​
Code

cookbook/apps/playground/upload_files.py
from agno.agent import Agent
from agno.knowledge.combined import CombinedKnowledgeBase
from agno.knowledge.csv import CSVKnowledgeBase
from agno.knowledge.docx import DocxKnowledgeBase
from agno.knowledge.json import JSONKnowledgeBase
from agno.knowledge.pdf import PDFKnowledgeBase
from agno.knowledge.text import TextKnowledgeBase
from agno.models.google.gemini import Gemini
from agno.models.openai import OpenAIChat
from agno.playground import Playground
from agno.storage.postgres import PostgresStorage
from agno.vectordb.pgvector import PgVector

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

knowledge_base = CombinedKnowledgeBase(
    sources=[
        PDFKnowledgeBase(
            vector_db=PgVector(table_name="recipes_pdf", db_url=db_url), path=""
        ),
        CSVKnowledgeBase(
            vector_db=PgVector(table_name="recipes_csv", db_url=db_url), path=""
        ),
        DocxKnowledgeBase(
            vector_db=PgVector(table_name="recipes_docx", db_url=db_url), path=""
        ),
        JSONKnowledgeBase(
            vector_db=PgVector(table_name="recipes_json", db_url=db_url), path=""
        ),
        TextKnowledgeBase(
            vector_db=PgVector(table_name="recipes_text", db_url=db_url), path=""
        ),
    ],
    vector_db=PgVector(table_name="recipes_combined", db_url=db_url),
)

file_agent = Agent(
    name="File Upload Agent",
    agent_id="file-upload-agent",
    role="Answer questions about the uploaded files",
    model=OpenAIChat(id="gpt-4o-mini"),
    storage=PostgresStorage(
        table_name="agent_sessions", db_url=db_url, auto_upgrade_schema=True
    ),
    knowledge=knowledge_base,
    show_tool_calls=True,
    markdown=True,
)


audio_agent = Agent(
    name="Audio Understanding Agent",
    agent_id="audio-understanding-agent",
    role="Answer questions about audio files",
    model=OpenAIChat(id="gpt-4o-audio-preview"),
    storage=PostgresStorage(
        table_name="agent_sessions", db_url=db_url, auto_upgrade_schema=True
    ),
    add_history_to_messages=True,
    add_datetime_to_instructions=True,
    show_tool_calls=True,
    markdown=True,
)

video_agent = Agent(
    name="Video Understanding Agent",
    model=Gemini(id="gemini-2.0-flash"),
    agent_id="video-understanding-agent",
    role="Answer questions about video files",
    storage=PostgresStorage(
        table_name="agent_sessions", db_url=db_url, auto_upgrade_schema=True
    ),
    add_history_to_messages=True,
    add_datetime_to_instructions=True,
    show_tool_calls=True,
    markdown=True,
)

playground = Playground(
    agents=[file_agent, audio_agent, video_agent],
    name="Upload Files Playground",
    description="Upload files and ask questions about them",
    app_id="upload-files-playground",
)
app = playground.get_app()

if __name__ == "__main__":
    playground.serve(app="upload_files:app", reload=True)

​
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 keys

export OPENAI_API_KEY=xxx
export GOOGLE_API_KEY=xxx
3

Install libraries


pip install -U agno "uvicorn[standard]" openai google-generativeai psycopg-binary

pip install -U "agno[pdf,csv,docx,json,text]"
4

Run Agent

python cookbook/apps/playground/upload_files.py