📖 Bible Chatbot using FastAPI, ChromaDB, and OpenAI

This is my first AI-powered Bible chatbot, built as an experiment to combine artificial intelligence with Scripture. The goal is to allow anyone to ask questions like:

“What does the Bible say about love?”
“Who was Moses?”
“Where is the fruit of the Spirit mentioned?”

…and get clear, contextual answers backed by actual Bible verses.

🚀 Why I Built It

I've always been passionate about creating useful digital experiences. With the rise of large language models, I saw an opportunity to build something meaningful—an AI assistant for spiritual guidance, Bible study, or simple curiosity.

I wanted to remove the friction of traditional Bible search and allow more natural, conversational interaction with biblical content.

⚙️ Tech Stack

• FastAPI – For building a lightweight and high-performance backend in Python.
• ChromaDB – As a vector database for semantic search of Bible texts.
• Sentence Transformers – To generate local embeddings from Bible text.
• OpenAI API – For generating contextual, natural responses.
• dotenv – To securely manage environment variables and API keys.
• Docker - To use a container for ChromaDB
• Next.js + Tailwind CSS – For building a clean and reactive frontend UI.

🧠 Architecture Overview

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
import chromadb
from chromadb.config import Settings
from sentence_transformers import SentenceTransformer
from openai_service import OpenAIService
import os
from dotenv import load_dotenv

1. Environment & Embedding Model

load_dotenv()
embedding_model = SentenceTransformer("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")

1. Initialize ChromaDB

chroma_client = chromadb.HttpClient(
    host="localhost",
    port=8000
)
collection = chroma_client.get_or_create_collection("bible_verses")

1. Store Embeddings

# Bible verses were loaded from a CSV file in the format:
---
book,chapter,verse,text
Genesis,1,1,In the beginning God created the heavens and the earth.
---

# Then vectorized and added to Chroma batch by batch:

embedding = embedding_model.encode(verse_text)
batch_size = 100
for i in range(0, len(docs), batch_size):
    batch_docs = docs[i:i+batch_size]
    batch_embeddings = embeddings[i:i+batch_size]
    batch_ids = [str(uuid.uuid4()) for _ in batch_docs]

    collection.add(
        ids=batch_ids,
        documents=batch_docs,
        embeddings=[e.tolist() for e in batch_embeddings],
        metadatas=[{ "book": book, "chapter": chap, "verse": verse }],
    )
    print(f"Adding batch {i} - {i+batch_size}")

1. Semantic Search + GPT Response

# When a user asks a question:

query = "What does the Bible say about forgiveness?"
query_embedding = embedding_model.encode([query])
results = collection.query(query_embeddings=[query_embedding], n_results=3)

Then a context is built from those verses and sent to OpenAI:

relevant_verse = results['documents'][0] if results['documents'] else []
prompt = f"Use the following biblical context to answer:\n{relevant_verse}\n\nQuestion: {query}"
response = chat.completions.create(
            model=self.model_name,
            messages=[
                {"role": "system", "content": "You are a helpful Bible assistant."},
                {"role": "user", "content": prompt}
            ],
            max_tokens=max_toks,
            temperature=float(os.getenv("OPENAI_TEMPERATURE", "0.7"))
       )
print("response", response)
# Extract the response
answer = response.choices[0].message.content.strip()

The openai_service.py file handles interaction with the OpenAI Chat API (gpt-4 or gpt-3.5-turbo).

💻 Frontend

The frontend was built with Next.js (App Router) and styled with Tailwind CSS. It sends the user’s question to the FastAPI backend and displays the AI-generated response in real time.

🔍 Sample Interaction

User Input:

What does the Bible say about love?

Bot Response:

>> 1 Corinthians 13:4-5 — Love is patient, love is kind. It does not envy, it does not boast, it is not proud. It does not dishonor others, it is not self-seeking…

⚠️ Challenges

• Token limits with GPT-4 and how much biblical context to send.
• Ensuring accuracy and respect when handling sacred texts.
• Balancing semantic relevance vs literal match in ChromaDB.
• Avoiding AI “hallucinations” and anchoring responses in scripture only.

🔮 Next Steps

• Add support for multiple Bible versions (NIV, ESV, RVR60).
• Enable audio replies (TTS).
• Add study plan suggestions.
• Multilingual support.
• Deploy backend via Docker, Render, or Vercel Functions.

🙏 Final Thoughts

This project helped me bridge faith and technology in a way that feels personal and impactful. I learned a lot about vector databases, semantic search, and prompt engineering—and I hope this tool can help others in their spiritual journey.