AI understands the domain. Texomy makes it executable.
Texomy Language expresses a semantic model and the rules that recognize it in text. AI designs and verifies that specification; the Texomy runtime executes it deterministically on every record.
- Microseconds, not secondsCompiled parser, native runtime
- Zero per-record LLM costAI helps author the spec; the runtime runs it
- Deterministic & testableSame input, same output — reviewable in PRs
Money:
$fields:
currency: Currency
amount: Number
$match:
- '#{currency}\s?#{amount}'
- '#{amount}\s?#{currency}'
Payment:
$fields:
money: Money
date: Date
$match: '#{money} due on #{date}'USD 50.5 due on May 31, 2025- moneyUSD 50.5Money
- currencyUSDCurrency::USD
- amount50.5Number::Decimal
- dateMay 31, 2025Date::TextualDate
- monthMayMonth::MAY
- day31Number::Integer
- year2025Number::Integer
Text often carries real structure — without a formal schema.
Webhooks, log streams, invoices, statements, clinical notes, and financial sentences all contain recurring concepts and relationships. The structure is real, but it is embedded in text rather than declared as a schema.
- Hand-written patterns. Fast and inexpensive at runtime, but hard to grow into a readable domain model.
- Per-request LLM extraction. Flexible for unfamiliar text, but costly, variable, and difficult to audit at scale.
Texomy combines the useful parts: AI helps develop the semantic model and specification; Texomy Language makes that knowledge explicit; the runtime compiles it into a deterministic parser. Every matching record becomes typed JSON according to an inspectable specification.
Let AI reason. Let Texomy run.
- Understands a domain and its concepts
- Builds the semantic model first
- Authors a Texomy specification
- Checks parse results against the intended model
- Refines the specification as text evolves
- Works with an expert's domain judgement
- Compiles the specification into a parser
- Produces the same result for the same input
- Runs at native speed without an LLM call
- Returns typed, inspectable semantic results
- Replays specification tests before release
- Fits wherever your system needs structured data
The specification is the contract.
A Texomy specification records the domain's concepts, their structure, and the rules that recognize their instances in text. It is the explicit contract between the domain expert, AI, and the runtime — not an implicit prompt or hidden model behaviour.
AI can propose changes to that contract. Texomy runs representative text through the specification, and AI or a domain expert can inspect the resulting semantic structure before accepting the change. At runtime, the parser executes the accepted specification literally; the model is not in the request path.
The specification can be versioned, reviewed, tested, and audited like source code. Its accumulated tests replay before release, keeping changes inspectable and catching regressions before they reach production.
Where Texomy creates leverage
Ingest normalization
Convert every incoming webhook, email, or upload into typed JSON at the boundary. Downstream never touches raw strings.
Structured-data pre-processor
Peel off the structured parts of a message deterministically before downstream processing. Hand only the genuinely free-form residue to an LLM when one is needed.
Gradual LLM extraction migration
Replace known recurring structures behind an existing LLM-based extractor. Keep the output contract, lower marginal cost, and migrate incrementally.
Rule & alert engines
Replace a thicket of Splunk/ELK regexes with one spec. Rules are written against a typed model, not raw log lines.
Compliance & audit
The spec is a file you can review, sign off, and hand to auditors. LLM behaviour is not.
Feature engineering for ML
Text → typed features → downstream model. Deterministic features are essential for reproducible training and scoring.
Try it in 60 seconds.
Studio is a browser playground with one-click imports of every example on this site. No install, no signup.