GovPath https://govpath-tawny.vercel.app/

GovPath is an interview-ready MVP that maps a startup or defense-relevant company to government adoption pathways using source-grounded defense budget, acquisition, and strategy documents.

The app accepts:

• pasted company text
• uploaded pitch deck PDFs
• a visible but intentionally stubbed website URL field

Every recommendation must be traceable to the provided PDF source set. If support is missing, the app returns: Not found in the provided source documents.

Source Documents

Place these PDFs in data/sources/:

1. 2026 DoW Directory Update 7.pdf
2. DoD FY 26 Procurement Budget Estimates.pdf
3. FY26-30 DoW Strategic Plan 6Apr2026 r18.pdf
4. FY2026_Budget_Request_Overview_Book.pdf

The included source manifest documents what each file is used for and what known limitation should be surfaced in the UI.

How Matching Works

1. GovPath extracts company keywords, capability tags, mission areas, and context from pasted text or an uploaded pitch deck.
2. A local indexing script parses the four PDFs, chunks text by page sections, and stores page-aware retrieval data in data/index/source-index.json.
3. The app runs keyword and BM25-style retrieval over that local index.
4. Recommendations, budget themes, and outreach pathways are generated only from retrieved passages.
5. Results are labeled as direct_source_match, inferred_from_source, or not_found.

Local Setup

npm install
cp "<path-to>/2026 DoW Directory Update 7.pdf" data/sources/
cp "<path-to>/DoD FY 26 Procurement Budget Estimates.pdf" data/sources/
cp "<path-to>/FY26-30 DoW Strategic Plan 6Apr2026 r18.pdf" data/sources/
cp "<path-to>/FY2026_Budget_Request_Overview_Book.pdf" data/sources/
npm run index
npm run dev

Open http://localhost:3000.

Build the Local Index

npm run index

This regenerates data/index/source-index.json from the PDFs in data/sources.

Run Locally

npm run dev

Production Build

npm run build
npm run start

Deploy to Vercel

GovPath is designed as a standalone Vercel project.

vercel

For a production deployment:

vercel --prod

Push to GitHub

Initialize or verify the nested standalone repo inside govpath, then push it to its own GitHub repository:

git init
git add .
git commit -m "Initial GovPath MVP"
git branch -M main
git remote add origin <YOUR_GITHUB_REPO_URL>
git push -u origin main

Project Structure

• app/ - routes, layout, homepage, and API route
• components/ - input flow, results dashboard, methodology, citations, and source library UI
• lib/source-ingestion/ - PDF source manifest and parsing helpers
• lib/retrieval/ - local index loading, tokenization, and search
• lib/analysis/ - company parsing and report generation
• data/sources/ - authoritative PDFs
• data/index/ - generated local retrieval index
• types/ - GovPath data contracts

Limitations

• Website crawling is intentionally stubbed for the MVP.
• The app uses deterministic heuristics rather than an LLM by default.
• The DoW Directory is not an official U.S. government publication and may be outdated.
• Budget documents show alignment and procurement context, not guaranteed funding availability.

Future Improvements

• Optional retrieval-constrained LLM synthesis behind environment variables
• More robust office and stakeholder extraction from PDF structure
• Website crawling and ingestion
• Better uploaded pitch deck summarization and slide-aware extraction