langchain-gigachat-lc1 0.4.0b5

An integration package connecting GigaChat and LangChain

English | Русский

langchain-gigachat

LangChain integration for GigaChat (chat models, embeddings, tool calling, and attachments).

This library is part of GigaChain.

Quick Install

pip install -U langchain-gigachat

🤔 What is this?

This package provides:

• Chat model: langchain_gigachat.GigaChat (sync/async, streaming, tool calling, structured output)
• Embeddings: langchain_gigachat.GigaChatEmbeddings
• Tools helper: langchain_gigachat.tools.giga_tool.giga_tool (extends LangChain @tool with GigaChat-specific extras)
• Attachments: upload files and send them as message content_blocks (images/audio/documents)

Requirements

• Python 3.10+
• Access to GigaChat API (credentials, access token, or other supported auth methods)
• TLS root certificate (recommended). If your environment requires it, configure a CA bundle via GIGACHAT_CA_BUNDLE_FILE / ca_bundle_file.

For details on auth and certificates, see:

• GigaChat SDK README
• GigaChat API docs

Quickstart

Chat

from langchain_gigachat import GigaChat

llm = GigaChat(
    credentials="YOUR_AUTHORIZATION_KEY",
    verify_ssl_certs=False,  # dev-only (recommended: configure CA bundle instead)
)

msg = llm.invoke("Hello, GigaChat!")
print(msg.content)

Streaming

from langchain_gigachat import GigaChat

llm = GigaChat(credentials="YOUR_AUTHORIZATION_KEY", verify_ssl_certs=False)

for chunk in llm.stream("Write a short poem about programming"):
    print(chunk.content, end="", flush=True)
print()

Async

import asyncio

from langchain_gigachat import GigaChat


async def main() -> None:
    llm = GigaChat(credentials="YOUR_AUTHORIZATION_KEY", verify_ssl_certs=False)
    msg = await llm.ainvoke("Explain quantum computing in simple terms.")
    print(msg.content)


asyncio.run(main())

Embeddings

from langchain_gigachat import GigaChatEmbeddings

emb = GigaChatEmbeddings(
    credentials="YOUR_AUTHORIZATION_KEY",
    verify_ssl_certs=False,
    model="Embeddings",
)

vector = emb.embed_query("Привет!")
print(len(vector))

Tool calling

Use giga_tool (a drop-in alternative to LangChain @tool with extra fields supported by GigaChat).

from langchain_gigachat import GigaChat
from langchain_gigachat.tools.giga_tool import giga_tool


@giga_tool
def get_weather(city: str) -> str:
    """Get current weather for a city."""
    return f"{city}: sunny"


llm = GigaChat(credentials="YOUR_AUTHORIZATION_KEY", verify_ssl_certs=False)
llm_with_tools = llm.bind_tools([get_weather], tool_choice="auto")

msg = llm_with_tools.invoke("What's the weather in Tokyo?")
print(msg.tool_calls)

Notes:

• tool_choice="any" is not supported by the GigaChat API. Use "auto", "none", or a specific tool name. If you must accept "any" from upstream code, set allow_any_tool_choice_fallback=True in GigaChat(...) to convert it to "auto".

Structured output

from pydantic import BaseModel, Field

from langchain_gigachat import GigaChat


class Answer(BaseModel):
    """Structured answer."""

    text: str = Field(description="Final answer")
    confidence: float = Field(ge=0, le=1, description="Confidence 0..1")


llm = GigaChat(credentials="YOUR_AUTHORIZATION_KEY", verify_ssl_certs=False)
chain = llm.with_structured_output(Answer)
parsed = chain.invoke("Answer briefly and provide confidence.")
print(parsed)

You can also use JSON mode: llm.with_structured_output(Answer, method="json_mode").

Attachments (images/audio/documents)

Upload a file via the GigaChat Files API and pass it as a standard LangChain content_blocks attachment.

from langchain_core.messages import HumanMessage

from langchain_gigachat import GigaChat

llm = GigaChat(credentials="YOUR_AUTHORIZATION_KEY", verify_ssl_certs=False)

with open("image.png", "rb") as f:
    uploaded = llm.upload_file(("image.png", f.read()))

msg = HumanMessage(
    content_blocks=[
        {"type": "text", "text": "Describe the image."},
        {"type": "image", "file_id": uploaded.id_},
    ]
)

reply = llm.invoke([msg])
print(reply.content)

Configuration

All SDK parameters can be passed to GigaChat(...) / GigaChatEmbeddings(...) directly, or configured via environment variables (prefix GIGACHAT_).

Notes:

• If you embed Base64 data URLs into image_url / audio_url / document_url blocks, you can enable auto_upload_attachments=True to auto-upload them. This is not recommended for production; prefer explicit upload_file(...).
• Retries are handled by the underlying gigachat SDK (max_retries, retry_backoff_factor, retry_on_status_codes). Avoid combining SDK retries with LangChain retries (e.g. .with_retry()), otherwise the effective attempts multiply.

Common variables:

Variable	Meaning
GIGACHAT_CREDENTIALS	OAuth credentials (recommended default)
GIGACHAT_ACCESS_TOKEN	Pre-obtained access token (JWT)
GIGACHAT_SCOPE	API scope (GIGACHAT_API_PERS, GIGACHAT_API_B2B, GIGACHAT_API_CORP)
GIGACHAT_BASE_URL	API base URL
GIGACHAT_VERIFY_SSL_CERTS	Enable/disable TLS verification
GIGACHAT_CA_BUNDLE_FILE	Path to CA bundle file

📖 Documentation

• Source code: langchain_gigachat/
• GigaChat SDK: README

💁 Contributing

See CONTRIBUTING.md. Development happens under libs/gigachat (run uv sync, then make lint_package / make test).