Summary

Implement complete OpenTelemetry monitoring integration allowing users to send AI call traces to Jaeger, Grafana Tempo, and other OTLP-compatible backends.

Changes

• Add OpenTelemetry SDK dependencies
• Create telemetry module for SDK management
• Extend experimental.openTelemetry configuration from boolean to full object
• Integrate tracer into LLM stream and Agent generation
• Support environment variables (OTEL_EXPORTER_OTLP_ENDPOINT, OTEL_SERVICE_NAME)
• Add LOC (Lines of Code) statistics tracking for file modification tools

Configuration Example

{
  "experimental": {
    "openTelemetry": {
      "enabled": true,
      "endpoint": "http://localhost:4318",
      "serviceName": "opencode",
      "sampleRate": 1.0
    }
  }
}

Tracked Data

AI Operations

• Model ID, provider, operation type
• Token usage (prompt/completion)
• Session ID, agent name, user ID
• Tool calls with arguments
• Request duration and spans

File Modifications (LOC Tracking)

• File path and tool name (edit/write/apply_patch)
• Lines added, deleted, and net change
• Change type (add/update/delete/move)
• Batch operations with aggregated statistics

Test plan

• Type checking passed
• Related tests passed (edit, write, apply_patch)