Audits whether autonomous agents can complete real buyer flows.
Rendered crawl, operability checks, simulated navigation, evidence reports.
openingress.dev · LinkedIn · Features
OpenIngress crawls rendered pages, runs static operability checks, simulates agent-style navigation, and produces evidence-heavy reports with fix prompts and remediation exports.
This repository includes the open-source backend and an optional Vue frontend. The backend is the main artifact. The hosted site at openingress.dev offers a demo report and contact flow.
AEO and GEO help AI systems find and cite your store. That does not mean an agent can shop it.
| Layer | What it solves |
|---|---|
| AEO / GEO | |
| Brand visibility in AI answers — "this store sells workout gear" | |
| Intra-site navigation | |
| Machine-readable commerce on the live site — sizes in stock, cart state, checkout handoff |
OpenIngress audits the second layer: whether an agent can traverse catalog → product → cart → checkout, not whether your marketing pages rank in an answer box. More detail in docs/ARCHITECTURE.md.
Rendered crawl— screenshots, HTML, and accessibility trees** Static operability**— label coverage, DOM health,llms.txt
, hydration gapsAgent exploration— heuristic or LLM-guided navigation through discovery → product → cart → checkout** Buyer-flow evidence**— funnel steps with pass/fail, blockers, and screenshot replay** Readiness score**— headline metric plus revenue-ranked blockers** Fix packs**— implementation prompts and Markdown/PDF exports** Dashboard**— Vue app for audits, runs, reports, and fixes
Sample report: openingress.dev or /reports/demo
locally (no backend required).
Requirements: Python 3.10+, Node.js 18+, Chromium (via Playwright)
git clone https://github.com/Open-Ingress/OpenIngress.git
cd OpenIngress
cp .env.example .env
cp backend/.env.example backend/.env
cp frontend/.env.example frontend/.env
make install
make backend # terminal 1 → http://127.0.0.1:5055
make frontend # terminal 2 → http://localhost:5175
Open ** http://localhost:5175** →
New audit→ enter a storefront URL.
Browser automation uses Playwright for the crawl layer. Exploration and scoring sit on top: accessibility-tree inspection, navigation-graph analysis, operability checks, and report generation.
flowchart LR
URL[Storefront URL] --> Crawl[Rendered crawl]
Crawl --> Static[Static checks]
Crawl --> Explore[Agent exploration]
Static --> Report[Readiness report]
Explore --> Report
Report --> Fixes[Fix prompts + exports]
Crawl— discover pages, capture HTML, screenshots, and aria snapshots** Analyze**— score operability, map navigation, detect commerce funnel steps** Explore**— agent-style clicks through the buyer journey** Report**— scores, blockers, traces, and remediation exports
| Path | Purpose |
|---|---|
backend/ |
frontend/
docs/
frontend/public/demo-report/
/reports/demo
Dockerfile
| Guide | Description |
|---|---|
API referenceDeploymentDevelopment| Variable | Where | Purpose |
|---|---|---|
LLM_API_KEY |
backend/.env |
OpenAI-compatible key for agent exploration |
LLM_BASE_URL |
backend/.env |
Any compatible endpoint (Ollama, Azure, etc.) |
VITE_API_URL |
frontend/.env |
Backend API base (http://127.0.0.1:5055/api/ingress ) |
Full reference: docs/SELF_HOST.md
make test
make smoke # backend must be running
See CONTRIBUTING.md and docs/DEVELOPMENT.md.