Building an Observable AI Market Research Agent with SigNoz

A developer built an observable AI market research agent using Google's Agent Development Kit (ADK) and instrumented it with OpenTelemetry to export telemetry data to SigNoz. The agent tracks distributed tracing, token usage, latency, and errors for every tool call and LLM generation, providing full visibility into variable latency and unpredictable logic paths.

AI Market Research Agent 🤖 SigNoz Hackathon Submission 👁️ Observability & Monitoring with SigNoz This agent is fully instrumented using OpenTelemetry to export telemetry data to SigNoz . Because AI agents involve variable latency and unpredictable logic paths, full visibility is critical. What is Tracked via openinference-instrumentation-google-adk: - Distributed Tracing: Every tool call like google search and LLM generation is traced. - Token Usage & Latency: Tracks exact time taken by Gemini 2.0 Flash to synthesize data. - Error Tracking: API rate limits or failed search queries are caught in the trace. 🚀 How Judges Can Run This Locally - Ensure SigNoz is running on your machine. - Install dependencies: pip install -r requirements.txt - Export the SigNoz endpoint: export OTEL EXPORTER OTLP ENDPOINT="http://localhost:4318/v1/traces" - Set your Gemini key: export GOOGLE API KEY="your api key" - Run the agent using the ADK CLI: adk run agent.py