-
Notifications
You must be signed in to change notification settings - Fork 2
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
8 changed files
with
972 additions
and
2 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,4 @@ | ||
# update this with your private key | ||
# you can get one from the Google AI Studio | ||
|
||
GEMINI_API_KEY = 'YOUR_PRIVATE_GEMINI_API_KEY' |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,2 +1,32 @@ | ||
# Gemini_2.0_Live_API_Tutorials | ||
Gemini_2.0_Live_API_Tutorials | ||
# Gemini_2.0_Live_API_Tutorials | ||
|
||
These two Python files provide a demo of the newly release Google Gemini 2.0 Live API. | ||
|
||
## live_api_starter_cv.md | ||
The Live API Starter is a Python application that implements real-time audio and video interaction with Google's Gemini AI model. It creates a bidirectional communication channel where users can send text, audio, and video input while receiving audio and text responses from the model in real-time. | ||
The code shares the video from your webcam with the Gemini model, while you can also do a voice chat. | ||
|
||
## live_api_starter_desk.py | ||
This application is a desktop assistant that combines audio input/output capabilities with screen capture functionality to interact with Google's Gemini API. It creates an interactive experience where users can communicate with the Gemini model through both voice and text while sharing their screen. | ||
The code shares your desktop with the Gemini model, while you can also do a voice chat. | ||
|
||
# References: | ||
https://github.com/google-gemini/cookbook/blob/main/gemini-2/README.md | ||
https://github.com/google-gemini/cookbook/blob/main/gemini-2/live_api_starter.py | ||
|
||
## The new Gemini 2.0 Live API requires: | ||
https://pypi.org/project/google-genai/ | ||
https://github.com/googleapis/python-genai | ||
|
||
To learn more, see the Python SDK reference: | ||
https://googleapis.github.io/python-genai/ | ||
|
||
# Possible bugs | ||
During testing I've noticed that the Gemini model sometimes fails to "see" the webcam or the desktop when you request that via the text prompts, however, it usually works well if you make the same request via the voice prompt. | ||
This might be due to the nature of the experimental release at the time of the tests. | ||
|
||
# Please note | ||
The code is provided as-is for learning purposes, please don't expect any updates in the future. | ||
I've made some changes to the code in the Google cookbook to add some logging and troubleshooting details. | ||
This code was tested with Python 3.12 running on Windows 11. | ||
Unfortuantely, I can't provide any support. |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,112 @@ | ||
# Live API Starter Documentation | ||
|
||
References: | ||
https://github.com/google-gemini/cookbook/blob/main/gemini-2/README.md | ||
https://github.com/google-gemini/cookbook/blob/main/gemini-2/live_api_starter.py | ||
|
||
The new Live API requires: | ||
https://pypi.org/project/google-genai/ | ||
https://github.com/googleapis/python-genai | ||
|
||
This PyPI package doesn't support Live API: | ||
https://pypi.org/project/google-generativeai/ | ||
https://github.com/google-gemini/generative-ai-python | ||
|
||
Notes as of 2024-12-13: | ||
`google-generativeai` - old python sdk, for Gemini API in Google AI Studio only | ||
`google-vertexai` - more complex, for Gemini LLM models in Google Vertex AI only | ||
`google-genai` - new one sdk, for both VertexAI and Gemini API. Supports the Live API. | ||
|
||
## The NEW GenAI API: | ||
|
||
The new Google Gen AI SDK provides a unified interface to Gemini 2.0 through both the Gemini Developer API and the Gemini Enterprise API ( Vertex AI). | ||
With a few exceptions, code that runs on one platform will run on both. The Gen AI SDK also supports the Gemini 1.5 models. | ||
|
||
Python | ||
The Google Gen AI SDK for Python is available on PyPI and GitHub: | ||
google-genai on PyPI --> `pip install google-genai` | ||
python-genai on GitHub | ||
|
||
To learn more, see the Python SDK reference: | ||
https://googleapis.github.io/python-genai/ | ||
|
||
Quickstart | ||
1. Import libraries | ||
``` python | ||
from google import genai | ||
from google.genai import types | ||
``` | ||
2. Create a client | ||
``` python | ||
client = genai.Client(api_key='YOUR_API_KEY') | ||
``` | ||
3. Generate content | ||
``` python | ||
response = client.models.generate_content( | ||
model='gemini-2.0-flash-exp', contents='What is your name?' | ||
) | ||
print(response.text) | ||
``` | ||
|
||
## Overview | ||
The Live API Starter is a Python application that implements real-time audio and video interaction with Google's Gemini AI model. It creates a bidirectional communication channel where users can send text, audio, and video input while receiving audio and text responses from the model in real-time. | ||
|
||
## Key Features | ||
- Real-time audio input/output processing | ||
- Video capture and streaming | ||
- Text-based interaction | ||
- Asynchronous operation | ||
- Bidirectional communication with Gemini AI model | ||
|
||
## Class Documentation | ||
|
||
### AudioLoop | ||
Main class that manages the audio/video streaming pipeline and communication with the Gemini AI model. | ||
|
||
## Method Documentation | ||
|
||
### AudioLoop.__init__ | ||
Initializes queues for audio/video processing and sets up session management. | ||
|
||
### AudioLoop.send_text | ||
Handles text input from the user and sends it to the Gemini session. | ||
|
||
### AudioLoop._get_frame | ||
Captures and processes a single video frame, converting it to JPEG format with size constraints. | ||
|
||
### AudioLoop.get_frames | ||
Continuously captures video frames from the default camera and adds them to the video queue. | ||
|
||
### AudioLoop.send_frames | ||
Sends captured video frames to the Gemini session. | ||
|
||
### AudioLoop.listen_audio | ||
Sets up and manages audio input stream from the microphone. | ||
|
||
### AudioLoop.send_audio | ||
Sends audio chunks from the output queue to the Gemini session. | ||
|
||
### AudioLoop.receive_audio | ||
Processes responses from the Gemini model, handling both text and audio data. | ||
|
||
### AudioLoop.play_audio | ||
Manages audio playback of responses received from the model. | ||
|
||
### AudioLoop.run | ||
Main execution method that coordinates all the async tasks and manages the session lifecycle. | ||
|
||
## Global Constants | ||
|
||
- FORMAT: Set to pyaudio.paInt16 for audio format | ||
- CHANNELS: Set to 1 for mono audio | ||
- SEND_SAMPLE_RATE: 16000Hz for input audio | ||
- RECEIVE_SAMPLE_RATE: 24000Hz for output audio | ||
- CHUNK_SIZE: 512 bytes for audio processing | ||
- MODEL: Uses "models/gemini-2.0-flash-exp" for AI interactions | ||
|
||
## Technical Details | ||
- Uses asyncio for concurrent operations | ||
- Implements PyAudio for audio handling | ||
- Uses OpenCV (cv2) for video capture | ||
- Integrates with Google's Genai client | ||
- Supports Python 3.11+ with fallback for earlier versions |
Oops, something went wrong.