Copilot Telemetry Analyzer (Local-First)

A local-first Next.js app to anonymize GitHub Copilot telemetry JSON files and visualize KPIs, aligned with GitHub’s Metrics Viewer guidance.

Quick start

1. Place raw telemetry files in ./telemetry (workspace root, sibling of the Next.js app). Each file is an array of day objects matching the schema described below.
2. Install and run:

cd copilot-telemetry
npm install
npm run dev

3. Visit /tools to anonymize. This reads ../telemetry and writes ../telemetry_anonymized. You can also download a ZIP of the output.
4. Visit / for the dashboard. Use the period picker (7/28/90/100 days) to change the window and see comparison deltas vs the previous period.

Environment:

• Optional: ANON_SALT to control deterministic hashing. If not set, a .anonymizer-salt file is generated in the app root on first run.

Schema (DayMetrics)

• date (YYYY-MM-DD)
• total_active_users
• total_engaged_users
• copilot_ide_code_completions with editors/models/languages per the GitHub Copilot Metrics API example
• copilot_ide_chat (totals per editor and model)
• copilot_dotcom_chat
• copilot_dotcom_pull_requests (repositories may contain sensitive slugs)

See src/types/copilot.ts for strong types.

Anonymization

• Deterministic pseudonyms using SHA-256(salt + value) (hex, 8-char prefix)
• Repository names: org/repo → repo-<hash8>
• Custom model names (is_custom_model===true): model-<hash8>, removes custom_model_training_date
• Preserves editor names (vscode, neovim, jetbrains) and language names
• Strips unknown fields by reconstructing the typed shape
• Run at /tools with the Anonymize button; ZIP download available

Dashboard metrics

Top KPIs (with tooltips):

• Active users: users with an active Copilot license using Copilot that day.
• Engaged users: users who engaged with a Copilot feature that day (e.g., accepted a suggestion, used chat). Breakdowns by IDE and language.
• IDE code completions engaged users: sum of copilot_ide_code_completions.total_engaged_users over the period.
• Chat engaged users: sum of copilot_ide_chat.total_engaged_users over the period.
• Acceptance rate (suggestions) = accepted suggestions / total suggestions.
• Acceptance rate (lines) = accepted lines / suggested lines.

Charts:

• Acceptance rate trend (7-day moving average)
• Suggestions vs Acceptances stacked bars
• Breakdowns: By IDE and by Language (top items with Others bucket)

Note: Acceptance rate is a usefulness indicator, not a direct productivity measure.

Tests

• Run npm test to execute Jest tests for anonymization and metrics aggregation.

Screenshots

• Add screenshots of /tools and / here after first run.