LangGraph Integration

Add trust verification to LangGraph StateGraphs. Gate agent execution on Nerq Trust Scores.

nerq-langgraph v0.1.0 Python

Installation

pip install nerq-langgraph

Quick Start

Add a trust-check node to your LangGraph StateGraph. The node queries the Nerq API and blocks execution if the agent fails verification.

# 1. Import the trust check node
from nerq_langgraph import create_trust_check_node

# 2. Create the node with your threshold
trust_check = create_trust_check_node(
    api_key="your-nerq-api-key",
    min_score=60,        # minimum trust score (0-100)
    agent_field="agent_name"  # state field containing the agent ID
)

# 3. Add to your StateGraph
from langgraph.graph import StateGraph

graph = StateGraph(MyState)
graph.add_node("trust_check", trust_check)
graph.add_node("agent", my_agent_node)
graph.add_node("fallback", fallback_node)

# 4. Route based on trust result
graph.add_edge("trust_check", "agent")     # passes if score >= threshold
graph.set_entry_point("trust_check")
app = graph.compile()

Handling Trust Failures

from nerq_langgraph import TrustCheckFailed

try:
    result = app.invoke({"agent_name": "untrusted-agent-123"})
except TrustCheckFailed as e:
    print(f"Blocked: {e.agent_name} scored {e.score} (min: {e.threshold})")
    # Route to fallback, log event, or alert team

Configuration

Parameter Type Default Description
api_keystrenv NERQ_API_KEYYour Nerq API key
min_scoreint60Minimum trust score to pass (0-100)
agent_fieldstr"agent_name"State field containing agent identifier
cache_ttlint300Cache duration in seconds

FAQ

What is nerq-langgraph?
nerq-langgraph is a Python package that adds trust verification to LangGraph StateGraphs. It provides a trust-check node you insert into your graph that queries the Nerq API and blocks execution if an agent's trust score is below your threshold.
How does the trust check work?
The trust check node calls the Nerq API with the agent ID or name. Nerq returns a trust score (0-100) and grade. If the score is below your configured minimum (default: 60), the node raises a TrustCheckFailed exception, preventing downstream nodes from executing.
What happens if an agent fails the trust check?
When an agent fails the trust check, a TrustCheckFailed exception is raised with details including the agent name, score, and your threshold. You can catch this exception and route to a fallback agent, log the event, or alert your team.
← All integrations