<!-- ephemeral -->
# TeaQL Java Runtime Customization

```java
import io.teaql.core.DataServiceExecutor;
import io.teaql.core.MutationPolicyApprovalProvider;
import io.teaql.core.MutationPolicyRegistry;
import io.teaql.core.QueryPolicy;
import io.teaql.core.SchemaExecutor;
import io.teaql.core.meta.EntityMetaFactory;
import io.teaql.runtime.AppAuditEventSink;
import io.teaql.runtime.DefaultUserContext;
import io.teaql.runtime.RuntimeLogSink;
import io.teaql.runtime.TeaQLRuntime;
import java.util.List;
import java.util.Map;

public final class RuntimeCustomization {
    private RuntimeCustomization() {}

    public static TeaQLRuntime buildRuntime(
            EntityMetaFactory metadata,
            DataServiceExecutor provider,
            QueryPolicy queryPolicy,
            MutationPolicyRegistry mutationPolicies,
            MutationPolicyApprovalProvider mutationPolicyApprovals,
            RuntimeLogSink runtimeLogSink) {
        if (metadata == null || provider == null || queryPolicy == null
                || mutationPolicies == null || mutationPolicyApprovals == null) {
            throw new IllegalArgumentException("metadata, provider and policies are required");
        }
        return TeaQLRuntime.builder().metadata(metadata).dataService("default", provider)
                .queryPolicy(queryPolicy)
                .mutationPolicyRegistry(mutationPolicies)
                .mutationPolicyApprovalProvider(mutationPolicyApprovals)
                .logSink(runtimeLogSink).build();
    }

    public static DefaultUserContext requestContext(
            TeaQLRuntime runtime, String trustedTenant, AppAuditEventSink appAuditSink) {
        if (runtime == null || trustedTenant == null || trustedTenant.isBlank() || appAuditSink == null) {
            throw new IllegalArgumentException("trusted runtime, tenant and App Audit Sink are required");
        }
        DefaultUserContext context = new DefaultUserContext(runtime);
        context.putAttribute("trustedTenant", trustedTenant);
        context.putAttribute(AppAuditEventSink.class.getName(), appAuditSink);
        return context;
    }

    public static void readiness(DefaultUserContext context) {
        if (context.getAttribute("trustedTenant", String.class) == null) {
            throw new IllegalStateException("missing trusted tenant");
        }
        context.ensureSchema();
    }

    public static void rejectGovernanceOverride(Object input) {
        if (input instanceof Map<?, ?> values) {
            for (Map.Entry<?, ?> entry : values.entrySet()) {
                String key = String.valueOf(entry.getKey());
                if (List.of("tenant", "provider", "queryPolicy", "mutationPolicy",
                        "mutationPolicyApproval", "auditSink", "hardLimit",
                        "continuousPage").stream().anyMatch(value -> value.equalsIgnoreCase(key))) {
                    throw new IllegalArgumentException("forbidden governance override: " + key);
                }
                rejectGovernanceOverride(entry.getValue());
            }
        } else if (input instanceof Iterable<?> values) {
            values.forEach(RuntimeCustomization::rejectGovernanceOverride);
        }
    }
}
```

The generated workspace owns application-scoped provider construction. `UserContext`
initialization is the trusted boundary for tenant data and the customizable App Audit Sink;
query/save still receive only that context. Raw row audit remains separate. Readiness invokes
the real `SchemaExecutor` and propagates provider failure. Reject governance keys recursively
from JSON/TFP. The integration gate must also run generated query and audited mutation against
SQLite; a route-only `/health` smoke test is insufficient.

## Runtime telemetry

Observability is optional and application-owned. Build
`io.teaql.opentelemetry.OpenTelemetryRuntimeTelemetry` from the application's
OpenTelemetry tracer, meter and logger, then pass it to
`TeaQLRuntime.builder().telemetry(telemetry)`. Keep the no-op default when it is
not configured. The application owns bounded SDK processors, OTLP exporters,
`forceFlush` and shutdown; telemetry failure must never change a query, save,
audit or readiness result. Installing telemetry does not call `ensureSchema`.
TeaQL derives `teaql.error.category` from the native error type. Sampling never
controls or replaces App Audit Sink delivery. Do not generate a Collector,
additional exporters, auto-discovery, or a telemetry configuration DSL.

