The Gauntletslasso.com
The Gauntlet · slasso.com

Get your RAG certified

A reproducible benchmark run of your retrieval stack, scored against published thresholds, with a scorecard you can link in sales conversations.

Submit your embedding endpoint

Your endpoint must accept POST { "texts": ["..."] } and return { "vectors": [[...], ...] } over HTTPS. We grade it on the same harness that gates our own deploys.

Standard tier · from $1,500 per run (Operator pricing gate) · results typically in minutes once the worker picks up your submission

Integrate directly

Call the certification API from your CI or staging environment. Submit an HTTPS endpoint; the worker embeds a fixed public catalog slice through your service and grades it with the same metric code that gates our own deploys. Typical turnaround: minutes.

POST /v1/certify
Authorization: Bearer <workspace-api-key>
{"endpointUrl": "https://your-service.com/embed"}

# Your endpoint contract (required):
POST {"texts": ["...", "..."]}
→ {"vectors": [[...], [...]]}

Poll GET /v1/certify/{submissionId} for status and scorecard. Payment checkout attaches at the Operator gate; the measured scorecard is independent of billing state.

1 · Scope

You describe the stack (embedder, index, reranker) and pick the tier. We agree the eval set: yours under NDA, or our public suites.

2 · Run

We execute the benchmark on the shared harness (nDCG@10, effective rank, latency) with pinned versions and a reproduction script you keep.

3 · Scorecard

You get the full report. If the run clears the tier's thresholds, the certification is published on the reference leaderboard with a dated, linkable scorecard.

What certification means here

Every published number is measured on the same harness that gates our own deploys (Spec 0008): nDCG@10 against relevance judgments and effective rank of the serving embeddings. Methodology and pinned versions ship with every scorecard. A certification you cannot reproduce is not one we publish.

Need a guided sprint instead?Preview the harnessPublished scorecards