Skip to content

progress.md

Latest commit

 

History

History
30 lines (28 loc) · 3.72 KB

progress.md

File metadata and controls

30 lines (28 loc) · 3.72 KB
Raw

Progress

2026-03-19 Session

  • Reviewed planning skill instructions and initialized planning files in the project root.
  • Confirmed the current code still has legacy auth, leaderboard metric naming, and GitHub ranking limits that do not match the new requirements.
  • Read the auth, leaderboard, persistence, and GitHub ranking codepaths end-to-end to identify where old token semantics are still embedded.
  • Probed live GitHub REST endpoints to validate response shape and rate limits for repository search, contributors, and contributor statistics.
  • Implemented canonical host redirect logic, added GitHub auth provider wiring, and exposed Google/GitHub login choices in the main UI.
  • Replaced the old yearly-token semantics with shared AI-coding-era math across mock data, proofs, repo intake, hero ticker, and leaderboard rendering.
  • Added a new global GitHub ranking API route and rewrote the ranking service to search recent active repositories, blend in benchmark repos, and aggregate up to 500 contributors.
  • Verified npm run lint, npm run build, and vercel build after the refactor.
  • Verified the built serverless output loads under Node and can successfully evaluate vercel/next.js, producing month-burn numbers in the multi-billion range and cumulative totals since January 2025.
  • Verified the full global snapshot path completes after adding per-repo fallback; without GITHUB_TOKEN it now degrades gracefully to a smaller snapshot instead of failing outright.
  • Added GITHUB_TOKEN to the Vercel project environments (Production, Preview, Development) and triggered a new production deployment.
  • Confirmed the deployment reached Ready and was re-aliased to https://token-forbes.vercel.app.
  • Replaced the client-side Gemini SDK path with a server-side OpenRouter proxy route at api/ai, keeping the API key out of the browser bundle.
  • Configured OPENROUTER_API_KEY, OPENROUTER_BASE_URL, and OPENROUTER_MODEL in Vercel and redeployed production.
  • Verified direct OpenRouter chat-completions calls succeed with stepfun/step-3.5-flash:free; after increasing token budget for the reasoning model, the API returns normal assistant text instead of empty content.

2026-03-21 Session

  • Restored planning context and extended the existing plan to cover a dedicated mobile responsive pass.
  • Identified the main mobile risk areas before editing: header/nav, hero ticker, leaderboard density, proof flows, and wide dashboard cards.
  • Read the key desktop-first components and confirmed the highest-impact responsive fixes should target table-like layouts, the floating chatbot panel, and top-level spacing/navigation behavior.
  • Implemented a mobile-friendly header, smaller hero/ticker spacing, responsive chatbot sizing, and reduced paddings across major narrative sections.
  • Reworked the leaderboard and GitHub audit contributor list so phones get card layouts while desktop keeps the existing table presentation.
  • Adjusted proof/share flows for better button stacking and spacing on smaller screens, then verified npm run lint, npm run build, and vercel build.
  • Reworked the proof badge layout so ultra-large token values wrap onto multiple lines and the certificate grows vertically with content instead of clipping at a fixed aspect ratio.
  • Verified the long-token proof badge fix with npm run lint and npm run build.
  • Audited the current data stack end-to-end across README, methodology copy, GitHub ranking services, enterprise ranking services, proof persistence, and leaderboard merge logic to distinguish implemented sources from aspirational methodology text.
  • Added a concrete README TODO roadmap focused on objective accuracy, source provenance, calibration, verification, identity resolution, and long-term methodology governance.