---

## TeaQL seven-language assist contract

Apply the verified Rust semantic ceiling while using only the exact JAVA generated and
runtime APIs. Discover APIs through the generated application AGENTS.md and progressive
model-aware Assist. Do not inspect generated domain-library source.

- Do not create plurals by appending `s` or `es`; use the centralized generated plural.
- Human and non-human entities use different generated predicate vocabularies. Preserve
  forms such as “who are active” and “whose email is”; never infer them from English.
- Configure filters, projection, paging, and other query options before `purpose(...)`.
  Comment may appear anywhere in the chain. Purpose enters the executable stage; execution
  requires both values, but comment does not have to immediately precede purpose.
- Every execute/list/stream and every save accepts exactly one context argument:
  `UserContext`. Name that argument `context`, never `runtime`; data services and global
  policy are injected when the context is built. Reserve `runtime` for process-level
  runtime ownership, provider/pool setup, and module assembly.
- Tenant, merchant, identity, permissions, request policy, purpose policy, hard limit,
  and continuous-page cursor policy come only from trusted context, never dynamic JSON or TFP.
- If the required operation is absent after current entity/action and required field
  Assist, stop that path and report MISSING_ASSIST. Do not guess an API or search the
  generated library as a fallback.
- Create each application-owned source file once. After its first compile attempt,
  repair only the smallest block identified by the exact compiler or test diagnostic.
  Preserve unrelated code; do not rewrite the complete file as an error-recovery loop.
- Before a repair that would replace more than 25% of an existing application file,
  stop and report LARGE_REWRITE_REQUEST with the file, exact diagnostic, reason, and
  estimated scope. Initial creation and model-driven regeneration are not repairs.

Capability: `runtime-custom`.

- Keep trusted dependencies and global runtime policy in UserContext initialization.
  Custom providers, policy hooks, and audit sinks must not add execute/save arguments.
- Preserve immutable row audit events and a separate customizable App Audit Sink.
  Include health, integration, and negative governance tests for every customization.


## Standard OpenTelemetry module

Use the optional standard module supplied by the selected TeaQL runtime. Match
its version to the runtime dependency selected by the model and workspace;
keep local source or internal Registry overrides during development. Do not
upgrade to an unrelated public release just to enable telemetry. A runtime
without this module reports `module_unavailable`; package installation alone
does not enable export.

On a supported backend, configure these trusted process variables before startup, never
through a query, JSON or TFP payload:

```bash
export TEAQL_TELEMETRY_ENABLED=true
export TEAQL_TELEMETRY_SERVICE_NAME=payment-service
export TEAQL_TELEMETRY_OTLP_ENDPOINT=http://localhost:4318
```

Without enablement, startup reports `disabled` and all three activation variables;
no SDK export resources are created. Enabled startup reports safe effective
settings, full recording and masking. Invalid settings produce a fixed safe
diagnostic, not a claim of successful export. Never print credentials or raw
configuration values.

The module automatically records supported Q, graph mutation and physical SQL
operations as traces, logs and metrics. SQL is expanded with masked parameters;
plaintext debugging never authorizes plaintext telemetry export. Full recording
does not promise lossless delivery when bounded queues or exporters fail.
Do not add per-method spans or replace the customer's global SDK providers.

Keep schema reconciliation explicit. Run the generated ensure-schema and seed
flow through the same configured runtime to observe startup settings and audited,
idempotent seeding. Telemetry setup must not run ensure-schema implicitly.
Flush and shut down the owned pipeline at application shutdown, not per query.
Exporter failures do not change business results; durable business audit remains
an independent path.


### Java assembly

Add the optional Maven artifact `io.teaql:teaql-opentelemetry` to the application,
using the exact same version as its TeaQL runtime dependency. Keep it out of the
generated domain library's mandatory dependencies. TeaQLRuntime discovers the
module through ServiceLoader at startup; no application SDK initialization is
required. Set the three variables before starting the JVM.