromaDB's Persistent Client for ingesting and querying documents, and Sentence-Transformers from Hugging Face for embedding generation. It supports document types such as PDF, DOCX, and TXT. The server allows users to ingest documents, extract their contents, generate embeddings, and query the stored documents efficiently.
Document Ingestion: Upload and ingest documents in PDF, DOCX, and TXT formats. The text from the document is extracted and embedded using Sentence-Transformers. Querying: Send a user query to search through ingested documents using embeddings. ChromaDB: Persistent document storage with efficient querying and retrieval based on embeddings. Asynchronous Processing: Non-blocking endpoints with support for concurrent requests using FastAPI and Python’s asyncio.
Python 3.9 or higher FastAPI ChromaDB Sentence-Transformers Uvicorn (for running the server) You can install the required dependencies using pip:
pip install fastapi chromadb sentence-transformers uvicornFor processing PDF and DOCX files, the following libraries are used:
PyMuPDF (for PDF text extraction) python-docx (for DOCX text extraction) Install these dependencies as well:
pip install pymupdf python-docxThe server will run on http://127.0.0.1:8000.
Endpoints POST /ingest Ingest a document (PDF, DOCX, or TXT) and store its text and embedding in ChromaDB.
Request Content-Type: multipart/form-data File: The document file to ingest (PDF, DOCX, or TXT). Example Request:
curl -X 'POST' \
'http://127.0.0.1:8000/ingest' \
-F '[email protected]'Response Status Code: 201 (Created) Body: { "message": "Document ingested successfully" } GET /query Query ingested documents using a user query string. This endpoint uses the embeddings generated from the query to retrieve relevant documents.
Request Query Parameters: user_query (string) — The query string. Example Request:
curl -X 'GET' \
'http://127.0.0.1:8000/query?user_query=your+search+query'Response Status Code: 200 (OK) Body: A JSON object containing the query results. Example:
{
"results": [
{
"document_id": "1",
"score": 0.89,
"content": "Extracted content of the matching document..."
}
]
}FastAPI: For building the lightweight and high-performance web API. ChromaDB: For persistent document storage and efficient query handling based on embeddings. Sentence-Transformers: For generating text embeddings to represent the content of documents and queries. Uvicorn: ASGI server to run the FastAPI app. PyMuPDF (fitz): For extracting text from PDF files. python-docx: For extracting text from DOCX files.
This project is licensed under the MIT License - see the LICENSE file for details.