@axiom-lattice/core 3.0.2 → 3.0.4

package/dist/index.d.mts

@@ -5276,89 +5276,6 @@ declare function setBindingRegistry(r: BindingRegistry): void;
  */
 declare function getBindingRegistry(): BindingRegistry;
 
-/**
- * Service contract for controlling evaluation runs.
- *
- * The gateway layer implements this interface and registers it via
- * {@link setEvalRunService} so that agent tools (e.g., `run_eval`) can
- * start, abort, and monitor evaluation runs without knowing the
- * underlying run-time infrastructure.
- *
- * @see {@link setEvalRunService}
- * @see {@link getEvalRunService}
- */
-interface EvalRunService {
-    /**
-     * Start a new evaluation run for the given project.
-     *
-     * @param tenantId - Tenant that owns the project
-     * @param projectId - Project to evaluate
-     * @param suiteIds - Optional suite filter — only these suites run.
-     *   Omit to run all suites. Used to keep the validation suite
-     *   untouched during the fix loop (hold-out isolation).
-     * @returns The newly created run ID
-     */
-    startRun(tenantId: string, projectId: string, suiteIds?: string[]): Promise<string>;
-    /**
-     * Abort a running evaluation.
-     *
-     * @param runId - The run to abort
-     * @returns `true` if the run was successfully aborted
-     */
-    abortRun(runId: string): Promise<boolean>;
-    /**
-     * Check whether a run's runner process is still alive.
-     *
-     * @param runId - The run to check
-     * @returns `true` if the runner process is still active
-     */
-    isRunning(runId: string): boolean;
-}
-/**
- * Sets the global {@link EvalRunService} instance used by agent evaluation tools.
- *
- * The service enables the `run_eval` tool to start, abort, and monitor evaluation
- * runs. This must be called **at gateway startup** before any agent invocation that
- * expects `run_eval` to be functional.
- *
- * @example
- * ```ts
- * import { setEvalRunService } from "@axiom-lattice/core";
- *
- * const evalSvc: EvalRunService = {
- *   startRun: async (tid, pid) => { ... },
- *   abortRun: async (rid) => { ... },
- *   isRunning: (rid) => { ... },
- * };
- * setEvalRunService(evalSvc);
- * ```
- *
- * @param s - An {@link EvalRunService} implementation
- * @see {@link getEvalRunService}
- */
-declare function setEvalRunService(s: EvalRunService): void;
-/**
- * Returns the globally registered {@link EvalRunService}.
- *
- * Must be called **after** {@link setEvalRunService}. Used by the `run_eval`
- * tool to start, abort, and monitor evaluation runs.
- *
- * @returns The active {@link EvalRunService} instance
- * @throws {Error} If the service has not been initialized
- * @see {@link setEvalRunService}
- */
-declare function getEvalRunService(): EvalRunService;
-/**
- * Resets the global {@link EvalRunService} to `null`.
- *
- * Restricted to the `test` environment for test isolation only. Production
- * code must never call this function.
- *
- * @throws {Error} If `NODE_ENV` is not `"test"`
- * @see {@link setEvalRunService}
- */
-declare function clearEvalRunService(): void;
-
 interface LatticeAgentStepConfig {
     agent_id: string;
     override_message?: string;
@@ -5497,6 +5414,114 @@ interface CaseRunResult {
     logs: LatticeEvalLogEvent[];
 }
 
