Traditional RAG with PgVector

This example demonstrates traditional RAG implementation using PgVector database with OpenAI embeddings, where knowledge context is automatically added to prompts without search functionality.

Code

1from kern.agent import Agent
2from kern.knowledge.embedder.openai import OpenAIEmbedder
3from kern.knowledge.knowledge import Knowledge
4from kern.models.openai import OpenAIResponses
5from kern.vectordb.pgvector import PgVector, SearchType
6
7db_url = "postgresql+psycopg://ai:ai@localhost:5532/ai"
8
9knowledge = Knowledge(
10    # Use PgVector as the vector database and store embeddings in the `ai.recipes` table
11    vector_db=PgVector(
12        table_name="recipes",
13        db_url=db_url,
14        search_type=SearchType.hybrid,
15        embedder=OpenAIEmbedder(id="text-embedding-3-small"),
16    ),
17)
18
19knowledge.insert(
20    url="https://kern-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
21)
22
23agent = Agent(
24    model=OpenAIResponses(id="gpt-5.2"),
25    knowledge=knowledge,
26    # Enable RAG by adding context from the `knowledge` to the user prompt.
27    add_knowledge_to_context=True,
28    # Set as False because Agents default to `search_knowledge=True`
29    search_knowledge=False,
30    markdown=True,
31)
32agent.print_response(
33    "How do I make chicken and galangal in coconut milk soup", stream=True
34)

Usage

Set up your virtual environment

1uv venv --python 3.12
2source .venv/bin/activate

1uv venv --python 3.12
2.venv\Scripts\activate

Install dependencies

1uv pip install -U kern-ai openai sqlalchemy psycopg pgvector

Setup PgVector

Start PostgreSQL with pgvector extension and update the connection string in the code as needed.

Export your OpenAI API key

1export OPENAI_API_KEY=your_openai_api_key_here

Run Agent

1python traditional_rag_pgvector.py