Compreface

{
  "service_type": "platform",
  "base_url": "https://exadel.com/accelerator-showcase/compreface",
  "auth_method": "none",
  "auth_config": {},
  "endpoints": [],
  "pricing_model": {
    "type": "free",
    "details": {
      "note": "Free and open-source; self-hosted deployment model"
    }
  },
  "rate_limits": {},
  "capabilities": [
    "face recognition",
    "face detection",
    "face verification",
    "face identification",
    "open-source",
    "self-hosted deployment",
    "docker-based installation",
    "REST API",
    "role-based access control",
    "user management"
  ],
  "raw_analysis": "CompreFace by Exadel is a free, open-source face recognition platform designed for developers and organizations that need to add facial recognition capabilities to their applications. It is distributed as a self-hosted Docker-based solution, so its base URL is not a public API host but rather its showcase page on the Exadel site. The real API becomes available once the user deploys CompreFace locally or on their infrastructure; the API's base URL is user-specific (e.g., http://localhost:8000 or a custom domain).\n\n**What it does:** CompreFace provides a REST API for face detection, verification, identification, and recognition. It supports multiple face recognition models (including state-of-the-art ones like FaceNet, ArcFace, and InsightFace) and allows users to create applications, manage face collections, and perform recognition tasks. It also includes a web UI for management and testing.\n\n**Who it's for:** Developers, data scientists, and organizations building applications that require face recognition without relying on third-party cloud services. Its open-source nature makes it attractive for privacy-conscious users and those who want to avoid vendor lock-in.\n\n**Maturity:** CompreFace is a mature project with active development, regular updates, and a growing community. It is backed by Exadel, a software engineering company, and has been available since 2020. It supports horizontal scaling, GPU acceleration, and various deployment options.\n\n**Integrations:** Being a self-hosted tool, integrations are primarily through its REST API, which can be consumed by any HTTP client. It also provides plugins and examples for integration with other systems. Since it's open-source, users can modify and extend it as needed.\n\n**API specifics:** The API is RESTful and uses JSON. Authentication is typically done via API keys or JWT tokens if enabled, but by default, the deployment might allow unauthenticated access for initial setup. However, the platform supports role-based access control, so authentication methods can be configured. The base URL is determined by the deployment. Rate limits are not predefined; they depend on the hardware and configuration.\n\n**Pricing:** CompreFace is completely free to use and modify under the Apache 2.0 license. There are no paid tiers or cloud-hosted options provided by Exadel directly, though third-party cloud providers may offer managed instances.\n\n**Note:** The provided URL is a showcase page, not the actual API endpoint. For API access, users must deploy the software themselves and refer to the API documentation."
}

1/3 tests passed

{
  "overall": 58,
  "dimensions": {
    "token_efficiency": 7.5,
    "first_try_success": 6.0,
    "response_parseability": 6.0,
    "error_clarity": 5.5,
    "doc_quality": 5.5,
    "auth_simplicity": 8.5,
    "latency": 8.5,
    "consistency": 6.0
  },
  "pricing_normalized": {
    "model": "free_open_source",
    "cost_per_month_usd": 0,
    "notes": "Free and open-source; self-hosted deployment. No SaaS pricing tiers. Total cost of ownership falls on user infrastructure (Docker, compute, storage)."
  },
  "issues": [
    "No robots.txt or llms.txt served (both 404) — agents and crawlers lack a machine-readable entry point or guidance, hurting discoverability and agent onboarding",
    "No security headers detected on responses — weaker signal for production-grade reliability/trust",
    "Marketing site is hosted on exadel.com (agency site), not a dedicated product domain — reduces clarity of value prop and product identity for agents summarizing to users",
    "Self-hosted/open-source model means no instant API key flow; agent cannot onboard a user to a working endpoint in minutes without infra setup",
    "No confirmed structured API response examples or OpenAPI spec in check data, so response_parseability is inferred rather than verified"
  ],
  "recommendations": [
    "Publish a dedicated product domain or clearly scoped product landing page with a crisp one-line value prop (e.g., 'open-source face recognition API, self-hosted in one Docker command')",
    "Add /robots.txt and /llms.txt with concise capability, deployment, and API-doc pointers so agents can onboard users efficiently",
    "Ship an OpenAPI/Postman spec plus example JSON request/response payloads to maximize agent parseability and first-try integration success",
    "Provide a quickstart that gets a user from clone to a working /verify endpoint in under 5 minutes (docker compose up single command), with curl examples",
    "Add security headers and a public status/roadmap page to strengthen consistency and reliability perception",
    "Document hard limits explicitly (image size, latency, GPU requirements, concurrency) in a machine-readable 'limitations' section to improve error_clarity",
    "Offer an optional hosted/demo API sandbox so agents can demonstrate a working call before users commit to self-hosting"
  ]
}