+/**
+ * Service contract for controlling evaluation runs.
+ *
+ * The gateway layer implements this interface and registers it via
+ * {@link setEvalRunService} so that agent tools (e.g., `run_eval`) can
+ * start, abort, and monitor evaluation runs without knowing the
+ * underlying run-time infrastructure.
+ *
+ * @see {@link setEvalRunService}
+ * @see {@link getEvalRunService}
+ */
+interface EvalRunService {
+    /**
+     * Start a new evaluation run for the given project.
+     *
+     * @param tenantId - Tenant that owns the project
+     * @param projectId - Project to evaluate
+     * @param suiteIds - Optional suite filter — only these suites run.
+     *   Omit to run all suites. Used to keep the validation suite
+     *   untouched during the fix loop (hold-out isolation).
+     * @param caseIds - Optional case filter — only these cases run (across
+     *   the suites selected by `suiteIds`, or all suites if omitted).
+     *   Suites with no matching cases are skipped; an error is thrown if
+     *   no case matches at all.
+     * @param runConfig - Runtime environment (workspace/project) the eval
+     *   executes in. Comes from the CALLER's runConfig, NOT from the eval
+     *   project's targetServerConfig. Test cases are environment-agnostic;
+     *   the run environment is decided at run time.
+     * @returns The newly created run ID
+     */
+    startRun(tenantId: string, projectId: string, suiteIds?: string[], caseIds?: string[], runConfig?: {
+        workspaceId?: string;
+        projectId?: string;
+    }): Promise<string>;
+    /**
+     * Abort a running evaluation.
+     *
+     * @param runId - The run to abort
+     * @returns `true` if the run was successfully aborted
+     */
+    abortRun(runId: string): Promise<boolean>;
+    /**
+     * Wait for a run to reach a terminal state and return its batch report.
+     *
+     * Resolves when the run finishes (completed, failed, or aborted) and its
+     * status has been persisted to the store — the caller can then read the
+     * run and its results without polling. Enables synchronous tool execution
+     * (`run_eval start` with `wait: true`).
+     *
+     * @param runId - The run to wait for
+     * @returns The final batch report of the run
+     * @throws {Error} If the run is not tracked in this process's memory
+     *   (already finished and cleaned up, or owned by another gateway instance)
+     */
+    waitForRun(runId: string): Promise<LatticeEvalBatchReport>;
+    /**
+     * Check whether a run's runner process is still alive.
+     *
+     * @param runId - The run to check
+     * @returns `true` if the runner process is still active
+     */
+    isRunning(runId: string): boolean;
+}
+/**
+ * Sets the global {@link EvalRunService} instance used by agent evaluation tools.
+ *
+ * The service enables the `run_eval` tool to start, abort, and monitor evaluation
+ * runs. This must be called **at gateway startup** before any agent invocation that
+ * expects `run_eval` to be functional.
+ *
+ * @example
+ * ```ts
+ * import { setEvalRunService } from "@axiom-lattice/core";
+ *
+ * const evalSvc: EvalRunService = {
+ *   startRun: async (tid, pid) => { ... },
+ *   abortRun: async (rid) => { ... },
+ *   isRunning: (rid) => { ... },
+ * };
+ * setEvalRunService(evalSvc);
+ * ```
+ *
+ * @param s - An {@link EvalRunService} implementation
+ * @see {@link getEvalRunService}
+ */
+declare function setEvalRunService(s: EvalRunService): void;
+/**
+ * Returns the globally registered {@link EvalRunService}.
+ *
+ * Must be called **after** {@link setEvalRunService}. Used by the `run_eval`
+ * tool to start, abort, and monitor evaluation runs.
+ *
+ * @returns The active {@link EvalRunService} instance
+ * @throws {Error} If the service has not been initialized
+ * @see {@link setEvalRunService}
+ */
+declare function getEvalRunService(): EvalRunService;
+/**
+ * Resets the global {@link EvalRunService} to `null`.
+ *
+ * Restricted to the `test` environment for test isolation only. Production
+ * code must never call this function.
+ *
+ * @throws {Error} If `NODE_ENV` is not `"test"`
+ * @see {@link setEvalRunService}
+ */
+declare function clearEvalRunService(): void;
+
 interface JudgeVerdict {
     pass?: boolean;
     final_score?: number;
@@ -5641,6 +5666,7 @@ declare class LatticeEvalProject {
     calibrateJudge(): Promise<{
         ok: boolean;
         reason?: string;
+        bypassed?: boolean;
     }>;
     /**
      * Run all suites as a batch and build an in-memory report.
@@ -7552,11 +7578,12 @@ declare function serializePluginMeta(plugin: Plugin): PluginMetaOutput;
 declare const BUILTIN_PLUGINS: Plugin[];
 
 /**
- * Document Learning Plugin
+ * Capability Learning Plugin
  *
- * Provides a document-learner agent and learn-document procedural skill.
- * The agent guides users through turning documents into structured skill
- * systems with evaluations — following a supervised learning paradigm.
+ * Provides a document-learner agent and learn-capability procedural skill.
+ * The agent guides users through turning source material (documents, API
+ * specs, conversations, spreadsheets) into structured skill systems with
+ * evaluations — following a supervised learning paradigm.
  */
 
 declare const documentLearningPlugin: Plugin;

package/dist/index.d.ts

@@ -5276,89 +5276,6 @@ declare function setBindingRegistry(r: BindingRegistry): void;
  */
 declare function getBindingRegistry(): BindingRegistry;
 
