Rust CLI for evaluating TeaQL semantic models, generating language-native libraries and workspaces, and discovering generated APIs through model-aware Assist.
# Discover the targets currently offered by the service.
cargo teaql services
# Evaluate before generation.
cargo teaql evaluate --input models/
# Generate a library or runnable workspace.
cargo teaql rust-lib-core --input models/ --output build/
cargo teaql java-app-console --input models/ --output build/
# Discover the generated API progressively.
cargo teaql rust-assist-query/school --input models/
cargo teaql rust-assist-query/school.established_date --input models/
# Local maintenance commands.
cargo teaql show-config
cargo teaql config
cargo teaql ping
cargo teaql check --tests
cargo teaql rust-expression-check --source src
cargo teaql java-expression-check --source src
cargo teaql kotlin-expression-check --source src
cargo teaql python-expression-check --source src
cargo teaql csharp-expression-check --source src
cargo teaql golang-expression-check --source src
cargo teaql swift-expression-check --source Sources
cargo teaql typescript-expression-check --source srcGeneration and Assist targets are provided dynamically by the TeaQL service,
so the fixed Commands list in --help is not their complete inventory. Run
cargo teaql services to discover the current targets and cargo teaql version
to inspect their versions.
Use --input for the KSML model file or directory. A dynamic target with model
input performs evaluation, generation, or Assist; a remote information target
such as services or version performs a GET request.
If no command is provided, the CLI defaults to cargo teaql services.
cargo-teaql is an equivalent direct invocation.
cargo teaql rust-expression-check parses application-owned Rust source and
fails when a TeaQL read query bypasses the generated E expression facade or
uses select_self_fields() outside a mutation flow. It also recognizes
centralized expression gateways such as LoadedAsset::from_selected, so a
service can evaluate a partial entity once and return an application read
model.
TeaQL-owned diagnostics use the TQL- prefix so they are distinct from Rust
compiler and Clippy diagnostics:
TQL-RUST-EXPR-001: a TeaQL query result is consumed without E-expression evaluation.TQL-RUST-EXPR-002: a read query usesselect_self_fields()instead of an explicit projection.
cargo teaql rust-expression-check \
--source rust-web-axum/src \
--exclude rust-web-axum/src/generatedThe default output is designed for humans and coding agents. Use
--format json for CI integrations. Projects can keep audited exceptions in
.teaql/rust-expression-check.yml:
version: 1
source_roots:
- rust-web-axum/src
exclude_roots:
- rust-lib-core
allow:
- path: rust-web-axum/src/services/asset_service.rs
function: find_by_path
rules: [TQL-RUST-EXPR-001, TQL-RUST-EXPR-002]
reason: full entity is loaded only for mutation hydrationExceptions are exact function-level records with mandatory reasons; they are
not wildcard suppressions. Keep the policy file under code review. Generated
directories such as rust-lib-core, generate-lib, and bizcore are excluded
by default.
cargo teaql java-expression-check parses Java with Tree-sitter and reports
direct getter chains across TeaQL entity relations. It discovers entity and
relation types from generated E.java and entity sources, so ordinary Java,
Android, and framework getter chains are not treated as TeaQL expressions.
TQL-JAVA-EXPR-001: a TeaQL entity relation is accessed through a direct getter chain instead ofE.
cargo teaql java-expression-check \
--source app/src/main/java \
--format jsonGenerated java-lib-core, target, build, and generated directories are
excluded from application diagnostics. The generated library is still read as
the domain catalog. When the application module is the working directory, a
sibling java-lib-core directory is discovered automatically.
Audited exceptions can be kept in .teaql/java-expression-check.yml:
version: 1
source_roots:
- app/src/main/java
allow:
- path: app/src/main/java/example/LegacyView.java
function: renderStatus
rules: [TQL-JAVA-EXPR-001]
reason: legacy view awaiting E-expression migrationThe same local policy workflow is available for six additional TeaQL ecosystems:
| Command | Direct-access rule | Capability rule |
|---|---|---|
kotlin-expression-check |
TQL-KOTLIN-EXPR-001 |
TQL-KOTLIN-EXPR-002 |
python-expression-check |
TQL-PYTHON-EXPR-001 |
TQL-PYTHON-EXPR-002 |
csharp-expression-check |
TQL-CSHARP-EXPR-001 |
TQL-CSHARP-EXPR-002 |
golang-expression-check |
TQL-GOLANG-EXPR-001 |
TQL-GOLANG-EXPR-002 |
swift-expression-check |
TQL-SWIFT-EXPR-001 |
TQL-SWIFT-EXPR-002 |
typescript-expression-check |
TQL-TYPESCRIPT-EXPR-001 |
TQL-TYPESCRIPT-EXPR-002 |
The 001 rules report direct TeaQL relation traversal. The 002 rules fail
closed when the model relation metadata or a functional generated E facade
cannot be found. This prevents a project from appearing compliant when its
runtime cannot express the required safe access.
