Connect LangChain to Anoman AI

Step 1: Install dependencies

pip install langchain langchain-openai

Step 2: Configure Anoman as the LLM backend

Set the environment variables before initializing your LLM. The base URL points to Anoman's OpenAI-compatible endpoint.

import os
from langchain_openai import ChatOpenAI

# Set Anoman as the LLM backend
os.environ["OPENAI_API_KEY"] = "anm-sk-your-key-here"
os.environ["OPENAI_BASE_URL"] = "https://api.anoman.io/v1"

llm = ChatOpenAI(model="gpt-4o-mini")

Step 3: Use in a chain

Your existing chains work without modification. Every call goes through Anoman's guardrail pipeline automatically.

from langchain_core.messages import HumanMessage
from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    ("human", "{input}"),
])

chain = prompt | llm

response = chain.invoke({"input": "Summarize this document."})
print(response.content)

Using with LangGraph

from langgraph.prebuilt import create_react_agent

agent = create_react_agent(llm, tools=[search_tool, calculator_tool])
result = agent.invoke({"messages": [HumanMessage(content="Find the population of Jakarta.")]})

When a guardrail triggers

If a request is blocked by the guardrail pipeline, the API returns a 403 with a structured error. LangChain surfaces this as an exception.

{
  "error": {
    "code": "guardrail_triggered",
    "message": "Request blocked by guardrail pipeline",
    "guardrail": "injection",
    "score": 0.94
  }
}

This example is illustrative. Check docs/openapi/ for the current error schema.

Guardrail transparency headers

Every response includes guardrail result headers. Check these in your API logs to verify the pipeline is running correctly.

X-Anoman-Guardrail-Injection: pass score=0.07
X-Anoman-Guardrail-Pii: pass
X-Anoman-Guardrail-Content: pass
X-Anoman-Cache: none