Akashic Computer

Akashic Computer is an open-weight model atlas for navigating model families, parameter sizes, variants, quantized artifacts, benchmarks, runtime support, and hardware fit.

The product is designed around the path a user actually takes:

Family -> Release -> Size -> Variant -> Artifact -> Runtime/Benchmark/Fit

Example:

Qwen -> Qwen 3.6 -> 27B -> Instruct -> NVFP4 -> 18-24 GB VRAM

Stack

• Next.js App Router
• TypeScript
• Tailwind CSS
• Convex backend/database
• Hugeicons

Local Development

npm install
npm run dev

Open http://localhost:3000.

Website analytics

Cloudflare Web Analytics is installed in src/app/layout.tsx using the site's public beacon token. It loads after hydration, only in production builds on akashic.computer; localhost and preview hostnames do not load the beacon. Cloudflare handles client-side route tracking without additional page-view events.

Use manual JS snippet installation for this site in Cloudflare, rather than also enabling automatic injection. After deployment, check Web Analytics for incoming page views and verify navigation between catalog pages. These metrics measure website activity separately from Convex function calls and controller polling.

Convex

Convex is the only runtime source of truth for the public catalog.

To create and sync a Convex dev deployment:

npx convex dev

The first run will open the Convex login/project setup flow.

Configure monitored sources and run the first audit after setting an operator secret:

npx convex env set CATALOG_ADMIN_SECRET '<random-secret>'
npx convex run admin:syncSourceConfig '{"secret":"<random-secret>"}'
npx convex run admin:runAudit '{"secret":"<random-secret>"}'

See docs/catalog-sync.md for production deployment, Hugging Face webhooks, daily reconciliation, validation rules, and operational checks.

Verification

npm test
npm run typecheck
npm run lint
npm run build
npm run test:e2e

Personal local workspace

The Akashic Computer companion connects your LAN fleet to the signed-in website and desktop MCP App. Public model discovery remains at the homepage. Account conversations and job progress sync through Convex; inference runs on your hardware. Existing device-only sessions stay local.

See connected workspace setup for GitHub authentication, enrollment, deployment flags, and rollback.

Planning Docs

The optional deployment comparison workflow adds exact quantized files, component memory estimates, GitHub account profiles, private KLD/Top-1 and performance imports, saved decisions and export-only pinned recipes. Enable it with NEXT_PUBLIC_DEPLOYMENT_COMPARISON=true after preparing the backend. Public signup is separately controlled by the server's WORKSPACE_ACCESS_MODE=github. Automated benchmarking is deferred.

• PRODUCT.md
• DESIGN.md
• docs/product-brief.md
• docs/design-agent-prompt.md