> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tokenlab.sh/llms.txt
> Use this file to discover all available pages before exploring further.

# LangChain

> Integrate TokenLab with LangChain using standard OpenAI-compatible chat and embeddings

## Overview

TokenLab works well with LangChain's `ChatOpenAI` and `OpenAIEmbeddings` integrations when you stay on the standard OpenAI-compatible chat and embeddings surface.

This page intentionally covers the standard OpenAI-compatible LangChain surface, not provider-native LangChain features beyond that surface.

<Note>
  Current LangChain docs note that `ChatOpenAI` targets official OpenAI-compatible request/response shapes. If you need provider-specific, non-standard response fields, use a provider-specific LangChain integration instead of relying on `ChatOpenAI`.
</Note>

<Note>
  **Type**: Framework or Platform

  **Primary Path**: OpenAI-compatible standard surface

  **Support Confidence**: Supported standard surface
</Note>

## Installation

```bash theme={null}
pip install langchain langchain-openai langchain-community faiss-cpu
```

## Basic Configuration

```python theme={null}
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="gpt-5.4",
    api_key="sk-your-tokenlab-key",
    base_url="https://api.tokenlab.sh/v1",
)

response = llm.invoke("Explain TokenLab in one sentence.")
print(response.content)
```

## Using Different Models

```python theme={null}
from langchain_openai import ChatOpenAI

gpt = ChatOpenAI(
    model="gpt-5.4",
    api_key="sk-your-key",
    base_url="https://api.tokenlab.sh/v1",
)

claude = ChatOpenAI(
    model="claude-sonnet-4-6",
    api_key="sk-your-key",
    base_url="https://api.tokenlab.sh/v1",
)

gemini = ChatOpenAI(
    model="gemini-3.5-flash",
    api_key="sk-your-key",
    base_url="https://api.tokenlab.sh/v1",
)

deepseek = ChatOpenAI(
    model="deepseek-r1",
    api_key="sk-your-key",
    base_url="https://api.tokenlab.sh/v1",
)
```

## Message History

```python theme={null}
from langchain_core.messages import HumanMessage, SystemMessage

messages = [
    SystemMessage(content="You are a helpful assistant."),
    HumanMessage(content="What is the capital of France?")
]

response = llm.invoke(messages)
print(response.content)
```

## Streaming

```python theme={null}
for chunk in llm.stream("Write a short poem about coding."):
    if chunk.content:
        print(chunk.content, end="", flush=True)
```

## Embeddings

```python theme={null}
from langchain_openai import OpenAIEmbeddings

embeddings = OpenAIEmbeddings(
    model="text-embedding-3-small",
    api_key="sk-your-key",
    base_url="https://api.tokenlab.sh/v1",
)

vector = embeddings.embed_query("Hello world")
print(vector[:5])
```

## Simple RAG Example

```python theme={null}
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough

embeddings = OpenAIEmbeddings(
    model="text-embedding-3-small",
    api_key="sk-your-key",
    base_url="https://api.tokenlab.sh/v1",
)

texts = [
    "TokenLab provides one API for many AI models.",
    "TokenLab supports OpenAI-compatible integrations."
]

vectorstore = FAISS.from_texts(texts, embeddings)
retriever = vectorstore.as_retriever()

prompt = ChatPromptTemplate.from_template(
    "Answer using the context below.\\n\\nContext:\\n{context}\\n\\nQuestion:\\n{question}"
)

rag_chain = (
    {"context": retriever, "question": RunnablePassthrough()}
    | prompt
    | llm
)

response = rag_chain.invoke("What does TokenLab provide?")
print(response.content)
```

## Agents

<Note>
  For new agentic projects, LangChain recommends considering LangGraph for more explicit control over long-running and tool-using workflows.
</Note>

```python theme={null}
from langchain.agents import create_openai_tools_agent, AgentExecutor
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.tools import tool

@tool
def search(query: str) -> str:
    """Search for information."""
    return f"Search results for: {query}"

tools = [search]

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant with access to tools."),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}")
])

agent = create_openai_tools_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools)

result = executor.invoke({"input": "Search for TokenLab pricing"})
print(result["output"])
```

## Best Practices

<AccordionGroup>
  <Accordion title="Pass base_url explicitly">
    The most reliable TokenLab setup is to pass `base_url="https://api.tokenlab.sh/v1"` directly to `ChatOpenAI` and `OpenAIEmbeddings` instead of depending on older environment-variable aliases.
  </Accordion>

  <Accordion title="Use standard features here">
    Stick to standard chat, tool calling, streaming, and embeddings on `ChatOpenAI`. If you need vendor-native extras, switch to the vendor's own LangChain integration.
  </Accordion>

  <Accordion title="Use cheaper models for retrieval">
    Use embedding models like `text-embedding-3-small` for retrieval and reserve stronger chat models for the final answer step.
  </Accordion>
</AccordionGroup>