Language Support Overview
TeaQL generates typed domain, query, expression, relation, aggregation and mutation APIs for seven language runtimes. They share one model and one semantic contract, while retaining idiomatic casing, async models and package layouts.
| Language | Guide | Verified SQL databases | Typical fit |
|---|---|---|---|
| Java | Java Stack | PostgreSQL, MySQL, SQLite, Oracle, DB2, DM8, SAP HANA, SQL Server, DuckDB | Enterprise services and the broadest database matrix |
| Rust | Rust Stack | PostgreSQL, MySQL, SQLite | Efficient native services and explicit runtime composition |
| TypeScript | TypeScript Guide | PostgreSQL, MySQL, SQLite | Node services plus a browser-safe federal client profile |
| Swift | Swift Guide | SQLite | Apple-platform applications and Swift services |
| Python | Python Guide | PostgreSQL, MySQL, SQLite | Async services, automation and data-oriented applications |
| C# / .NET | .NET Guide | PostgreSQL, MySQL, SQLite, SQL Server | .NET services and native SQL Server integration |
| Go | Go Guide | PostgreSQL, MySQL, SQLite | Small services and straightforward deployment |
“Verified” means generated code was compiled and executed against a real database in the dated acceptance campaign. It is not a promise about every database version or a commercial support SLA. See the runtime/database matrix for evidence boundaries.
A complete path for each language
Every runtime has a language-native entry point and the same Begin/Prove/Extend journey. Java and Rust retain deeper stack guides behind those common entry points.
The seven-runtime contract is shared, but each path uses that ecosystem's project layout, toolchain, async/concurrency model, provider lifecycle, packaging expectations, and verification style. A user should not have to read five other language guides to ship the one runtime they chose.
The generator selections and independently published package versions are tracked separately in Latest Versions. Always prefer the dependency coordinates emitted by the generator for a newly generated workspace.
Shared native query profile
All seven native runtimes support the same baseline query surface: scalar and relation predicates including nested child queries, typed projection and ordering, explicit relation selection, grouped aggregates, relation statistics, facets, ordinary and stateful pagination, and exact per-parent Top-N. The cross-language query contract defines the full catalog and its source anchors. This native-runtime statement does not imply that the separate TFP wire profile or every database provider has the same evidence level.
For a task-oriented explanation of reads, writes, grouped and relation statistics, Facets, and result shapes, start with Query, Mutation, Analytics, and Facets, then open the selected language's Scenarios page for native syntax.
For diagnosis, start with Debugging, Audit, and Observability, then open the selected language's Debugging page. Those pages distinguish safe SQL evidence, operator-only rendered SQL, RuntimeTelemetry, application audit, and tenant-policy verification instead of treating them as one logging stream.
Runtime infrastructure is a separate portability dimension. The Cache, Lock, and Cloud Runtime Infrastructure guide records Local/Redis cache, Local/Remote Lock, Nacos, Consul, health, metrics, and graceful-shutdown coverage without treating source presence as live-provider verification.
Shared contract
Every runtime follows these rules:
- plural and predicate names are decided centrally; target templates do not guess English;
- human entities use
whose/whoAre, while non-human entities usewith/whichAre; - the public parameter name is
contextin every language; do not replace it with a runtime, service, data-service, or shortened alias; - list queries return typed entities in the runtime's
SmartList, never raw database rows or the host language's bare list/array type; - non-empty comment and purpose must both be supplied before execution; either order is valid and they need not be adjacent;
- mutations require an audit reason;
- tenant, actor, permissions and policy come from trusted context, not dynamic client JSON;
- SQL runtimes provide native COUNT, SUM and GROUP BY and exact partition/window per-parent Top-N;
- the runtime emits a row/mutation audit event and a separately attributable masked application event.
save(context)returns the persisted typed entity and preserves loaded/null/not-loaded state;- a Runtime Module is a passive, composable manifest; installing it never changes a database;
- schema reconciliation is an explicit startup or administrative call using that module;
- RuntimeTelemetry is optional and no-op by default; sampling and audit delivery remain independent.
The conformance model defines ten Assist categories for every language: create, query, update, delete, list/page, expression, debug, tool API, runtime customization, and ensure schema. Their publication evidence levels are tracked separately; see Assist and Generated Workspaces. Use the TeaQL Agent Kit as the evolving Harness Engineering workflow when an AI coding agent consumes these generated contracts.
Java remains the naming and Request-semantics gold standard. Other targets preserve the same meaning with language-appropriate casing rather than translating method names by intuition.
Choosing a runtime
Choose by application ecosystem and deployment needs, then verify the exact runtime/provider/driver combination you plan to ship. Do not select a database merely because a driver or template exists. Regenerate from the same model when comparing languages so public API differences remain reviewable.
The earlier six-language Order Search campaign executed 18 core language/database cells, 252 positive dynamic scenarios, governance gates, audited saves and real TypeScript-to-Rust federation. Read the acceptance article for the result and the provider compatibility reference for support terminology.