Examples

See how Quantlix acts as the control and verification layer around your models: policy at the boundary, optional retrieval, full traces, and eval scores—not just hosting weights.

End-to-end (illustrative)

A governed AI request—not just model hosting

Quantlix sits between your app and the model: contracts and policies first, optional retrieval, then execution, evals, and a full trace. Step through a typical path (simulated output).

Step 1 / 6User prompt

Structured input hits your deployment contract.

POST /run
{
  "deployment_id": "…",
  "input": {
    "prompt": "Summarize our refund policy for EU customers."
  }
}

Explore the platform

Inspect a trace

View run details, spans, and observability

Open dashboard →

See a failed eval regression

Compare versions, inspect failing examples

Open dashboard →

Try a grounded RAG workflow

Knowledge bases, retrieval, citations

Open dashboard →

Run a citation-aware answer flow

Workflows with retrieval and reranking

Open dashboard →

Enforcement scenarios

Schema, policy, and budget enforcement at the boundary.

Schema violation

Request blocked when input type or structure does not match the contract. No type coercion.

SCHEMA_INVALID_TYPE

Retry amplification

Block repeated retries that would multiply cost. Enforce retry ceiling per request.

BUDGET_RETRY_CEILING

Budget limit exceeded

Block when compute or request rate exceeds plan limits. 429 with Retry-After.

BUDGET_COMPUTE

Policy rejection

Block when policy rules fail: guardrails, feature drift, risk thresholds.

GUARDRAIL_BLOCK

Interactive enforcement examples

Run valid and invalid requests to see the enforcement decision.

Credit risk schema enforcement

Model: CreditRisk-v3

Contract

{
  "input_schema": {
    "type": "object",
    "required": ["age", "income", "score"],
    "properties": {
      "age": { "type": "number" },
      "income": { "type": "number" },
      "score": { "type": "number", "minimum": 0, "maximum": 850 }
    }
  }
}

Policy

Reject if score < 300. Warn if income < 20k.

Budget

Max 100 req/min, 3600s compute/month

Request input

Content policy and budget

Model: ContentMod-v2

Contract

{
  "input_schema": {
    "type": "object",
    "required": ["prompt"],
    "properties": {
      "prompt": { "type": "string", "maxLength": 2000 }
    }
  }
}

Policy

Block PII, block harmful content. Rate limit per user.

Budget

50 req/min, 1800s compute/month

Request input

Public example deployments

Reference deployments for trying the runtime API; metrics are illustrative. Use them with your API key to exercise enforcement and tracing—the demo model is a small stand-in for your own stack.

Model	Status	Deploy time	Latency	Cost	Metrics	Endpoint
qx-example (demo deployment) Reference deployment for contracts, policy, traces, and evals—backed by a small open model for fast iteration.	CPU	~12s	<50ms	Free tier	Requests today: 2,341 Blocked: 12%	API docs
qx-example GPU Same governed demo path on GPU when you need lower latency (plan limits apply).	GPU	~18s	<30ms	Pro: €0.50/hr	Requests today: 892 Blocked: 8%	API docs

Try a governed request end to end

Sign up, open Quick Start (deployment → policy pack → Playground run), or deploy the demo model from the dashboard and use your deployment ID in the cURL above. You'll see enforcement decisions, traces, and eval hooks on real runs—not only raw model output.