Google Audio Input Local File Upload
Send a local MP3 to Gemini with Audio(filepath) and stream the analysis.
The current local-audio helper checks for a SUCCESS state, so it re-uploads an existing ACTIVE file instead of reusing it. Use the current adaptation below instead of running the source snapshot unchanged.
"""
Google Audio Input Local File Upload
====================================
Cookbook example for `google/gemini/audio_input_local_file_upload.py`.
"""
from pathlib import Path
from agno.agent import Agent
from agno.media import Audio
from agno.models.google import Gemini
# ---------------------------------------------------------------------------
# Create Agent
# ---------------------------------------------------------------------------
agent = Agent(
model=Gemini(id="gemini-3.7-flash"),
markdown=True,
)
# Please download a sample audio file to test this Agent and upload using:
audio_path = Path(__file__).parent.joinpath("sample.mp3")
agent.print_response(
"Tell me about this audio",
audio=[Audio(filepath=audio_path)],
stream=True,
)
# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------
if __name__ == "__main__":
passCurrent adaptation
Place sample.mp3 beside this script. You can set GOOGLE_FILE_NAME to a previously uploaded file's exact files/... name to reuse it; otherwise this example uploads and removes its own file. Both queries complete before cleanup.
Save this helper as google_files.py beside the runnable example. It accepts a local file or an existing Files API name, waits up to five minutes for ACTIVE, and rejects failed or incomplete uploads. It deletes only files it uploaded itself; an existing file remains owned by its caller. Developer API uploads otherwise expire after 48 hours.
from contextlib import contextmanager
from pathlib import Path
from time import monotonic, sleep
@contextmanager
def ready_file(client, path: Path, existing_name: str | None = None):
existing_name = existing_name or None
if existing_name is None and not path.is_file():
raise FileNotFoundError(path)
uploaded = (
client.files.get(name=existing_name)
if existing_name
else client.files.upload(file=path)
)
owned_name = uploaded.name if existing_name is None else None
try:
deadline = monotonic() + 300
while True:
state = uploaded.state.name if uploaded.state else None
if state == "ACTIVE":
if not uploaded.uri or not uploaded.mime_type:
raise RuntimeError("Active file has no URI or MIME type")
yield uploaded
return
if state == "FAILED":
raise RuntimeError(f"File processing failed: {uploaded.name}")
if state != "PROCESSING" or not uploaded.name:
raise RuntimeError(f"Unexpected file state: {state}")
if monotonic() >= deadline:
raise TimeoutError("File processing exceeded five minutes")
sleep(2)
uploaded = client.files.get(name=uploaded.name)
finally:
if owned_name:
client.files.delete(name=owned_name)from os import environ
from pathlib import Path
from agno.agent import Agent
from agno.db.in_memory import InMemoryDb
from agno.media import File
from agno.models.google import Gemini
from google_files import ready_file
model = Gemini(id="gemini-3.7-flash")
agent = Agent(
model=model,
db=InMemoryDb(),
add_history_to_context=True,
markdown=True,
)
path = Path(__file__).parent / "sample.mp3"
with ready_file(model.get_client(), path, environ.get("GOOGLE_FILE_NAME")) as uploaded:
agent.print_response(
"Tell me about this audio.",
files=[File(external=uploaded)],
stream=True,
)
agent.print_response("Summarize the main point from our conversation.")File(external=...) passes the uploaded file URI and MIME type. InMemoryDb retains this conversation only within the process. File access must remain valid while history containing its URI is reused.
Run the Example
Set up your virtual environment
uv venv --python 3.12
source .venv/bin/activateInstall dependencies
uv pip install -U agno google-genaiExport your Google API key
export GOOGLE_API_KEY="your_google_api_key_here"Run the example
Save the current adaptation as query_uploaded_audio.py and any helper beside it, then run:
python query_uploaded_audio.pyFull source: cookbook/90_models/google/gemini/audio_input_local_file_upload.py