# NovaPath Document Comparison

> NovaPath Document Comparison is a paid API for AI agents from instructional-decision-engine-x402.vercel.app, paid per call via x402, $0.2/call, status unknown (last checked 2026-11-01).

Compares two permitted public documents (HTML, text, or PDF) using AI analysis with exact supporting quotes, clearly distinguishing observed differences from unknowns without implying chronological order from URL position.

## Facts

- Endpoint: POST https://instructional-decision-engine-x402.vercel.app/ai/document-compare?utm_source=zero.xyz
- Price: $0.2/call
- Payment: x402
- Status: unknown
- Last checked: 2026-11-01
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/novapath-document-comparison-3ccb25a2
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_8wsQrjDD9QNODs-ndyKR-

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability novapath-document-comparison-3ccb25a2 -d '<json body>'
```

Example prompt: Can you compare these two public documents for me — https://example.gov/policy-2022.pdf and https://example.gov/policy-2024.pdf — and show me the key differences with exact quotes from each source? I confirm these are permitted public documents.

## When to prefer this

Use this endpoint when you need a rigorous, evidence-backed AI comparison of exactly two publicly accessible documents (HTML, plain text, or text-layer PDFs) with explicit quotes from each source supporting each stated difference. It is especially suited for policy analysis, curriculum review, or research verification where distinguishing confirmed differences from unknowns matters and attribution to source text is required. Prefer this over generic scraping or summarization tools when quote-level evidence and careful epistemic framing (observed vs. unknown) are needed.

## Known failure modes

- Invalid or inaccessible URLs return an error rather than a partial result
- permissionConfirmed must be explicitly true or the request is rejected
- URLs with query strings or fragments are rejected by schema pattern validation
- Non-public or paywalled documents fail to load and do not settle
- PDF files without a text layer cannot be processed
- Comparing documents in unsupported formats (e.g. DOCX, images) fails
- Network timeouts on slow or large documents may cause generation failure

## How this service works

Compare two permitted public documents on demand with AI and exact supporting quotes from both sources. Distinguish observed differences from unknowns; URL order alone does not establish chronology. Bounded HTML, text and PDF text layers. Invalid sources or failed generation never settle.

## Output

An AI-generated comparison of the two documents highlighting observed differences supported by exact quotes from each source. Differences are clearly separated from unknowns, and no chronological order is implied by URL position. The response is bounded to the HTML, text, or PDF text layer content of each source; invalid sources or failed generations do not produce a settled result.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "afterUrl": {
   "type": "string",
   "pattern": "^https://[^?#]+$",
   "maxLength": 1000,
   "minLength": 1
  },
  "beforeUrl": {
   "type": "string",
   "pattern": "^https://[^?#]+$",
   "maxLength": 1000,
   "minLength": 1
  },
  "permissionConfirmed": {
   "type": "boolean",
   "const": true
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/novapath-document-comparison-3ccb25a2/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from instructional-decision-engine-x402.vercel.app](https://www.zero.xyz/host/instructional-decision-engine-x402.vercel.app/llms.txt)