URL, Repo, API, Perf: Four Ways to Analyse Your Application

Not every quality question is the same. "Does the deployed site work?" is different from "Are my API contracts honoured?" which is different from "Will it handle 100 concurrent users?"

QuantumVerifi has four analysis modes, each targeting a different layer.

URL Mode — Test the Deployed Site#

Input: A live URL (e.g., https://your-app.com)

What happens:

1. AI crawls your site, discovering pages, forms, navigation paths
2. Generates Playwright E2E tests for critical user journeys
3. Executes tests in sandboxed Chromium browsers
4. Runs security scanning (OWASP Top 10)
5. Checks accessibility (WCAG 2.1 AA)
6. Measures performance (Lighthouse + Core Web Vitals)

Output: A composite grade (A-F) across all four dimensions, with per-dimension breakdowns.

When to use: After deploying to staging or production. This is the "does it actually work for users?" check.

Example:

Bash
Copy
curl -X POST https://quantumverifi.com/api/v1/analyses \
  -H "X-API-Key: $QV_API_KEY" \
  -d '{"live_url": "https://staging.your-app.com"}'

Repo Mode — Test the Codebase#

Input: A Git repository URL (e.g., https://github.com/org/repo)

What happens:

1. Clones the repo (shallow, temp directory)
2. Detects language, framework, entry points
3. Assesses current test coverage and health
4. AI generates unit and integration tests
5. Validates: syntax check → compile → execute
6. Self-heals failing tests (up to 3 attempts)
7. Generates documentation if missing

Output: Test files, security findings, coverage report, documentation.

When to use: During development, before merging a PR, or to bootstrap test coverage for a project that has none.

API Mode — Test the Contracts#

Input: An API spec URL or endpoint (REST, GraphQL, gRPC, WebSocket, or SOAP)

What happens:

1. Auto-detects the protocol from the spec content or URL
2. Discovers endpoints via protocol-specific mechanisms (OpenAPI parsing, GraphQL introspection, gRPC reflection, WSDL parsing)
3. AI generates protocol-aware contract tests (Jest or Pytest)
4. Executes tests against the live API
5. Runs protocol-specific security analysis
6. Self-heals failing tests with protocol-aware error clustering

Supported protocols:

Output: API contract test results with pass/fail per endpoint, security findings, and protocol-specific coverage.

When to use: When you have a documented API and want to verify it matches the spec. Catches schema drift, missing required fields, and undocumented error responses — across any protocol.

Example (REST):

JSON
Copy
{
  "api_spec": "https://api.your-app.com/openapi.json",
  "api_base_url": "https://api.your-app.com",
  "execute_api_tests": true
}

Example (GraphQL):

JSON
Copy
{
  "api_spec": "https://api.your-app.com/graphql",
  "api_protocol": "graphql",
  "execute_api_tests": true
}

Read the deep dive on multi-protocol API testing for architecture details and code examples.

Perf Mode — Test Under Load#

Input: A target URL

What happens:

1. AI selects weighted test targets based on site structure
2. Generates k6 load test scripts
3. Executes in sandboxed environment with configurable profiles
4. Measures latency (p50, p95, p99), throughput, error rates
5. Compares against SLA thresholds

Output: Performance grade with detailed metrics.

Profiles:

When to use: Before a launch, after infrastructure changes, or as a recurring check to catch performance regressions.

Combining Modes#

The quality gate API (POST /api/v1/cicd/gate) works with any mode. Run a URL analysis after deploy, then gate your pipeline on the result. Run perf tests weekly and alert if the grade drops below B.

Each mode produces a grade. The grade feeds into the same quality gate. One API, four perspectives on quality.

Getting Started#

1. Start with URL mode — point it at your staging environment
2. Add API mode if you have an OpenAPI spec
3. Add Perf mode before launches or as a weekly job
4. Use Repo mode to bootstrap test coverage for under-tested codebases

Every mode is available on every plan, including the free Developer tier (5 analyses/month). The only gate is credits — not features.