SpecForge is a local-first workflow studio for turning open-source trend evidence into a spec-driven portfolio project.
It was designed as a GitHub-profile piece: polished enough to demo, technical enough to inspect, and grounded in current high-star project patterns instead of generic app ideas.
Live demo: https://nripankadas07.github.io/specforge/
Star SpecForge if you want a compact, inspectable example of:
- turning OSS trend evidence into product direction without runtime scraping;
- ranking project inspirations with a transparent scoring model;
- simulating a guarded spec-first workflow before building;
- exporting a README-ready blueprint from product and engineering signals;
- building a polished React/TypeScript portfolio project with tests and CI.
The highest-signal repositories I verified on 2026-06-13 cluster around four durable ideas:
- Build-your-own learning and technical depth.
- Curated developer knowledge and awesome-list discovery.
- Visual AI workflow builders and agent orchestration.
- Spec-driven development, guardrails, and MCP-style tooling.
SpecForge combines those ideas into a runnable product: pick high-signal inspirations, simulate a spec-first build workflow, inspect confidence and risk, then export a README-ready blueprint.
- Trend radar backed by a fixed, source-linked dataset of high-star GitHub repositories.
- Repository scoring engine that weighs stars, technical depth, demo appeal, feasibility, moat, and risk.
- Interactive workflow graph with deterministic event simulation.
- Guardrail and evidence toggles that change the ship score and node status.
- Exportable Markdown blueprint for project planning or README drafts.
- Pure TypeScript scoring and workflow modules covered by Vitest tests.
- Responsive dashboard UI with real repository avatars and no API key requirement.
| Layer | File | Why It Matters |
|---|---|---|
| Evidence dataset | src/data/repositories.ts | Fixed, source-linked project signals keep demos stable and reviewable. |
| Scoring model | src/lib/scoring.ts | Converts stars, technical depth, demo appeal, feasibility, moat, and risk into a portfolio score. |
| Workflow engine | src/lib/workflow.ts | Generates deterministic workflow nodes, events, risks, confidence, and ship score. |
| Blueprint export | src/lib/exporters.ts | Produces a Markdown project plan from the selected evidence and workflow. |
| Product shell | src/App.tsx | Wires the dashboard, selectors, graph, event timeline, and export flow. |
- React 19
- TypeScript 6
- Vite 8
- Vitest
- Lucide React
npm install
npm run devQuality gates:
npm run lint
npm run test
npm run buildsrc/
data/
repositories.ts verified source dataset
lib/
scoring.ts portfolio scoring and ranking logic
workflow.ts deterministic workflow simulator
exporters.ts Markdown export utilities
App.tsx product shell and interaction wiring
App.css dashboard visual system
Read more in docs/ARCHITECTURE.md.
- GitHub high-star baseline: https://api.github.com/search/repositories?q=stars:%3E100000&sort=stars&order=desc
- GitHub AI repository query: https://api.github.com/search/repositories?q=topic:ai%20stars:%3E20000&sort=stars&order=desc
- GitHub MCP repository query: https://api.github.com/search/repositories?q=topic:mcp%20stars:%3E5000&sort=stars&order=desc
- GitHub Trending: https://github.com/trending
- OSSInsight AI trending: https://ossinsight.io/trending/ai
- Hacker News spec-driven workflow discussion: https://news.ycombinator.com/item?id=48413629
- Hacker News composable agent discussion: https://news.ycombinator.com/item?id=47350516
Details are in docs/RESEARCH.md.
Discussions are open for scoring ideas, workflow suggestions, and project planning use cases: https://github.com/nripankadas07/specforge/discussions
High-value next improvements:
- add scoring profiles for backend, AI, data, and frontend portfolios;
- add a CLI exporter for generating blueprints from a checked-in config;
- add an optional live GitHub fetcher with cached fallback data;
- add repeatable screenshot and demo-video generation;
- publish example blueprints for real portfolio projects.
The public launch kit is in docs/LAUNCH_KIT.md.
SpecForge is intentionally local-first. It does not call LLM APIs, scrape live social media, or require credentials at runtime. The repository data is fixed inside the app so demos are stable, repeatable, and reviewable.
MIT