-/**
- * Service contract for controlling evaluation runs.
- *
- * The gateway layer implements this interface and registers it via
- * {@link setEvalRunService} so that agent tools (e.g., `run_eval`) can
- * start, abort, and monitor evaluation runs without knowing the
- * underlying run-time infrastructure.
- *
- * @see {@link setEvalRunService}
- * @see {@link getEvalRunService}
- */
-interface EvalRunService {
-    /**
-     * Start a new evaluation run for the given project.
-     *
-     * @param tenantId - Tenant that owns the project
-     * @param projectId - Project to evaluate
-     * @param suiteIds - Optional suite filter — only these suites run.
-     *   Omit to run all suites. Used to keep the validation suite
-     *   untouched during the fix loop (hold-out isolation).
-     * @returns The newly created run ID
-     */
-    startRun(tenantId: string, projectId: string, suiteIds?: string[]): Promise<string>;
-    /**
-     * Abort a running evaluation.
-     *
-     * @param runId - The run to abort
-     * @returns `true` if the run was successfully aborted
-     */
-    abortRun(runId: string): Promise<boolean>;
-    /**
-     * Check whether a run's runner process is still alive.
-     *
-     * @param runId - The run to check
-     * @returns `true` if the runner process is still active
-     */
-    isRunning(runId: string): boolean;
-}
-/**
- * Sets the global {@link EvalRunService} instance used by agent evaluation tools.
- *
- * The service enables the `run_eval` tool to start, abort, and monitor evaluation
- * runs. This must be called **at gateway startup** before any agent invocation that
- * expects `run_eval` to be functional.
- *
- * @example
- * ```ts
- * import { setEvalRunService } from "@axiom-lattice/core";
- *
- * const evalSvc: EvalRunService = {
- *   startRun: async (tid, pid) => { ... },
- *   abortRun: async (rid) => { ... },
- *   isRunning: (rid) => { ... },
- * };
- * setEvalRunService(evalSvc);
- * ```
- *
- * @param s - An {@link EvalRunService} implementation
- * @see {@link getEvalRunService}
- */
-declare function setEvalRunService(s: EvalRunService): void;
-/**
- * Returns the globally registered {@link EvalRunService}.
- *
- * Must be called **after** {@link setEvalRunService}. Used by the `run_eval`
- * tool to start, abort, and monitor evaluation runs.
- *
- * @returns The active {@link EvalRunService} instance
- * @throws {Error} If the service has not been initialized
- * @see {@link setEvalRunService}
- */
-declare function getEvalRunService(): EvalRunService;
-/**
- * Resets the global {@link EvalRunService} to `null`.
- *
- * Restricted to the `test` environment for test isolation only. Production
- * code must never call this function.
- *
- * @throws {Error} If `NODE_ENV` is not `"test"`
- * @see {@link setEvalRunService}
- */
-declare function clearEvalRunService(): void;
-
 interface LatticeAgentStepConfig {
     agent_id: string;
     override_message?: string;
@@ -5497,6 +5414,114 @@ interface CaseRunResult {
     logs: LatticeEvalLogEvent[];
 }
 
