Document Chunking

DocumentChunking packs double-newline-separated paragraphs toward chunk_size. It splits an oversized paragraph at sentence boundaries.

Create a Python file

document_chunking.py

from agno.agent import Agent
from agno.knowledge.chunking.document import DocumentChunking
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_document_chunking", db_url=db_url),
)

knowledge.insert(
    url="https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf",
    reader=PDFReader(
        name="Document Chunking Reader",
        split_on_pages=False,
        chunking_strategy=DocumentChunking(),
    ),
)

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

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

Set up your virtual environment

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

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

Install dependencies

uv pip install -U agno sqlalchemy psycopg pgvector pypdf openai

Export your OpenAI API key

Set OpenAI Key

Set your OPENAI_API_KEY as an environment variable. You can get one from OpenAI.

export OPENAI_API_KEY=sk-***

setx OPENAI_API_KEY sk-***

Run PgVector

docker run -d \
  -e POSTGRES_DB=ai \
  -e POSTGRES_USER=ai \
  -e POSTGRES_PASSWORD=ai \
  -e PGDATA=/var/lib/postgresql \
  -v pgvolume:/var/lib/postgresql \
  -p 5532:5432 \
  --name pgvector \
  agnohq/pgvector:18

docker run -d `
  -e POSTGRES_DB=ai `
  -e POSTGRES_USER=ai `
  -e POSTGRES_PASSWORD=ai `
  -e PGDATA=/var/lib/postgresql `
  -v pgvolume:/var/lib/postgresql `
  -p 5532:5432 `
  --name pgvector `
  agnohq/pgvector:18

Run the script

python document_chunking.py

The example sets split_on_pages=False so PDFReader combines the pages before applying DocumentChunking. Keep the default value of True to chunk each page independently.

Chunk Size

chunk_size is a target rather than a strict ceiling. Paragraph separators, a sentence longer than the target, and overlap can produce a longer chunk. Overlap is prepended without an added separator.