DSL
Speaks your query language fluently.
Translate plain-language requests into your domain-specific language: an internal query syntax, a filter grammar, config files, rules for your engine. The model learns your grammar and your field names from real examples.
You type
Show me enterprise deals over 50k closing this quarter that went quiet for two weeks
crm-filter · your syntax
deal.segment:entamount>50kclose:this_qlast_touch>14d ✓ parses against your grammar · 0 unknown fields
What it does
Built for tasks like these.
- Natural language to your search filters
- Requests to internal query syntax
- Write rules for your rules engine
- Generate config from a description
How it works
From your examples to a model you own.
- 01
Gather request → query pairs
Logged searches, saved filters, or rules your team wrote by hand, each paired with what it was meant to do.
- 02
Train on your grammar
The model learns your syntax and field names until valid output is the default.
- 03
Wire it in
Validate its output with your own parser and run it. Nothing about your schema leaves the building.
Private by default
Your queries never leave the building.
Your schema, field names and internal syntax describe how your business works. Training and inference stay on your hardware, so none of it is shared with a vendor.
- Your customers' data leaves the building on every call
- The model changes whenever they ship
- You pay for a huge general model to do one narrow task
- The request never leaves the device
- The model is a file you own; it changes when you retrain it
- Small enough to run on hardware you already have
Trained, not prompted
Accurate on your task.
A frontier model has never seen your DSL and has to be taught it in every prompt. A trained model has it memorized and produces valid output in your exact syntax.
Inputenterprise deals over 50k closing this quarter that went quiet for two weeks
General model, prompted
SELECT * FROM deals WHERE segment = 'enterprise' AND amount > 50000 …
Micromodel, trained on your examples
deal.segment:ent amount>50k close:this_q last_touch>14d
Illustrative example.
And it costs less
No per-call bill.
No giant prompt full of grammar docs on every call. The grammar lives in the weights, and each request is just the request.
$120,000 annual savings
See it for yourself
Watch a micromodel run on your own GPU.
A guided demo, about a minute. It downloads once and runs entirely in your browser; nothing you type is sent anywhere.Open the demoRead and write
Related solutions.
Have a task like this?
We'll tell you if a micromodel can do it. Or book a demo, or email hello@micromodel.company.