Data Sync Overview

The monorepo has a packages/data-sync package (13 tests) responsible for pulling external data into the build pipeline. Different apps have different data sources.

LLM Timeline data sync

bun run sync

Pulls from two sources:

• Curated: 785 models from a Google Sheets source (maintained by Duyet)
• Epoch AI: 3156 models from the Epoch AI dataset

The last sync ran 2026-03-25, producing 3937 unique models after deduplication (4 duplicates removed). Year range expanded to 1950–2026.

Data lands in apps/llm-timeline/lib/data.ts. That file is out of scope for duyetbot — it contains research facts.

CCUsage (insights)

The insights app shows Claude API usage statistics sourced from CCUsage. Data is synced to a static JSON or API that the insights app reads at build time.

In Cycle 4, ccusage-utils was a 674-line file with extensive duplication. It was deduplicated as part of the improvement sweep.

Blog posts

Blog posts are not synced — they live as local .md files in apps/blog/_posts/. The packages/libs post-loading utilities (readPublicJson(), getPostBySlug()) process them at build time.

Wakatime

Wakatime coding activity data is referenced in the monorepo but specific sync details are not captured in current memory. It likely feeds apps/insights or apps/ai-percentage.

Test coverage

packages/data-sync has 13 tests. The packages/api package (8 tests) handles external API calls used by some data sync flows.