Data Chat Sandbox

The fastest way to trust a semantic model is to put real questions to it. Solid integrates directly with Data Chat, giving you a natural-language sandbox to test any model through conversational questions — no query language required.

What it does

Connect any of your semantic models to Data Chat and ask questions the way you'd ask a colleague. Every response comes back with:

• A real answer grounded in your semantic model, using the same SQL generation pipeline Solid uses in production
• A chart or visualization built in — not just a table of numbers, so you can evaluate answers at a glance
• Full traceability — every response is backed by generated SQL you can inspect

Use cases

Stress-test before rollout

Run a model through its paces in a conversational setting before making it available to agents or end users. Catch edge cases and gaps faster than you would with scripted benchmark questions alone.

Validate accuracy interactively

Instead of running a formal benchmark, ask a handful of representative business questions in plain language and see how the model responds in real time. This is especially useful for validating new tables or metrics just added to a model.

Share with stakeholders

Give a business stakeholder a natural-language interface to a specific semantic model so they can explore it themselves — without SQL access or knowledge of the underlying schema.

Why it complements benchmarking

Benchmark questions measure accuracy against known cases. The Data Chat sandbox helps you discover unknown cases — questions you didn't think to benchmark for, or edge cases that only surface in a real conversation. A model that holds up under open-ended questioning is a model you can trust in production.

How to access

Data Chat Sandbox is available from your Solid workspace. Select a semantic model and open the sandbox to start posing questions against it. Results appear alongside the generated SQL so you can trace every answer back to its source.