agent-observability-langchain

Agent Observability for LangChain

LangChain integration for the Agent Observability platform — structured logging, cost tracking, and compliance audit trails for production AI agents.

Installation

pip install agent-observability-langchain

Quick Start

Step 1: Get Your Free API Key (One-Time)

import requests

response = requests.post(
    "https://api-production-0c55.up.railway.app/v1/register",
    json={"agent_id": "my-langchain-agent"}
)
api_key = response.json()["api_key"]
print(f"Your API key: {api_key}")
# Save this! Example: ao_live_abc123...

Or via curl:

curl -X POST https://api-production-0c55.up.railway.app/v1/register \
  -H "Content-Type: application/json" \
  -d '{"agent_id": "my-langchain-agent"}'

Step 2: Set Environment Variable

export AGENT_OBS_API_KEY=ao_live_your_key_here

Step 3: Add to Your Agent

from langchain.agents import initialize_agent, AgentType
from langchain_openai import ChatOpenAI
from agent_observability_langchain import AgentObservabilityTool

# Create observability tool
obs_tool = AgentObservabilityTool()

# Create agent with observability
agent = initialize_agent(
    tools=[obs_tool],
    llm=ChatOpenAI(model="gpt-4"),
    agent=AgentType.OPENAI_FUNCTIONS,
    verbose=True
)

# Use your agent — it can now log events for compliance
result = agent.run(
    "Log an API call to OpenAI with cost $0.05 and latency 1200ms"
)

Features

Use Cases

1. Cost Analysis

Track spending across all your agents and LLM providers:

from agent_observability_langchain import AgentObservabilityTool

tool = AgentObservabilityTool(agent_id="research-bot-v2")

# Log each API call with cost
tool.invoke({
    "event_type": "api_call",
    "metadata": {
        "provider": "openai",
        "model": "gpt-4",
        "cost_usd": 0.03,
        "latency_ms": 1250,
        "tokens": 1500
    }
})

2. Decision Auditing

For regulated industries (finance, healthcare, legal):

tool.invoke({
    "event_type": "decision",
    "metadata": {
        "decision": "approve_loan",
        "confidence": 0.92,
        "factors": ["credit_score", "income", "debt_ratio"],
        "customer_id": "cust_12345"
    },
    "severity": "info"
})

3. Error Tracking

tool.invoke({
    "event_type": "error",
    "metadata": {
        "error_type": "RateLimitError",
        "provider": "openai",
        "retry_count": 3,
        "resolved": False
    },
    "severity": "error"
})

4. With LangGraph

from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI
from agent_observability_langchain import AgentObservabilityTool

# Create agent with observability
agent = create_react_agent(
    ChatOpenAI(model="gpt-4"),
    tools=[AgentObservabilityTool()]
)

result = agent.invoke({
    "messages": [("user", "Research AI trends and log the costs")]
})

5. Custom Agent ID per Task

# Different agent IDs for different workflows
research_tool = AgentObservabilityTool(agent_id="research-agent")
writer_tool = AgentObservabilityTool(agent_id="writer-agent")
reviewer_tool = AgentObservabilityTool(agent_id="reviewer-agent")

API Reference

AgentObservabilityTool

AgentObservabilityTool(
    api_key: str = None,        # Or set AGENT_OBS_API_KEY env var
    agent_id: str = "langchain-agent",  # Your agent identifier
    api_base: str = None        # Override API URL (for self-hosted)
)

Input Schema

Common Metadata Fields

Pricing

Full pricing: https://api-production-0c55.up.railway.app/pricing.json

Convenience Function

from agent_observability_langchain import create_observability_tool

# Quick creation
tool = create_observability_tool(agent_id="my-agent")

Environment Variables

Error Handling

The tool handles errors gracefully and returns error messages instead of raising exceptions:

result = tool.invoke({
    "event_type": "api_call",
    "metadata": {"provider": "openai"}
})

if "Failed" in result:
    print(f"Logging failed: {result}")
else:
    print(f"Success: {result}")

Related Packages

• agent-observability — Core Python SDK
• agent-observability-mcp — Claude Desktop MCP integration
• agent-observability-autogpt — AutoGPT plugin
• agent-observability-crewai — CrewAI integration

License

MIT License — see LICENSE for details.

Links

• API Documentation
• OpenAPI Spec
• Pricing

Contributing

Contributions welcome! Please open an issue or submit a pull request.