The TypeScript check scans both .ts and .tsx. For example,
item.platform?.name is reported, while
E.workItem(item).platform().name().eval() is accepted.
These checks parse each language with its Tree-sitter grammar and derive relation names from workspace KSML/XML models. Generated code, dependency caches, and build output are excluded from application diagnostics. If a model is supplied only as a package, relation members can be declared explicitly:
version: 1
source_roots:
- src
relation_members:
- status
- platform
allow:
- path: src/legacy_view.py
function: render_status
rules: [TQL-PYTHON-EXPR-001]
reason: legacy view awaiting E-expression migrationStore the policy at .teaql/<language>-expression-check.yml, using kotlin,
python, csharp, golang, swift, or typescript as the language name.
All commands support --source, --exclude, --config, and
--format text|json.
cargo teaql java-app-console --input models/ \
--endpoint-prefix https://api.teaql.io/latest/ \
--api-key ******** \
--output ./build \
--timeout-seconds 300 \
--cwd /workspace/projectWhen a directory is provided as input, cargo-teaql will normally compress it into a zip archive. The server expects exactly one file in the zip to be named main.xml to serve as the entry point.
However, if a directory is provided but it does not contain a main.xml, the CLI will search for a single .xml or .ksml model file. If it finds exactly one such file, it will bypass compression and upload that single file directly. Single-file uploads do not require the name main.xml. If multiple files are found and no main.xml is present, the CLI will abort with an error.
If you create symlink aliases to the same binary, these names also work:
cargo teaql-java-app-console --input models/
cargo teaql-services
cargo teaql-version
cargo teaql-show-config
cargo teaql-configInstall the aliases automatically:
cargo-teaql install-linksYou can also target a custom directory with cargo-teaql install-links --dir /some/bin --force.
TeaQL provides specialized assist commands designed to help AI coding agents (and developers) quickly generate safe, compliant boilerplate templates for various business scenarios.
The entity-level command format is:
cargo teaql <lang>-assist-<action>/<entity> --input <model>
For field-specific help, append the KSML field name:
cargo teaql <lang>-assist-<action>/<entity>.<field> --input <model>
Currently supported rust-assist-* targets:
rust-assist-query: Generate a comprehensive query template with field selections and mandatory.purpose()/.comment()cascades.rust-assist-list-page: Generate a paginated query list template.rust-assist-create: Generate an entity creation template with proper.audit_as()constraints.rust-assist-update: Generate an entity update template.rust-assist-delete: Generate an entity deletion template safely.rust-assist-expression: Generate advanced business expression calculation templates usingE::.
(Note: The exact same variations are also available for java-assist-*)
Example Usage:
cargo teaql rust-assist-query/book --input modeling/bookstore.xml
cargo teaql rust-assist-query/book.published_at --input modeling/bookstore.xml
cargo teaql java-assist-create/customer --input modeling/bookstore.xmlCLI flag > Environment variable > config.yml > Built-in default
| Env var | Config key | Description |
|---|---|---|
TEAQL_ENDPOINT_PREFIX |
endpoint_prefix |
TeaQL service endpoint prefix |
TEAQL_LICENSE_FILE |
license_file |
License file path |
TEAQL_API_KEY |
api_key |
API Key for service access |
TEAQL_BUILD_DIR |
build_dir |
Output directory |
TEAQL_TIMEOUT_SECONDS |
timeout_seconds |
HTTP timeout in seconds |
The CLI uses a built-in default Out-Of-The-Box (OOTB) API key. This default key is fully functional and can be used freely for development, testing, and experimentation. Users and AI agents do not need to sign up or search for a real API key to get started. The service will accept this default key.
If you have a dedicated API key for higher rate limits or production usage, you can provide it via the --api-key CLI flag, the TEAQL_API_KEY environment variable, or in your config.yml.
TEAQL_SERVICE_URL is still accepted for compatibility, but new
configuration should use TEAQL_ENDPOINT_PREFIX.
Local config lives in ~/.teaql/config.yml.
endpoint_prefix: https://api.teaql.io/latest/
api_key: "YOUR_API_KEY" # optional — built-in free OOTB key used if omitted
build_dir: build
timeout_seconds: 300The endpoint prefix is combined with service methods. For example, generation
uses https://api.teaql.io/latest/generate, and cargo-teaql version uses
https://api.teaql.io/latest/version.
At startup, the CLI prints where each effective config value came from:
config (precedence: cli > env > config.yml > default):
endpoint_prefix = https://api.teaql.io/latest/ (from: environment variable)
api_key = ******** (from: built-in default)
build_dir = /workspace/build (from: built-in default)
timeout_seconds = 300 (from: ~/.teaql/config.yml)