hugging-chat-api

English | 简体中文

Unofficial HuggingChat Python API, extensible for chatbots etc.

Note
Recently new updates:

• Web search
• Memorize context
• Change LLMs supported. See more at Soulter#56 (v0.0.9)

Installation

pip install hugchat

or

pip3 install hugchat

Usage

API

from hugchat import hugchat
from hugchat.login import Login

# Log in to huggingface and grant authorization to huggingchat
sign = Login(email, passwd)
cookies = sign.login()

# Save cookies to the local directory
cookie_path_dir = "./cookies_snapshot"
sign.saveCookiesToDir(cookie_path_dir)

# Load cookies when you restart your program:
# sign = login(email, None)
# cookies = sign.loadCookiesFromDir(cookie_path_dir) # This will detect if the JSON file exists, return cookies if it does and raise an Exception if it's not.

# Create a ChatBot
chatbot = hugchat.ChatBot(cookies=cookies.get_dict())  # or cookie_path="usercookies/<email>.json"


# non stream response
query_result = chatbot.query("Hi!")
print(query_result) # or query_result.text or query_result["text"]

for resp in chatbot.query(
    "Hello",
    stream=True
):
    print(resp)  # stream response

# Use web search *new
query_result = chatbot.query("Hi!", web_search=True)
print(query_result) # or query_result.text or query_result["text"]
for source in query_result.web_search_sources:
    print(source.link)
    print(source.title)
    print(source.hostname)

# Create a new conversation
id = chatbot.new_conversation()
chatbot.change_conversation(id)

# Get conversation list
conversation_list = chatbot.get_conversation_list()

# Switch model (default: meta-llama/Llama-2-70b-chat-hf. )
chatbot.switch_llm(0) # Switch to `OpenAssistant/oasst-sft-6-llama-30b-xor`
chatbot.switch_llm(1) # Switch to `meta-llama/Llama-2-70b-chat-hf`

The query() function receives these parameters:

• text: Required[str].
• temperature: Optional[float]. Default is 0.9
• top_p: Optional[float]. Default is 0.95
• repetition_penalty: Optional[float]. Default is 1.2
• top_k: Optional[int]. Default is 50
• truncate: Optional[int]. Default is 1024
• watermark: Optional[bool]. Default is False
• max_new_tokens: Optional[int]. Default is 1024
• stop: Optional[list]. Default is ["</s>"]
• return_full_text: Optional[bool]. Default is False
• stream: Optional[bool]. Default is True
• use_cache: Optional[bool]. Default is False
• is_retry: Optional[bool]. Default is False
• retry_count: Optional[int]. Number of retries for requesting huggingchat. Default is 5

CLI

version 0.0.5.2 or newer

Simply run the following command in your terminal to start the CLI mode

python -m hugchat.cli

CLI params:

• -u <your huggingface email> : Provide account email to login.
• -p : Force request password to login, ignores saved cookies.
• -s : Enable streaming mode output in CLI.

Commands in cli mode:

• /new : Create and switch to a new conversation.

• /ids : Shows a list of all ID numbers and ID strings in current session.

• /switch <id> : Switches to the ID number or ID string passed.

• /del <id> : Deletes the ID number or ID string passed. Will not delete active session.

• /clear : Clear the terminal.

• /llm : Get available models you can switch to.

• /llm <index> : Switches model to given model index based on /llm.

• /sharewithauthor <on|off> : Changes settings for sharing data with model author. On by default.

• /exit : Closes CLI environment.

• /stream <on|off>: streaming the response.

• /web <on|off>: web search.

• /web-hint <on|off>: display web search hint.

• AI is an area of active research with known problems such as biased generation and misinformation. Do not use this application for high-stakes decisions or advice.

• Server resources are precious, it is not recommended to request this API in a high frequency. (Hugging Face's CTO🤗 just liked the suggestion)

Disclaimers

This is not an official Hugging Face product. This is a personal project and is not affiliated with Hugging Face in any way. Don't sue us.

Star History