Python Client for Bard, a Chat Based AI tool by Google.
Note
This is an unofficial client.
- Connect to Bard, Google's AI-powered personal assistant.
- Ask questions and have a conversation.
- Improve responses by defining the conversation tone and length.
- Use asyncio for efficient and non-blocking I/O operations.
- Python 3.10 or newer
- Google account with access to Bard
To install Bard.py, run the following command:
pip install bard-py
or, if you use poetry:
poetry add bard-py
Tip
Make sure you're using the latest version of Bard.py to ensure best compatibility with Bard.
To use Bard.py you first need to extract the __Secure-1PSID
and __Secure-1PSIDTS
cookies from the Bard web page. These cookies are used to authenticate your requests to the Bard API.
To get the cookies, follow these steps on Chrome:
- Go to the Bard web page.
- Write a message on the chat dialog that appears.
- Open the developer tools in your browser (usually by pressing
F12
or right-clicking on the chat dialog and selectingInspect
). - Select the
Application
tab and click on theCookies
option to view all cookies associated withhttps://bard.google.com
. - Look for the
__Secure-1PSID
and__Secure-1PSIDTS
cookies and click on them to expand their details. - Copy the values of the cookies (they should look like a long string of letters and numbers).
Then, set them as environment variables in your shell:
export SECURE_1PSID=<your-cookie>
export SECURE_1PSIDTS=<your-other-cookie>
or, in your Python code:
os.environ["SECURE_1PSID"] = "<your-cookie>"
os.environ["SECURE_1PSIDTS"] = "<your-other-cookie>"
You can use Bard.py to easily create a CLI client for Bard:
import asyncio
from bard import BardClient
async def main() -> None:
async with BardClient() as bard:
while True:
prompt = input("You: ")
if prompt == "!reset":
await bard.reset_conversation()
continue
elif prompt == "!exit":
break
response = await bard.ask(prompt)
print(f"Bard: {response}")
if __name__ == "__main__":
asyncio.run(main())
You can create a Bard Client and initialize a connection with Bard which starts a conversation:
bard = BardClient()
await bard.start_conversation()
# Conversation
await bard.end_conversation()
Alternatively, you can use the async with
statement to keep the code compact:
async with BardClient() as bard:
# Conversation
You can reset the conversation in order to make the client forget the previous conversation:
async with BardClient() as bard:
# Conversation
await bard.reset_conversation()
You can ask Bard questions and get the results:
async with BardClient() as bard:
response = await bard.ask("When was Bard released?")
print(response)
You can set the tone when having a conversation with Bard:
async with BardClient() as bard:
_ = await bard.ask("When was Bard released?")
response = await bard.ask("When was Bard released?", tone="Professional")
print(response)
The available options for the tone
parameter are:
Professional
Casual
Note
It is recommended to use the tone
parameter on subsequent prompts and not in the first one. This is because this feature is typically used to change the previous response, rather than define the entire conversation tone.
You can set the response length when having a conversation with Bard:
async with BardClient() as bard:
_ = await bard.ask("When was Bard released?")
response = await bard.ask("When was Bard released?", length="Short")
print(response)
The available options for the length
parameter are:
Short
Long
Note
It is recommended to use the length
parameter on subsequent prompts and not in the first one. This is because this feature is typically used to change the previous response, rather than define the entire conversation length.
When something goes wrong, Sydney.py might throw one of the following exceptions:
Exception | Meaning | Solution |
---|---|---|
CreateConversationException |
Failed to create conversation | Retry or use new cookies |
AskException |
Failed to get response from Bard | Retry or use new cookies |
NoResponseException |
Received an empty response from Bard | Wait and retry |
For more detailed documentation and options, please refer to the code docstrings.
This project is licensed under the MIT License - see the LICENSE file for details.