Agentic Chunking

Agentic chunking is an intelligent method of splitting documents into smaller chunks by using a model to determine natural breakpoints in the text. Rather than splitting text at fixed character counts, it analyzes the content to find semantically meaningful boundaries like paragraph breaks and topic transitions.

Create a Python file

agentic_chunking.py

import asyncio
from agno.agent import Agent
from agno.knowledge.chunking.agentic import AgenticChunking
from agno.knowledge.knowledge import Knowledge
from agno.knowledge.reader.pdf_reader import PDFReader
from agno.vectordb.pgvector import PgVector

db_url = "postgresql+psycopg://ai:ai@localhost:5532/ai"

knowledge = Knowledge(
    vector_db=PgVector(table_name="recipes_agentic_chunking", db_url=db_url),
)

asyncio.run(knowledge.ainsert(
    url="https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf",
    reader=PDFReader(
        name="Agentic Chunking Reader",
        chunking_strategy=AgenticChunking(),
    ),
))

agent = Agent(
    knowledge=knowledge,
    search_knowledge=True,
)

agent.print_response("How to make Thai curry?", markdown=True)

Set up your virtual environment

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

Install dependencies

uv pip install -U agno sqlalchemy psycopg pgvector

Run PgVector

docker run -d \
  -e POSTGRES_DB=ai \
  -e POSTGRES_USER=ai \
  -e POSTGRES_PASSWORD=ai \
  -e PGDATA=/var/lib/postgresql/data/pgdata \
  -v pgvolume:/var/lib/postgresql/data \
  -p 5532:5432 \
  --name pgvector \
  agno/pgvector:16

Run the script

python agentic_chunking.py

Custom Prompts

AgenticChunking(
    custom_prompt="Split at major section boundaries. Keep complete clauses together.",
    max_chunk_size=3000,
)

Custom prompts override default chunking behavior and are prioritized over default instructions.

Best Practices:

Always set max_chunk_size when using custom_prompt.
Focus custom_prompt on chunking logic only.
The default instructions automatically handle the output format constraints.

See Agentic Chunking with Custom Prompt for a complete example.