GitIntel is an AI-powered GitHub repository analyzer that leverages Gemini 2.5 Flash to evaluate code across eight engineering dimensions including security, testing, and documentation. Developed by Divya Singh, the tool builds repository-level insights and comprehensive developer profiles. As the system evolved into a distributed pipeline involving asynchronous GitHub API requests and multiple LLM calls, the developer encountered significant challenges in tracking performance, with end-to-end latency fluctuating between 8 secs and 52 secs.

To address these observability gaps, the application was instrumented with OpenTelemetry and connected to a self-hosted SigNoz instance. This integration provides granular visibility into execution flow, allowing for the correlation of traces, metrics, and logs. By centralizing observability in a dedicated telemetry file, the system captures performance data without requiring manual dashboard queries, automatically discovering the service as gitintel within the SigNoz interface.

Beyond standard HTTP monitoring, the implementation includes eight custom OpenTelemetry instruments to track domain-specific metrics. These include a gauge for remaining GitHub rate limits, counters for prompt and completion token consumption per repository, and a histogram measuring total assessment duration in seconds. By recording these metrics with repository-specific attributes, the developer can now identify expensive analysis workflows, monitor token usage trends, and pinpoint specific latency bottlenecks within the LLM-powered pipeline.

GitIntel is an AI-powered GitHub repository analyzer that leverages Gemini 2.5 Flash to evaluate code across eight engineering dimensions including security, testing, and documentation. Developed by Divya Singh, the tool builds repository-level insights and comprehensive developer profiles. As the system evolved into a distributed pipeline involving asynchronous GitHub API requests and multiple LLM calls, the developer encountered significant challenges in tracking performance, with end-to-end latency fluctuating between 8 secs and 52 secs.

To address these observability gaps, the application was instrumented with OpenTelemetry and connected to a self-hosted SigNoz instance. This integration provides granular visibility into execution flow, allowing for the correlation of traces, metrics, and logs. By centralizing observability in a dedicated telemetry file, the system captures performance data without requiring manual dashboard queries, automatically discovering the service as gitintel within the SigNoz interface.

Beyond standard HTTP monitoring, the implementation includes eight custom OpenTelemetry instruments to track domain-specific metrics. These include a gauge for remaining GitHub rate limits, counters for prompt and completion token consumption per repository, and a histogram measuring total assessment duration in seconds. By recording these metrics with repository-specific attributes, the developer can now identify expensive analysis workflows, monitor token usage trends, and pinpoint specific latency bottlenecks within the LLM-powered pipeline.