README.md

---
license: mit
tags:
  - tutorial
  - crazyrouter
  - langchain
  - llamaindex
  - autogen
  - ai-agents
  - llm
language:
  - en
  - zh
---

# 🔗 Crazyrouter + LangChain / LlamaIndex / AutoGen

> Use 624+ AI models in your favorite framework — zero config changes.

Since Crazyrouter is 100% OpenAI-compatible, it works out of the box with every major AI framework.

---

## LangChain

### Installation

```bash
pip install langchain langchain-openai
```

### Basic Chat

```python
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="gpt-4o"
)

response = llm.invoke("Explain microservices in one paragraph")
print(response.content)
```

### Switch Models on the Fly

```python
# Use Claude for analysis
analyst = ChatOpenAI(
    base_url="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="claude-sonnet-4-20250514"
)

# Use DeepSeek for coding (cheaper)
coder = ChatOpenAI(
    base_url="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="deepseek-chat"
)

# Use GPT-4o-mini for simple tasks (cheapest)
helper = ChatOpenAI(
    base_url="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="gpt-4o-mini"
)
```

### LangChain Chains

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

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful coding assistant."),
    ("user", "{question}")
])

chain = prompt | llm | StrOutputParser()

result = chain.invoke({"question": "How do I read a CSV file in Python?"})
print(result)
```

### RAG with Crazyrouter

```python
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain_community.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough

# Embeddings (use a cheap model)
embeddings = OpenAIEmbeddings(
    base_url="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="text-embedding-3-small"
)

# Chat model
llm = ChatOpenAI(
    base_url="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="gpt-4o-mini"
)

# Split your documents
splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
docs = splitter.create_documents(["Your document text here..."])

# Create vector store
vectorstore = FAISS.from_documents(docs, embeddings)
retriever = vectorstore.as_retriever()

# RAG chain
template = """Answer based on context:
{context}

Question: {question}"""

prompt = ChatPromptTemplate.from_template(template)

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

result = chain.invoke("What does the document say?")
print(result.content)
```

---

## LlamaIndex

### Installation

```bash
pip install llama-index llama-index-llms-openai-like
```

### Basic Usage

```python
from llama_index.llms.openai_like import OpenAILike

llm = OpenAILike(
    api_base="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="gpt-4o",
    is_chat_model=True
)

response = llm.complete("What is retrieval augmented generation?")
print(response)
```

### With OpenAI Class Directly

```python
from llama_index.llms.openai import OpenAI
import os

os.environ["OPENAI_API_KEY"] = "sk-your-crazyrouter-key"
os.environ["OPENAI_API_BASE"] = "https://crazyrouter.com/v1"

llm = OpenAI(model="gpt-4o-mini")
response = llm.complete("Hello!")
print(response)
```

---

## AutoGen

### Installation

```bash
pip install autogen-agentchat
```

### Multi-Agent Setup

```python
import autogen

config_list = [
    {
        "model": "gpt-4o",
        "base_url": "https://crazyrouter.com/v1",
        "api_key": "sk-your-crazyrouter-key",
    }
]

llm_config = {"config_list": config_list}

# Create agents
assistant = autogen.AssistantAgent(
    name="assistant",
    llm_config=llm_config,
)

user_proxy = autogen.UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",
    max_consecutive_auto_reply=3,
    code_execution_config={"work_dir": "coding"},
)

# Start conversation
user_proxy.initiate_chat(
    assistant,
    message="Write a Python script that fetches the top 10 Hacker News stories."
)
```

### Multi-Model Agents (Cost Optimization)

```python
# Expensive model for complex reasoning
senior_config = [{"model": "gpt-4o", "base_url": "https://crazyrouter.com/v1", "api_key": "sk-your-key"}]

# Cheap model for simple tasks
junior_config = [{"model": "gpt-4o-mini", "base_url": "https://crazyrouter.com/v1", "api_key": "sk-your-key"}]

senior = autogen.AssistantAgent("senior", llm_config={"config_list": senior_config})
junior = autogen.AssistantAgent("junior", llm_config={"config_list": junior_config})
```

---

## CrewAI

```python
from crewai import Agent, Task, Crew
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://crazyrouter.com/v1",
    api_key="sk-your-crazyrouter-key",
    model="gpt-4o"
)

researcher = Agent(
    role="Researcher",
    goal="Find accurate information",
    backstory="You are an expert researcher.",
    llm=llm
)

task = Task(
    description="Research the latest trends in AI API gateways",
    agent=researcher,
    expected_output="A summary of trends"
)

crew = Crew(agents=[researcher], tasks=[task])
result = crew.kickoff()
print(result)
```

---

## Environment Variables (Works Everywhere)

Set these once and most frameworks auto-detect:

```bash
export OPENAI_API_KEY="sk-your-crazyrouter-key"
export OPENAI_API_BASE="https://crazyrouter.com/v1"
export OPENAI_BASE_URL="https://crazyrouter.com/v1"
```

---

## Pro Tips

1. **Use cheap models for agents that do simple tasks** — `gpt-4o-mini` or `deepseek-chat` for routing, summarizing, formatting
2. **Use powerful models for reasoning** — `gpt-4o`, `claude-sonnet-4-20250514`, or `deepseek-reasoner` for complex analysis
3. **Mix providers freely** — one agent uses Claude, another uses GPT, another uses Gemini. All through one API key
4. **Monitor costs** — Crazyrouter dashboard shows per-model usage

---

## Links

- 🌐 [Crazyrouter](https://crazyrouter.com/?utm_source=huggingface&utm_medium=tutorial&utm_campaign=dev_community)
- 📖 [Getting Started Guide](https://huggingface.co/xujfcn/Crazyrouter-Getting-Started)
- 🤖 [Live Demo](https://huggingface.co/spaces/xujfcn/Crazyrouter-Demo)
- 💬 [Telegram](https://t.me/crazyrouter)
- 🐦 [Twitter @metaviiii](https://twitter.com/metaviiii)