LangChain

AgentV provides native support for LangChain, allowing you to evaluate chains, agents, and complex RAG workflows.

1. Setup Your Agent API

Expose your LangChain agent via a standard HTTP endpoint or using LangServe.

from langchain.chat_models import ChatOpenAI
from langchain.schema import HumanMessage
from fastapi import FastAPI

chat = ChatOpenAI()
app = FastAPI()


@app.post("/execute_task")
async def execute(request: dict):
    # Process turn
    result = chat([HumanMessage(content=request["task_description"])])
    return {"action": "final_answer", "summary": result.content}

2. Run Evaluation

Use the langchain protocol to connect the harness to your agent service.

agentv run \
  --path scenarios/loan_scenario.json \
  --protocol langchain \
  --agent langchain://localhost:8000/execute_task \
  --agent-name "LangChain-Retail-Bot"

3. Verify Grounding

If your LangChain agent uses RAG, AgentV can track and visualize Grounding Coverage heatmaps to show which parts of your knowledge base were used to answer specific tasks.

reports/coverage/

# Reports are generated automatically in: