https://github.com/asmuelle/tenderhook
Procurement radar for SMB government contractors: a daily agent over SAM.gov, state/local portals, and buyer budget docs maintains living dossiers per agency and opportunity — matching tenders to your capabilities, red-lining amendments, flagging expiring incumbents.
https://github.com/asmuelle/tenderhook
government-contracting govtech llm monitoring procurement rfp saas
Last synced: about 19 hours ago
JSON representation
Procurement radar for SMB government contractors: a daily agent over SAM.gov, state/local portals, and buyer budget docs maintains living dossiers per agency and opportunity — matching tenders to your capabilities, red-lining amendments, flagging expiring incumbents.
- Host: GitHub
- URL: https://github.com/asmuelle/tenderhook
- Owner: asmuelle
- Created: 2026-06-11T16:51:20.000Z (26 days ago)
- Default Branch: main
- Last Pushed: 2026-07-05T14:50:42.000Z (2 days ago)
- Last Synced: 2026-07-05T16:10:14.725Z (2 days ago)
- Topics: government-contracting, govtech, llm, monitoring, procurement, rfp, saas
- Language: TypeScript
- Homepage: https://asmuelle.github.io/tenderhook/
- Size: 762 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Security: SECURITY.md
- Agents: AGENTS.md
Awesome Lists containing this project
README
# Tenderhook
[](https://github.com/asmuelle/tenderhook/actions/workflows/ci.yml)
> Procurement radar for SMB government contractors: a daily agent over SAM.gov, state/local portals, and buyer budget documents maintains living dossiers per agency and opportunity — semantically matching tenders to your real capabilities, red-lining every amendment, and flagging incumbent contracts expiring 6-12 months out.
## Concept
Procurement radar for SMB government contractors: a daily agent over SAM.gov, state/local portals, and buyer budget documents maintains living dossiers per agency and opportunity — semantically matching tenders to your real capabilities, red-lining every amendment, and flagging incumbent contracts expiring 6-12 months out.
## Auto-Research Mechanic
Daily monitoring of SAM.gov APIs, state and major-city procurement portals, EU TED, grants.gov — plus the pre-RFP layer incumbents ignore at this price: agency budget documents, board/council minutes, capital-improvement plans that telegraph solicitations months early, and award databases revealing incumbent contracts nearing expiration. Capability profiles (NAICS codes + natural-language description of what the firm actually does) drive cheap-model semantic matching far beyond keywords ('cloud migration under an IT modernization umbrella' matches when keywords don't). Living dossier per tracked opportunity — every amendment red-lined against the prior version, Q&A addenda summarized — and per target agency (buying history, incumbent map, upcoming expirations). All sources are public government documents: deep-linked, verifiable, free of the AI-crawler-blocking problem.
## Product Surface
Email digest first (morning opportunity review is the BD ritual in this industry) + web SaaS for the pipeline board, agency dossiers, and amendment red-lines + Slack alerts for deadline changes on tracked bids.
## Tech Stack & Unit Economics
### Ingestion
* SAM.gov daily extracts + Get Opportunities API via non-federal system account (1,000 req/day tier; 10/day without a role — extracts are the real backbone)
* FPDS-NG/USASpending bulk downloads for the award + incumbent-expiration graph
* grants.gov XML extract, EU TED API
* SLED via per-platform adapters (Playwright workers with platform-level adapters for Bonfire, BidNet Direct, Periscope S2G, OpenGov, Ionwave, PlanetBids covers ~70% of volume; long tail is bespoke scrapers)
* Granicus/Legistar APIs for council minutes and agendas.
### Storage/diffing
* Postgres + pgvector
* S3 document store with content-hash versioning per solicitation; amendment detection is deterministic (poll source-of-record, checksum, pdf-to-text via Docling/Reducto, structural text diff) with LLM used only to summarize changed hunks — never to decide whether something changed.
### Models
* embeddings (voyage-3 or text-embedding-3-large) for candidate retrieval
* Haiku 4.5 / Gemini Flash-class for match scoring and digest copy
* Sonnet 4.6-class only for capture briefs and agency dossier synthesis on the top tier.
### Citation grounding
* every digest claim deep-links to source doc + page anchor
* quoted text string-verified against extracted text before send (no unverifiable quotes shipped)
### Orchestration
* boring queue workers (Temporal or SQS + cron) for the daily batch — no agent framework needed
* agentic browsing reserved for credentialed portal retrieval.