+/**
+ * Service contract for controlling evaluation runs.
+ *
+ * The gateway layer implements this interface and registers it via
+ * {@link setEvalRunService} so that agent tools (e.g., `run_eval`) can
+ * start, abort, and monitor evaluation runs without knowing the
+ * underlying run-time infrastructure.
+ *
+ * @see {@link setEvalRunService}
+ * @see {@link getEvalRunService}
+ */
+interface EvalRunService {
+    /**
+     * Start a new evaluation run for the given project.
+     *
+     * @param tenantId - Tenant that owns the project
+     * @param projectId - Project to evaluate
+     * @param suiteIds - Optional suite filter — only these suites run.
+     *   Omit to run all suites. Used to keep the validation suite
+     *   untouched during the fix loop (hold-out isolation).
+     * @param caseIds - Optional case filter — only these cases run (across
+     *   the suites selected by `suiteIds`, or all suites if omitted).
+     *   Suites with no matching cases are skipped; an error is thrown if
+     *   no case matches at all.
+     * @param runConfig - Runtime environment (workspace/project) the eval
+     *   executes in. Comes from the CALLER's runConfig, NOT from the eval
+     *   project's targetServerConfig. Test cases are environment-agnostic;
+     *   the run environment is decided at run time.
+     * @returns The newly created run ID
+     */
+    startRun(tenantId: string, projectId: string, suiteIds?: string[], caseIds?: string[], runConfig?: {
+        workspaceId?: string;
+        projectId?: string;
+    }): Promise<string>;
+    /**
+     * Abort a running evaluation.
+     *
+     * @param runId - The run to abort
+     * @returns `true` if the run was successfully aborted
+     */
+    abortRun(runId: string): Promise<boolean>;
+    /**
+     * Wait for a run to reach a terminal state and return its batch report.
+     *
+     * Resolves when the run finishes (completed, failed, or aborted) and its
+     * status has been persisted to the store — the caller can then read the
+     * run and its results without polling. Enables synchronous tool execution
+     * (`run_eval start` with `wait: true`).
+     *
+     * @param runId - The run to wait for
+     * @returns The final batch report of the run
+     * @throws {Error} If the run is not tracked in this process's memory
+     *   (already finished and cleaned up, or owned by another gateway instance)
+     */
+    waitForRun(runId: string): Promise<LatticeEvalBatchReport>;
+    /**
+     * Check whether a run's runner process is still alive.
+     *
+     * @param runId - The run to check
+     * @returns `true` if the runner process is still active
+     */
+    isRunning(runId: string): boolean;
+}
+/**
+ * Sets the global {@link EvalRunService} instance used by agent evaluation tools.
+ *
+ * The service enables the `run_eval` tool to start, abort, and monitor evaluation
+ * runs. This must be called **at gateway startup** before any agent invocation that
+ * expects `run_eval` to be functional.
+ *
+ * @example
+ * ```ts
+ * import { setEvalRunService } from "@axiom-lattice/core";
+ *
+ * const evalSvc: EvalRunService = {
+ *   startRun: async (tid, pid) => { ... },
+ *   abortRun: async (rid) => { ... },
+ *   isRunning: (rid) => { ... },
+ * };
+ * setEvalRunService(evalSvc);
+ * ```
+ *
+ * @param s - An {@link EvalRunService} implementation
+ * @see {@link getEvalRunService}
+ */
+declare function setEvalRunService(s: EvalRunService): void;
+/**
+ * Returns the globally registered {@link EvalRunService}.
+ *
+ * Must be called **after** {@link setEvalRunService}. Used by the `run_eval`
+ * tool to start, abort, and monitor evaluation runs.
+ *
+ * @returns The active {@link EvalRunService} instance
+ * @throws {Error} If the service has not been initialized
+ * @see {@link setEvalRunService}
+ */
+declare function getEvalRunService(): EvalRunService;
+/**
+ * Resets the global {@link EvalRunService} to `null`.
+ *
+ * Restricted to the `test` environment for test isolation only. Production
+ * code must never call this function.
+ *
+ * @throws {Error} If `NODE_ENV` is not `"test"`
+ * @see {@link setEvalRunService}
+ */
+declare function clearEvalRunService(): void;
+
 interface JudgeVerdict {
     pass?: boolean;
     final_score?: number;
@@ -5641,6 +5666,7 @@ declare class LatticeEvalProject {
     calibrateJudge(): Promise<{
         ok: boolean;
         reason?: string;
+        bypassed?: boolean;
     }>;
     /**
      * Run all suites as a batch and build an in-memory report.
@@ -7552,11 +7578,12 @@ declare function serializePluginMeta(plugin: Plugin): PluginMetaOutput;
 declare const BUILTIN_PLUGINS: Plugin[];
 
 /**
- * Document Learning Plugin
+ * Capability Learning Plugin
  *
- * Provides a document-learner agent and learn-document procedural skill.
- * The agent guides users through turning documents into structured skill
- * systems with evaluations — following a supervised learning paradigm.
+ * Provides a document-learner agent and learn-capability procedural skill.
+ * The agent guides users through turning source material (documents, API
+ * specs, conversations, spreadsheets) into structured skill systems with
+ * evaluations — following a supervised learning paradigm.
  */
 
 declare const documentLearningPlugin: Plugin;