How to Integrate LlamaIndex for retail banking with Supabase for RAG

Combining LlamaIndex for retail banking with Supabase gives you a clean RAG backend for customer support, policy lookup, and internal banking ops. LlamaIndex handles ingestion, chunking, retrieval, and query orchestration; Supabase gives you Postgres storage with auth, row-level security, and a simple vector store path for production.

For retail banking, that means you can answer questions from product PDFs, fee schedules, KYC policies, loan guides, and call-center playbooks without hardcoding logic into the agent.

Prerequisites

• •Python 3.10+
• •A Supabase project with:
  • •SUPABASE_URL
  • •SUPABASE_SERVICE_ROLE_KEY
• •A Postgres database in Supabase with the vector extension enabled
• •OpenAI API key or another embedding/model provider supported by LlamaIndex
• •Installed packages:
  • •llama-index
  • •llama-index-vector-stores-supabase
  • •supabase
  • •python-dotenv
• •Banking content ready for ingestion:
  • •PDFs
  • •DOCX files
  • •Markdown policy docs
  • •FAQ text

Install the core dependencies:

pip install llama-index llama-index-vector-stores-supabase supabase python-dotenv

Integration Steps

1. •Set up your environment and clients.

Use environment variables so your agent can run in dev, staging, and prod without code changes.

import os
from dotenv import load_dotenv

load_dotenv()

SUPABASE_URL = os.environ["SUPABASE_URL"]
SUPABASE_SERVICE_ROLE_KEY = os.environ["SUPABASE_SERVICE_ROLE_KEY"]
OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]

Initialize the Supabase client and LlamaIndex settings:

from supabase import create_client
from llama_index.core import Settings

supabase = create_client(SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY)

Settings.chunk_size = 512
Settings.chunk_overlap = 64

2. •Create or connect to a Supabase table for vectors.

LlamaIndex can write embeddings into Supabase through its vector store integration. The table must be ready before ingestion.

from supabase import create_client

supabase = create_client(SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY)

# Example: create the table manually in SQL editor first.
# Table name used below: banking_docs

table_name = "banking_docs"
print(f"Using Supabase table: {table_name}")

If you want to validate that the table exists from Python:

result = supabase.table("banking_docs").select("id").limit(1).execute()
print(result.data)

3. •Load retail banking documents and build nodes.

Use LlamaIndex readers to ingest policy documents, product guides, and knowledge base articles.

from llama_index.core import SimpleDirectoryReader

documents = SimpleDirectoryReader(
    input_dir="./banking_docs",
    recursive=True,
).load_data()

print(f"Loaded {len(documents)} documents")

Convert them into nodes so retrieval is chunk-based instead of document-based:

from llama_index.core.node_parser import SentenceSplitter

splitter = SentenceSplitter(chunk_size=512, chunk_overlap=64)
nodes = splitter.get_nodes_from_documents(documents)

print(f"Created {len(nodes)} nodes")

4. •Store embeddings in Supabase using LlamaIndex’s vector store adapter.

This is the key integration point. LlamaIndex generates embeddings; Supabase stores them in Postgres with vector search.

from llama_index.vector_stores.supabase import SupabaseVectorStore
from llama_index.core import StorageContext, VectorStoreIndex

vector_store = SupabaseVectorStore(
    postgres_connection_string=f"postgresql://postgres:{os.environ['SUPABASE_DB_PASSWORD']}@db.{os.environ['SUPABASE_PROJECT_REF']}.supabase.co:5432/postgres",
    collection_name="banking_docs",
)

storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex(nodes=nodes, storage_context=storage_context)

If you already have embeddings stored in Supabase and just want to reconnect later:

index = VectorStoreIndex.from_vector_store(vector_store=vector_store)

5. •Query the index from your AI agent.

Build a retriever-backed query engine that your agent can call during customer interactions.

query_engine = index.as_query_engine(similarity_top_k=3)

response = query_engine.query(
    "What documents do I need to open a joint savings account?"
)

print(response)

For retail banking workflows, add a stricter prompt layer so answers stay grounded in policy text:

from llama_index.core.prompts import PromptTemplate

qa_prompt = PromptTemplate(
    "You are a retail banking assistant. Answer only from the provided context.\n"
    "If the answer is not in context, say you don't know.\n\n"
    "Context:\n{context_str}\n\nQuestion:\n{query_str}\n\nAnswer:"
)

query_engine = index.as_query_engine(text_qa_template=qa_prompt)

Testing the Integration

Run a retrieval test against a known policy question. You want to confirm three things: ingestion worked, Supabase stored vectors, and LlamaIndex can retrieve relevant chunks.

test પ્રશ્ન? # invalid? 

Use this instead:

test_query_engine = index.as_query_engine(similarity_top_k=2)

result = test_query_engine.query("What is the fee for an international wire transfer?")
print(str(result))

Expected output should look like this:

The international wire transfer fee is $25 per transaction.
Additional correspondent bank fees may apply depending on destination.

If you get an empty or irrelevant answer:

• •check that embeddings were inserted into the correct Supabase table
• •verify your chunk size is not too large
• •confirm the docs actually contain the answer text
• •inspect RLS policies if queries work locally but fail in deployed environments

Real-World Use Cases

• •

  Customer support copilot

  • •Answer questions about account opening, debit card replacement, overdrafts, and wire transfers from approved banking content only.
• •

  Branch staff assistant

  • •Let employees query internal SOPs for KYC checks, AML escalation paths, and loan application requirements.
• •

  Product FAQ agent

  • •Power an AI assistant on your retail banking website that explains savings accounts, mortgage basics, credit card terms, and fee schedules with grounded answers.

Keep learning

• •The complete AI Agents Roadmap — my full 8-step breakdown
• •Free: The AI Agent Starter Kit — PDF checklist + starter code
• •Work with me — I build AI for banks and insurance companies

By Cyprian Aarons, AI Consultant at Topiax.