Using LangChain | Tutujin API Docs

LangChain connects to Tutujin API through the langchain-openai package: both ChatOpenAI and OpenAIEmbeddings accept a custom base_url. Set it to https://www.tutujin.com/v1, use your Tutujin key as api_key, and use a model name from the model hub. For the native Claude API, use ChatAnthropic from langchain-anthropic with base_url set to https://www.tutujin.com.

Install

pip install langchain-openai langchain-anthropic

npm install @langchain/openai @langchain/core

Chat models

import os
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="gpt-5.4-mini",
    base_url="https://www.tutujin.com/v1",
    api_key=os.environ["TUTUJIN_API_KEY"],
    temperature=0.3,
    max_retries=3,
    timeout=60,
)

print(llm.invoke("Explain vector databases in one sentence.").content)

# streaming
for chunk in llm.stream("Write a 60-word product description"):
    print(chunk.content, end="", flush=True)

import { ChatOpenAI } from '@langchain/openai'

const llm = new ChatOpenAI({
  model: 'gpt-5.4-mini',
  apiKey: process.env.TUTUJIN_API_KEY,
  configuration: { baseURL: 'https://www.tutujin.com/v1' },
  temperature: 0.3,
})

const res = await llm.invoke('Explain vector databases in one sentence.')
console.log(res.content)

// streaming
for await (const chunk of await llm.stream('Write a 60-word product description')) {
  process.stdout.write(String(chunk.content))
}

To switch models, change model. Every model the catalog lists for Chat Completions — DeepSeek, Qwen, Claude, Gemini and more — works with ChatOpenAI.

Building a chain

Python

from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("system", "You write e-commerce copy in under 30 words."),
    ("user", "Write one selling point for: {product}"),
])
chain = prompt | llm | StrOutputParser()
print(chain.invoke({"product": "500 ml stainless steel insulated bottle"}))

Embeddings

Python

import os
from langchain_openai import OpenAIEmbeddings

embeddings = OpenAIEmbeddings(
    model="text-embedding-3-small",
    base_url="https://www.tutujin.com/v1",
    api_key=os.environ["TUTUJIN_API_KEY"],
)
vectors = embeddings.embed_documents(["What is the return policy?", "How soon do orders ship?"])
query_vector = embeddings.embed_query("how do I return an item")

For embedding models from other vendors (for example qwen3-embedding-8b), add check_embedding_ctx_length=False. By default OpenAIEmbeddings tokenizes text with OpenAI's tokenizer and sends token IDs, which other vendors' models do not accept; turning the check off sends the raw text:

Python

embeddings = OpenAIEmbeddings(
    model="qwen3-embedding-8b",
    base_url="https://www.tutujin.com/v1",
    api_key=os.environ["TUTUJIN_API_KEY"],
    check_embedding_ctx_length=False,
)

Plug the embeddings into any vector store LangChain supports (PGVector, Milvus, Qdrant, Chroma…) for retrieval-augmented generation; the full design is in Enterprise knowledge base and RAG.

Native Claude calls

Python

import os
from langchain_anthropic import ChatAnthropic

claude = ChatAnthropic(
    model="claude-sonnet-5-5",
    base_url="https://www.tutujin.com",
    api_key=os.environ["TUTUJIN_API_KEY"],
    max_tokens=1024,
)
print(claude.invoke("Summarize these meeting notes in three points: ...").content)

The base URL is the site root, without /v1.

Troubleshooting

Symptom	Cause and fix
404	`ChatOpenAI` needs `/v1` in `base_url`; `ChatAnthropic` does not
400 from embeddings	A non-OpenAI embedding model without `check_embedding_ctx_length=False`
503 with `model_not_found`	The key's group lacks this model — see Choosing a group

Still stuck? Add our WeChat support JsnonoChat and send the request id from the error message, and we will look into it.