Savant is the enterprise platform for codifying expertise as governed, measurable, reusable skills.
- a
pnpm+ Turborepo monorepo root - a fully working Next.js App Router frontend in
apps/web - tenant-aware control-plane routes and DB-backed read surfaces in
apps/web - GitHub-first repository provisioning, scaffold/apply commits, and inline indexing in
apps/web - initial PostgreSQL control-plane schema and onboarding schema foundations under
db/ - spec-aligned placeholder folders for future API, worker, packages, skills, and infra
- CI that validates linting, typechecking, and production builds
- a Vercel runtime assessment kit for comparing Rust and Python before backend work begins
- extraction into dedicated
apps/apiandapps/workerruntimes until the Rust vs Python on Vercel assessment is complete - multi-provider repository write parity, background indexing jobs, and webhook ingestion
- BYO AI execution flows, eval execution pipelines, and release orchestration
- connector orchestration, signed bundle generation, and advanced sync agents
The runtime decision is documented in docs/architecture/runtime-assessment.md.
Tenant-authored skill content is expected to live in a provider-agnostic external Git environment; that architecture is documented in docs/architecture/tenant-skill-storage.md.
savant/
apps/
web/ # Next.js frontend, ready for feature work
api/ # placeholder until runtime assessment is complete
worker/ # placeholder until runtime assessment is complete
packages/
ui/
types/
schemas/
config/
db/
skills/
docs/
infra/
pnpm install
pnpm devUseful validation commands:
pnpm lint
pnpm lint:fix
pnpm test
pnpm security:scan
pnpm typecheck
pnpm build
pnpm checkFor a full local pre-commit sweep that installs dependencies, runs the security checks, fixes lint issues, runs tests, and finishes with the production validation gate:
bash ./precommit.sh
pwsh -NoProfile -ExecutionPolicy Bypass -File ./precommit.ps1- Core platform spec:
docs/product-specs/spec-core.md - Runtime assessment gate:
docs/architecture/runtime-assessment.md - Runtime assessment artifacts:
infra/vercel/runtime-assessment/ - Tenant skill storage architecture:
docs/architecture/tenant-skill-storage.md
apps/web is the initial admin console, tenant shell, and BFF for the first integrated beta flows.
The repo is intentionally not committed to Python, Rust, or any other backend runtime yet.
The skills/ directory in this repo is for templates and local reference assets, not the primary store for tenant-authored production skills.
The accepted beta runtime boundary is documented in docs/architecture/adr-0001-beta-runtime-boundary.md.