A runtime harness for AI-built business applications.
TeaQL connects a business model, generated domain APIs, and reusable runtimes. Its build-time tools help developers and AI agents evaluate models, discover APIs, and generate application foundations. Its runtime harness carries those contracts into application execution through typed requests, validation, trusted context, policy, transactions, and audit.
Start with an AI agent · Documentation · 中文
- Keep business meaning explicit. Use a shared domain model to describe entities, relationships, and validation, then generate APIs around that vocabulary.
- Make capabilities discoverable. Give agents model-aware evaluation and progressive Assist for the entities and operations they are working on.
- Carry rules into execution. Apply validation, query intent, mutation audit, and application-installed policies at the relevant TeaQL execution boundaries.
- Support change and verification. Regenerate domain libraries as models evolve, keep application code in its own workspace, and verify behavior against explicit language, provider, and version combinations.
flowchart LR
subgraph B["Build-time Harness"]
direction TB
A["teaql-agent-kit<br/>Development workflow and context governance"]
C["cargo-teaql CLI"]
F["teaql-forge-rs<br/>Open-source Rust generation service"]
S["TeaQL Model-Aware Services<br/>Commercial model services"]
A --> C
C --> F
C --> S
end
subgraph R["Runtime Harness"]
direction TB
W["Generated Workspace<br/>Business application"]
T["Generated Type Library<br/>Entity · Q · E · Checker"]
RT["TeaQL Runtime<br/>Seven language implementations"]
O["Peripheral tools · Future directions<br/>Observability platform · Policy management · Diagnostics"]
W --> T
T --> RT
RT -.-> O
end
B -->|"Generate domain libraries and application workspaces"| R
style O stroke-dasharray: 5 5
Build-time Harness runs in the development environment. Agent Kit supplies the workflow and agent integrations; the CLI connects it to model services. Developers can also use the CLI independently. The open-source Forge service focuses on Rust. Commercial TeaQL Agentic Model-Aware Services provide model evaluation, seven-language domain-library generation, target-specific workspaces, and progressive Assist. Their feature sets are documented separately.
Runtime Harness is the application's execution foundation. The Workspace hosts business code and assembles its dependencies. The generated Type Library provides model-specific entities, Q queries, E expressions, and Checker rules. The selected language Runtime executes supported operations through the application's configured context and providers. Type Library and Workspace are build-time outputs that become parts of the running application; application code stays editable, and domain-library changes come from the model and regeneration.
| Project | Role |
|---|---|
| teaql-agent-kit | Build TeaQL App Skill, Agent Context Kit, and host integrations. Enforcement depends on the adapter and entry point. |
| teaql-cargo-cli | The cargo-teaql client for model-service discovery, evaluation, generation, and Assist. |
| teaql-forge-rs | A local-first, open-source Rust generation service. |
| TeaQL Agentic Model-Aware Services | Commercial model-intelligence services, including generation and API guidance. |
The seven Runtime implementations share an architectural contract and retain their own language APIs, providers, releases, and verification boundaries:
| Language | Runtime repository |
|---|---|
| Java | teaql-java |
| Rust | teaql-rs |
| TypeScript | teaql-ts |
| Go | teaql-golang |
| Python | teaql-python |
| .NET | teaql-dotnet |
| Swift | teaql-swift |
The Agent Kit, CLI, open-source Forge, commercial model services, and seven Runtime implementations form the current project landscape. Generated workspaces depend on the selected target; Swift currently has domain-library generation without a separate application-workspace target.
Runtime telemetry, execution traces, structured logs, and audit already provide a foundation for inspecting application behavior. Independent observability products, policy-management tools, and operational diagnostics are future directions shown with a dashed border in the diagram. Audit and telemetry retain separate responsibilities. See Debugging, Audit, and Observability for the existing runtime facilities.
Published packages, generator-selected versions, and verified application scenarios have separate status. Use Latest Versions and the language guides to choose a starting point. Cross-language conformance checks, retained scenario evidence, and performance experiments support verification; each result applies to its recorded scope.
Start with TeaQL Agent Kit. Describe a small business application and choose a language and database. Follow the model → evaluate → generate → implement → verify workflow, then inspect the application's actual build, query, and audited mutation results.