@wix/evalforge-types 0.78.0 → 0.80.0

package/build/index.mjs

@@ -148,7 +148,8 @@ var ModelConfigSchema = z4.object({
     z4.number().min(0).max(1).optional()
   ),
   maxTokens: z4.preprocess(nullToUndefined, z4.number().min(1).optional()),
-  maxTurns: z4.preprocess(nullToUndefined, z4.number().int().min(1).optional())
+  /** Number of agentic turns. 0 = unlimited (agent runs until done or timeout). */
+  maxTurns: z4.preprocess(nullToUndefined, z4.number().int().min(0).optional())
 });
 
 // src/common/rule.ts
@@ -220,11 +221,23 @@ var AgentSchema = TargetSchema.extend({
   agentType: AgentTypeSchema.default(AgentType.CLI),
   /** Command to run the agent (required for CLI agents, absent for SDK agents) */
   runCommand: AgentRunCommandSchema.optional(),
-  /** Optional model configuration override */
+  /**
+   * @deprecated Use `config` bag instead. Retained for backward compatibility
+   * with existing DB rows.
+   */
   modelConfig: ModelConfigSchema.optional(),
   systemPrompt: z6.string().nullish().describe(
     "Override for eval runs. undefined=default instructions, null=raw agent, string=append to claude_code preset. See https://docs.anthropic.com/en/docs/claude-code/sdk/modifying-system-prompts"
-  )
+  ),
+  /**
+   * Unified agent configuration bag. Absorbs model params (model,
+   * temperature, maxTokens, maxTurns) plus agent-specific settings
+   * (permissions, thinking tokens, allowed tools, etc.).
+   *
+   * Adapters read from `config` first, falling back to `modelConfig`
+   * for backward compatibility with existing DB rows.
+   */
+  config: z6.record(z6.string(), z6.unknown()).optional()
 });
 var CreateAgentInputSchema = AgentSchema.omit({
   id: true,
@@ -234,7 +247,8 @@ var CreateAgentInputSchema = AgentSchema.omit({
 });
 var UpdateAgentInputSchema = CreateAgentInputSchema.partial().extend({
   modelConfig: ModelConfigSchema.optional().nullable(),
-  systemPrompt: z6.string().optional().nullable()
+  systemPrompt: z6.string().optional().nullable(),
+  config: z6.record(z6.string(), z6.unknown()).optional().nullable()
 });
 
 // src/target/skill.ts
@@ -1632,14 +1646,18 @@ var EvalRunSchema = TenantEntitySchema.extend({
     agentType: AgentTypeSchema.optional(),
     runCommand: AgentRunCommandSchema.optional(),
     systemPrompt: z30.string().nullable().optional(),
-    modelConfig: ModelConfigSchema.optional()
+    /** @deprecated retained for backward compat with stored snapshots */
+    modelConfig: ModelConfigSchema.optional(),
+    config: z30.record(z30.string(), z30.unknown()).optional()
   }).optional(),
   /** UUID linking all runs in a comparison group */
   comparisonGroupId: z30.string().optional(),
   /** Human-readable label for this variant (e.g., "MCP: Wix Stores") */
   comparisonLabel: z30.string().optional(),
   /** LLM-generated analysis of the completed run */
-  runAnalysis: RunAnalysisSchema.optional()
+  runAnalysis: RunAnalysisSchema.optional(),
+  /** IDs of folders this run belongs to (read-only, managed via AddRunToFolder / RemoveRunFromFolder) */
+  folderIds: z30.array(z30.string()).optional()
 });
 var CreateEvalRunInputSchema = EvalRunSchema.omit({
   id: true,
@@ -1898,8 +1916,67 @@ var CreateTemplateInputSchema = TemplateSchema.omit({
 });
 var UpdateTemplateInputSchema = CreateTemplateInputSchema.partial();
 
