llama-index-llms-gigachat-ru 0.3.0

llama-index llms gigachat integration

LlamaIndex Llms Integration: Gigachat

This package provides a Gigachat integration as LLM Module.

Installation

using poetry:

poetry add llama-index-llms-gigachat

or using pip:

pip install llama-index-llms-gigachat

Basic Usage

To initialize the Gigachat integration, you need to provide the credentials and optionally set verify_ssl_certs to False if you want to disable SSL certificate verification.

Then you will be able to use the complete method to get the completion.

from llama_index.llms.gigachat import GigaChatLLM

llm = GigaChatLLM(
    credentials="NjI0M2M4MzctNmEwMi00ZjhmLWIzYmEtMTBlMzdhZjI4NzNhOjgzYjM3YzFkLWQ3MTEtNGVhYi04Y2Q0LTkwODM5ZjI4MDg1Zg==",
    verify_ssl_certs=False,
)
resp = llm.complete("What is the capital of France?")
print(resp)

Also, you can use it asynchronously:

import asyncio
from llama_index.llms.gigachat import GigaChatLLM

llm = GigaChatLLM(
    credentials="NjI0M2M4MzctNmEwMi00ZjhmLWIzYmEtMTBlMzdhZjI4NzNhOjgzYjM3YzFkLWQ3MTEtNGVhYi04Y2Q0LTkwODM5ZjI4MDg1Zg==",
    verify_ssl_certs=False,
)


async def main():
    resp = await llm.acomplete("What is the capital of France?")
    print(resp)


asyncio.run(main())

And as a chat module:

import asyncio

from llama_index.core.chat_engine import SimpleChatEngine
from llama_index.llms.gigachat import GigaChatLLM

llm = GigaChatLLM(
    credentials="NjI0M2M4MzctNmEwMi00ZjhmLWIzYmEtMTBlMzdhZjI4NzNhOjgzYjM3YzFkLWQ3MTEtNGVhYi04Y2Q0LTkwODM5ZjI4MDg1Zg==",
    verify_ssl_certs=False,
)

chat = SimpleChatEngine.from_defaults(
    llm=llm,
)


async def main():
    resp = await chat.achat("What is the capital of France?")
    print(resp)


asyncio.run(main())

And as streaming completion:

import asyncio

from llama_index.llms.gigachat import GigaChatLLM

llm = GigaChatLLM(
    credentials="NjI0M2M4MzctNmEwMi00ZjhmLWIzYmEtMTBlMzdhZjI4NzNhOjgzYjM3YzFkLWQ3MTEtNGVhYi04Y2Q0LTkwODM5ZjI4MDg1Zg==",
    verify_ssl_certs=False,
)


async def main():
    async for resp in await llm.astream_complete(
        "What is the capital of France?"
    ):
        print(resp)


asyncio.run(main())