SubVendor AI

Autonomous multi-agent B2B operations engine for trade subcontractors, powered by Gemini 1.5 & Google Cloud.

Comment

💡 Inspiration

Commercial trade subcontractors (HVAC, plumbing, electrical, roofing, general contracting) represent the lifeblood of small businesses in the physical economy. Yet, they lose over 15 hours per week manually reading complex B2B Requests for Quote (RFQs), calculating Bill of Materials (BOM) cost breakdowns, verifying commercial liability insurance thresholds ($10M+ limits), and drafting proposal emails.

Small trade contractors lack dedicated estimating and legal compliance departments. SubVendor AI was built to level the playing field, providing small trade contractors with an autonomous multi-agent back-office engine that handles B2B quotes, regulatory audits, and commercial proposals in seconds.

🤖 What It Does & How AI Operates the Business

SubVendor AI operates using a 3-tier autonomous agent workflow with complete observability and structured audit logging:

  1. EstimatorAgent-v1 (Gemini 1.5 Flash):
    • Analyzes raw jobsite RFQ specifications.
    • Calculates itemized line items (Labor hours, Materials, Permits, Equipment).
    • Computes subtotal, 8% tax, applies optimal B2B profit margins (e.g. 20%), and sets payment schedule terms (30% deposit upfront).
  2. ComplianceAgent-v1 (Gemini Risk Audit Engine):
    • Audits generated quotes against General Liability insurance thresholds ($10M limits), local municipal building safety codes, and required street closure permits.
    • Generates a 0–100 Compliance Score and highlights critical safety risks.
  3. OutreachAgent-v1 (Commercial B2B Desk):
    • Formulates tailored commercial proposals with embedded compliance certificates and 30% upfront deposit payment gateways.

☁️ Google Cloud & Gemini Integration

  • Gemini API (@google/genai v2.17.1): Uses gemini-1.5-flash with JSON Schema structured outputs (responseSchema) for deterministic, structured B2B pricing and risk audit JSON.
  • Google Cloud Run: Serverless container hosting the Next.js 16 App Router multi-agent pipeline.
  • Google Cloud Logging: Emits structured JSON audit logs (severity, agentRole, thoughtProcess, executionTimeMs, tokensUsed) for full operational observability.

📊 Business Model & Financial Transparency

  • Business Model:
    • SaaS Subscription: $199/month per trade contractor.
    • Transaction Fee: 1.5% processing fee on upfront job deposit payments (30% upfront deposit on quotes).
  • Financial Statement & P&L (Hackathon Window May 19 – Aug 17, 2026):
    • Actual Third-Party Revenue: $0.00 USD (In full compliance with Devpost financial verification rules, actual third-party revenue earned during the hackathon is $0.00 USD. The financial metrics displayed inside the product dashboard serve as an interactive SaaS forecast demonstrator).
    • Total Expenses: $0.00 USD (Google Cloud Free Tier & Google AI Studio Free Tier).
    • Net Profit: $0.00 USD.

🛠️ How We Built It

  • Frontend & UI: Next.js 16 (App Router), React 19, TypeScript, Vanilla TailwindCSS, Lucide Icons, with instant Light/Dark mode toggles.
  • AI SDK: Official @google/genai Node.js SDK.
  • Containerization & Deployment: Multi-stage Docker build deployed on Google Cloud Run.
  • Observability: Structured JSON console logging parsed by GCP Cloud Logging.

🧠 Challenges & Learnings

Designing structured JSON schemas for Gemini 1.5 Flash to generate realistic B2B itemized quotes required careful schema definition and temperature tuning (0.1 - 0.2) to prevent price hallucination. Ensuring high-contrast accessibility across both Light and Dark themes was also a priority for busy jobsite contractors.

🚀 What's Next for SubVendor AI

  • Integrating OCR/multimodal Gemini vision analysis for blueprint PDF uploads.
  • Connecting Stripe Connect for direct automated deposit payouts to trade contractors.
  • Expanding municipal building permit auto-filing integrations.

Built With

Share this project:

Updates