-// src/schedule/eval-schedule.ts
+// src/agent/agent-config.ts
 import { z as z35 } from "zod";
+var BaseAgentConfigSchema = z35.object({
+  /** Model ID (Claude or OpenAI). */
+  model: AnyModelSchema.optional(),
+  /** Sampling temperature (0–1). */
+  temperature: z35.number().min(0).max(1).optional(),
+  /** Max output tokens per turn. */
+  maxTokens: z35.number().int().min(1).optional(),
+  /** Number of agentic turns. 0 = unlimited. */
+  maxTurns: z35.number().int().min(0).optional(),
+  /** Execution timeout in milliseconds. Overrides the default maxTurns-based calculation. */
+  maxDurationMs: z35.number().int().min(0).optional()
+});
+var EffortLevelSchema = z35.enum(["low", "medium", "high", "max"]);
+var ClaudeCodeConfigSchema = BaseAgentConfigSchema.extend({
+  /** Extended thinking token budget. */
+  maxThinkingTokens: z35.number().int().min(0).optional(),
+  /** Override the default allowedTools list passed to the SDK. */
+  allowedTools: z35.array(z35.string()).optional(),
+  /** Tools to remove from the model's context entirely. */
+  disallowedTools: z35.array(z35.string()).optional(),
+  /** Controls thinking depth: low, medium, high, max. */
+  effort: EffortLevelSchema.optional(),
+  /** Maximum USD spend per run. Stops execution when reached. */
+  maxBudgetUsd: z35.number().min(0).optional()
+});
+var PermissionValueSchema = z35.enum(["allow", "deny"]);
+var OpenCodePermissionSchema = z35.record(
+  z35.string(),
+  z35.union([PermissionValueSchema, z35.record(z35.string(), PermissionValueSchema)])
+);
+var ThinkingVariantSchema = z35.enum(["high", "low", "none"]);
+var OpenCodeConfigSchema = BaseAgentConfigSchema.extend({
+  /** Permission overrides (defaults: allow-all). */
+  permission: OpenCodePermissionSchema.optional(),
+  /** Maps to `--variant` CLI flag. 'none' omits --thinking entirely. Default: 'high'. */
+  thinkingVariant: ThinkingVariantSchema.optional(),
+  /** Nucleus sampling (0–1). Alternative to temperature. */
+  topP: z35.number().min(0).max(1).optional()
+}).omit({ maxTokens: true });
+var ReasoningEffortSchema = z35.enum(["low", "medium", "high"]);
+var SimpleAgentConfigSchema = BaseAgentConfigSchema.extend({
+  /** Anthropic thinking budget in tokens. Default: 10 000. */
+  thinkingBudgetTokens: z35.number().int().min(0).optional(),
+  /** Nucleus sampling (0–1). Alternative to temperature. */
+  topP: z35.number().min(0).max(1).optional(),
+  /** Integer seed for deterministic/reproducible results (if model supports it). */
+  seed: z35.number().int().optional(),
+  /** Stop sequences — model stops when generating any of these strings. */
+  stopSequences: z35.array(z35.string()).optional(),
+  /** OpenAI reasoning effort level. Default: 'high'. */
+  reasoningEffort: ReasoningEffortSchema.optional(),
+  /** Frequency penalty (−2 to 2). Reduces repetition of same tokens. */
+  frequencyPenalty: z35.number().min(-2).max(2).optional(),
+  /** Presence penalty (−2 to 2). Encourages topic diversity. */
+  presencePenalty: z35.number().min(-2).max(2).optional()
+});
+
+// src/schedule/eval-schedule.ts
+import { z as z36 } from "zod";
 var FrequencyType = /* @__PURE__ */ ((FrequencyType2) => {
   FrequencyType2["DAILY"] = "daily";
   FrequencyType2["WEEKDAY"] = "weekday";
@@ -1909,29 +1986,29 @@ var FrequencyType = /* @__PURE__ */ ((FrequencyType2) => {
 })(FrequencyType || {});
 var EvalScheduleSchema = TenantEntitySchema.extend({
   /** Whether the schedule is active */
-  enabled: z35.boolean(),
+  enabled: z36.boolean(),
   /** Test suite to run */
-  suiteId: z35.string(),
+  suiteId: z36.string(),
   /** Preset that provides agent + entities for this schedule */
-  presetId: z35.string(),
+  presetId: z36.string(),
   /** How often to run */
