Skip to main content
Build an AI assistant that answers questions using your own documentation. This example uses retrieval-augmented generation (RAG) to ensure accurate, documentation-grounded responses instead of hallucinations.

​
What You’ll Learn

By building this agent, you’ll understand:
  • How to create a vector database to store and search documentation
  • Why hybrid search (combining semantic and keyword matching) improves retrieval accuracy
  • How to maintain conversation history across multiple interactions
  • How to ensure agents search knowledge bases instead of relying solely on training data

​
Use Cases

Build documentation assistants, customer support agents, help desk systems, or educational tutors that need to reference specific knowledge bases.

​
How It Works

The agent uses retrieval-augmented generation (RAG) to answer questions:
  1. Search: Queries the vector database using hybrid search (semantic + keyword matching)
  2. Retrieve: Gets relevant documentation chunks from LanceDB
  3. Context: Combines retrieved docs with conversation history from SQLite
  4. Generate: LLM creates an answer grounded in the documentation
This ensures responses are based on your actual documentation, not just the model’s training data.

​
Code

agno_assist.py
import asyncio

from agno.agent import Agent
from agno.db.sqlite import SqliteDb
from agno.knowledge.embedder.openai import OpenAIEmbedder
from agno.knowledge.knowledge import Knowledge
from agno.models.openai import OpenAIChat
from agno.vectordb.lancedb import LanceDb, SearchType

# Create knowledge base with hybrid search
knowledge = Knowledge(
    vector_db=LanceDb(
        uri="tmp/lancedb",
        table_name="agno_assist_knowledge",
        search_type=SearchType.hybrid,  # Semantic + keyword search
        embedder=OpenAIEmbedder(id="text-embedding-3-small"),
    ),
)

# Load documentation asynchronously
asyncio.run(
    knowledge.add_content_async(
        name="Agno Docs", 
        url="https://docs.agno.com/llms-full.txt"
    )
)

# Create agent with knowledge and session persistence
agno_assist = Agent(
    name="Agno Assist",
    model=OpenAIChat(id="gpt-4o"),
    description="You help answer questions about the Agno framework.",
    instructions="Search your knowledge before answering the question.",  # Forces knowledge search
    knowledge=knowledge,
    db=SqliteDb(  # Stores conversation history
        session_table="agno_assist_sessions", 
        db_file="tmp/agents.db"
    ),
    add_history_to_context=True,
    add_datetime_to_context=True,
    markdown=True,
)

if __name__ == "__main__":
    agno_assist.print_response("What is Agno?")
    agno_assist.print_response("How do I create an agent with tools?")
    agno_assist.print_response("What vector databases does Agno support?")

​
What to Expect

The agent will answer questions by searching the Agno documentation. On the first run, it indexes the documentation into a local vector database (stored in tmp/ directory). Subsequent runs reuse the existing database for faster responses. Each answer is grounded in the actual documentation content, and the agent maintains conversation history so you can ask follow-up questions.

​
Usage

1

Create a virtual environment

Open the Terminal and create a python virtual environment.
python3 -m venv .venv
source .venv/bin/activate
2

Set your API key

export OPENAI_API_KEY=xxx
3

Install libraries

pip install -U agno openai lancedb tantivy sqlalchemy pandas
4

Run Agent

python agno_assist.py

​
Next Steps

  • Replace the URL in add_content_async() with your own documentation
  • Delete the tmp/ directory to reload with new content
  • Modify instructions to customize the agent’s behavior
  • Explore Knowledge Bases for advanced configuration