Add Telemetry Endpoint Support

[tanguc]

Describe the feature or problem you'd like to solve

Implement telemetry endpoint support in GitHub Copilot CLI, following Claude's philosophy

Proposed solution

Overview

Implement telemetry endpoint support in GitHub Copilot CLI, following Claude's philosophy of:

• User Control: Opt-in telemetry with clear transparency
• Privacy-First: Minimal data collection, user can customize what's sent
• Open Standards: Support for custom telemetry backends (not vendor-locked)
• Developer Experience: Simple configuration, no friction

Claude's Telemetry Philosophy

Claude supports telemetry through:

1. Configurable Endpoints: Users can specify custom telemetry URLs
2. Structured Metrics: Well-defined metrics for request/response tracking
3. User Session Tracking: Session IDs for correlating related requests
4. Cost Attribution: Track token usage and costs per session/user
5. Opt-In by Default: Users explicitly enable telemetry
6. No Automatic Sending: Telemetry happens only when configured
7. Custom Handlers: Ability to intercept and customize what gets sent

Proposed Implementation for Copilot CLI

Configuration

/telemetry                          # Configure telemetry
/telemetry --endpoint <url>         # Set custom telemetry endpoint
/telemetry --enable/--disable       # Toggle telemetry
/telemetry --list-metrics          # Show what would be sent

Environment Variables

COPILOT_TELEMETRY_ENDPOINT=https://your-service.com/events
COPILOT_TELEMETRY_ENABLED=true
COPILOT_TELEMETRY_API_KEY=...      # Optional auth

Metrics to Support

• Session metrics: session_id, start_time, duration, status
• Request metrics: model, prompt_tokens, completion_tokens, latency
• Tool usage: tools_used, success_rate, errors
• User context: org_id, user_id (optional, user-controlled)

Cost Metrics

• input_cost (based on tokens × model pricing)
• output_cost (based on tokens × model pricing)
• total_request_cost
• cumulative_session_cost
• cumulative_monthly_cost
• cost_per_user, cost_per_org
• model_cost_breakdown (cost distribution by model used)

Privacy Features

• Local-only option: store metrics locally without sending
• Data minimization: configurable metric collection level (minimal/standard/detailed)
• User anonymization: hash user IDs by default
• No automatic personal data: require explicit opt-in for PII

Benefits

• Enterprise monitoring and cost tracking
• Usage analytics for teams
• Performance insights
• Compliance with data policies
• Custom integration with existing observability platforms

Examples of Integration

• Send to DataDog, New Relic, Splunk via custom endpoints
• Local event log for audit trails
• Custom metrics processor for cost allocation

Example prompts or workflows

No response

Additional context

No response