Building Production RAG Pipelines with LangChain and PostgreSQL pgvector

Retrieval-Augmented Generation (RAG) is the go-to architecture for building LLM applications that need to reference private or up-to-date documents. Instead of baking all knowledge into model weights, RAG retrieves relevant document chunks at query time and passes them as context to the LLM.

Core RAG Architecture

A RAG pipeline has three main phases: (1) Ingestion—documents are chunked, embedded using a model like text-embedding-ada-002, and stored in a vector database. (2) Retrieval—user queries are embedded, and a similarity search returns the top-k relevant chunks. (3) Generation—chunks are injected into an LLM prompt as context, and the model generates a grounded response.

Setting Up pgvector in PostgreSQL

CREATE EXTENSION IF NOT EXISTS vector;
CREATE TABLE document_embeddings (
  id SERIAL PRIMARY KEY,
  content TEXT,
  embedding VECTOR(1536),
  metadata JSONB
);
CREATE INDEX ON document_embeddings USING ivfflat (embedding vector_cosine_ops);

LangChain Integration

LangChain's PGVector vectorstore class wraps all ingestion and retrieval operations. Combined with RetrievalQAChain, you get end-to-end RAG in fewer than 30 lines of Python. Use async retrieval with ainvoke for latency-critical applications.

Production pgvector Semantic Search Query

Here is an enterprise SQL query and corresponding Python connection logic utilizing pgvector to perform high-speed cosine similarity searches across document embeddings:

import numpy as np
from psycopg2.extras import execute_values

def query_semantic_search(db_conn, query_embedding: list, limit=5):
    # Cosine similarity operator '<=>' computes distance
    sql = """
        SELECT slug, title, summary, 1 - (embedding <=> %s::vector) AS similarity
        FROM posts
        ORDER BY embedding <=> %s::vector
        LIMIT %s;
    """
    embedding_array = np.array(query_embedding, dtype=np.float32)
    with db_conn.cursor() as cursor:
        cursor.execute(sql, (embedding_array, embedding_array, limit))
        results = cursor.fetchall()
        return [{
            'slug': row[0], 'title': row[1],
            'summary': row[2], 'similarity': float(row[3])
        } for row in results]

Model Performance & Retrieval Profiles

Below is the performance comparison profile for our processing pipeline tested in staging against sanitized validation datasets:

US & UK Regulatory Standards for Artificial Intelligence

Deploying machine learning models in the US and UK markets requires strict alignment with local regulatory frameworks. In the United States, applications must respect the guidelines set by the FTC regarding algorithmic transparency, alongside the Executive Order on Safe, Secure, and Trustworthy AI. In the United Kingdom, AI systems must comply with the UK General Data Protection Regulation (UK GDPR), which enforces strict rules on automated profiling (under Article 22). Conducting bias auditing and maintaining explainable decision paths is critical to avoiding compliance sanctions in both jurisdictions.