juchats

Juchats API wrapper

juchats lib is a Python library designed for interacting with the Juchats API, enabling seamless integration of chat functionalities into your applications. By utilizing this library, developers can leverage the power of advanced models like GPT-4, Claude Mezzo, and deepseek to perform various chat-related tasks.

Installation

pip3 install juchats

Usage

First, obtain your token from from Juchats official website, and place it into a .env file.

JTOKEN=your_token

Basic Chat Interaction

This example demonstrates a simple chat interaction where we ask the model to compare two floating-point numbers.

import os
from juchats.chat import Juchats
from dotenv import load_dotenv
import asyncio
load_dotenv()

async def api():
    token = os.getenv('JTOKEN')
    juchats = Juchats(token, model='deepseek-chat')

    async with juchats:
        await juchats.chat("3.11, 3.9 两个浮点数谁大？具体分析一下，给出你的原因", show_stream=True)

if __name__ == '__main__':
    asyncio.run(api())


''' Output
3.11 大，因为 3.11 > 3.9
'''

Structured JSON output

This example demonstrates how to obtain structured JSON output from the chat API.

import os
from juchats.chat import Juchats
from dotenv import load_dotenv
import asyncio
load_dotenv()

async def api():
    token = os.getenv('JTOKEN')
    juchats = Juchats(token, model='deepseek-chat')
    prompt = "每个月有多少天？以 JSON 格式给出答案，例如：{\"January\": 31, \"February\": 28, ...}"
    async with juchats:
        text = await juchats.chat(prompt)
        print(text)

''' Output
```json
{
    "January": 31,
    "February": 28,
    "March": 31,
    "April": 30,
    "May": 31,
    "June": 30,
    "July": 31,
    "August": 31,
    "September": 30,
    "October": 31,
    "November": 30,
    "December": 31
}
'''

Available Models

By 2024-07-25, the available models (may be outdated) are:

Model ID	Backend Model Name	Front Model Name
5	claude-3-haiku-20240307	Claude Mezzo
6	claude-3-opus-20240229	Claude3 Opus
7	mistralai/mixtral-8x22b-instruct	Mixtral Forte
9	gpt-4-turbo-2024-04-09	GPT Mezzo
10	gpt-4-turbo-2024-04-09	GPT Forte
11	dall-e-3	DALL · E3
12	meta-llama/llama-3-70b-instruct	Llama3 70B
13	google/gemini-pro-1.5	Gemini 1.5 Pro
14	deepseek-chat	Deepseek
15	google/gemini-flash-1.5	Gemini-flash
16	gpt-4o-2024-05-13	GPT4o
17	claude-3-opus-20240229	Claude3 Opus(100K)
18	Stable Image Ultra	Stable Diffusion 3 Ultra
19	claude-3-5-sonnet-20240620	Claude 3.5 Sonnet
20	gpt-4o-mini-2024-07-18	GPT4o-mini
21	meta-llama/llama-3.1-405b-instruct	Llama3.1 405B

Get real time available models

Dynamically retrieve the latest available models from the Juchats API.

import os
from juchats.chat import Juchats
from dotenv import load_dotenv
import asyncio
load_dotenv()
token = os.getenv('JTOKEN')
juchats = Juchats(token, model='gpt-4o-2024-05-13')
print(
    asyncio.run(juchats.get_models())
)

Note

• Streaming: Set show_stream=True to display the chat response in real-time. Use show_stream=False to disable it.
• Model Selection: Specify the backend model name with the model parameter. Refer to the available models table above for options.
• API Token: Obtain your token from Juchats and use it to authenticate requests.
• Rate Limiting: The API supports up to 3 queries per second (QPS). For higher limits, consider using the Deepseek API or OpenAI API.