Enterprise-grade observability for AI agents using OpenTelemetry
NexusHive Agent Insights provides comprehensive instrumentation, monitoring, and evaluation tools for AI agents, with a focus on the smolagents framework. Built on OpenTelemetry standards, it enables organizations to track critical agent metrics, evaluate performance, and ensure reliability in production environments.
- OpenTelemetry Integration: Seamlessly instrument agents with industry-standard telemetry
- Comprehensive Metrics: Track token usage, latency, errors, and custom KPIs
- Real-time Evaluation: Support for user feedback and LLM-as-judge evaluations
- Offline Analysis: Tools for benchmark testing (e.g., GSM8K) and performance comparison
- smolagents Framework Support: Built-in integration with smolagents architecture
- Agent Performance Tracking: Monitor costs, response times, and success rates
- Operational Oversight: Identify errors and bottlenecks in agent deployments
- Continuous Improvement: Evaluate and refine agents based on real-world usage
- Benchmark Testing: Compare agent implementations against standard datasets
# Install the package
pip install nexushive-agent-insights
# Basic usage
from nexushive.client import instrument_agent
from smolagents import Agent
# Instrument your agent
my_agent = Agent(...)
instrumented_agent = instrument_agent(my_agent)
# Use normally - telemetry is automatically collected
result = instrumented_agent.run(...)src/client/: Agent instrumentation librariessrc/collector/: Telemetry data collection and processingsrc/examples/: Usage examples and templates
MIT License