-  frequencyType: z35.nativeEnum(FrequencyType),
+  frequencyType: z36.nativeEnum(FrequencyType),
   /** Time of day in 24h format (HH:MM), hours 00-23, minutes 00-59 */
-  timeOfDay: z35.string().regex(/^([01]\d|2[0-3]):[0-5]\d$/),
+  timeOfDay: z36.string().regex(/^([01]\d|2[0-3]):[0-5]\d$/),
   /** Day of week (0=Sun, 6=Sat) for weekly schedules */
-  dayOfWeek: z35.number().min(0).max(6).optional(),
+  dayOfWeek: z36.number().min(0).max(6).optional(),
   /** Day of month (1-31) for monthly schedules */
-  dayOfMonth: z35.number().min(1).max(31).optional(),
+  dayOfMonth: z36.number().min(1).max(31).optional(),
   /** IANA timezone (e.g., 'America/New_York') */
-  timezone: z35.string(),
+  timezone: z36.string(),
   /** ID of the last eval run created by this schedule */
-  lastRunId: z35.string().optional(),
+  lastRunId: z36.string().optional(),
   /** Denormalized status of the last run */
-  lastRunStatus: z35.string().optional(),
+  lastRunStatus: z36.string().optional(),
   /** ISO timestamp of the last run */
-  lastRunAt: z35.string().optional(),
+  lastRunAt: z36.string().optional(),
   /** Next scheduled run time in UTC (pre-computed for efficient querying, set by backend) */
-  nextRunAt: z35.string().optional()
+  nextRunAt: z36.string().optional()
 });
 function isValidTimezone(tz) {
   try {
@@ -1944,14 +2021,14 @@ function isValidTimezone(tz) {
 function validateScheduleFields(data, ctx, options) {
   if (data.frequencyType === "weekly" /* WEEKLY */ && data.dayOfWeek == null) {
     ctx.addIssue({
-      code: z35.ZodIssueCode.custom,
+      code: z36.ZodIssueCode.custom,
       message: "dayOfWeek is required for weekly schedules",
       path: ["dayOfWeek"]
     });
   }
   if (data.frequencyType === "monthly" /* MONTHLY */ && data.dayOfMonth == null) {
     ctx.addIssue({
-      code: z35.ZodIssueCode.custom,
+      code: z36.ZodIssueCode.custom,
       message: "dayOfMonth is required for monthly schedules",
       path: ["dayOfMonth"]
     });
@@ -1959,7 +2036,7 @@ function validateScheduleFields(data, ctx, options) {
   const shouldValidateTz = options.partial ? data.timezone !== void 0 : true;
   if (shouldValidateTz && !isValidTimezone(data.timezone)) {
     ctx.addIssue({
-      code: z35.ZodIssueCode.custom,
+      code: z36.ZodIssueCode.custom,
       message: "Invalid IANA timezone",
       path: ["timezone"]
     });
@@ -2009,6 +2086,7 @@ export {
   AssertionSchema,
   AssertionTypeSchema,
   BATCH_IMPORT_LIMITS,
+  BaseAgentConfigSchema,
   BaseEntitySchema,
   BaseTestSchema,
   BatchAssertionLinkSchema,
@@ -2031,6 +2109,7 @@ export {
   CapabilityVersionOriginSchema,
   CapabilityVersionSchema,
   CapabilityWithLatestVersionSchema,
+  ClaudeCodeConfigSchema,
   ClaudeModel,
   ClaudeModelSchema,
   CommandExecutionSchema,
@@ -2063,6 +2142,7 @@ export {
   DiffContentSchema,
   DiffLineSchema,
   DiffLineTypeSchema,
+  EffortLevelSchema,
   EnvironmentSchema,
   EvalMetricsSchema,
   EvalRunFolderMembershipSchema,
@@ -2110,11 +2190,15 @@ export {
   OPENAI_RESPONSES_MODEL_IDS,
   OpenAIModel,
   OpenAIModelSchema,
+  OpenCodeConfigSchema,
+  OpenCodePermissionSchema,
+  PermissionValueSchema,
   PlaywrightNLTestSchema,
   PresetSchema,
   ProjectSchema,
   PromptResultSchema,
   RUN_COMMAND_LABELS,
+  ReasoningEffortSchema,
   RuleSchema,
   RuleTypeSchema,
   RunAnalysisFindingSchema,
@@ -2125,6 +2209,7 @@ export {
   SYSTEM_ASSERTION_IDS,
   ScenarioAssertionLinkSchema,
   ScenarioConversationSchema,
+  SimpleAgentConfigSchema,
   SiteConfigTestSchema,
   SkillFileSchema,
   SkillMetadataSchema,
@@ -2150,6 +2235,7 @@ export {
   TestTypeSchema,
   TextBlockSchema,
   ThinkingBlockSchema,
+  ThinkingVariantSchema,
   TimeAssertionSchema,
   TimeConfigSchema,
   TokenUsageSchema,