@wix/evalforge-evaluator 0.87.0 → 0.89.0

package/build/index.mjs

@@ -213,7 +213,21 @@ function applyParamsToAssertion(assertion, params) {
         );
       }
     }
-    return { ...assertion, prompt, systemPrompt };
+    return {
+      ...assertion,
+      prompt,
+      systemPrompt,
+      ...params.model !== void 0 && { model: params.model },
+      ...params.maxTokens !== void 0 && {
+        maxTokens: params.maxTokens
+      },
+      ...params.temperature !== void 0 && {
+        temperature: params.temperature
+      },
+      ...params.minScore !== void 0 && {
+        minScore: params.minScore
+      }
+    };
   }
   if (assertion.type === "skill_was_called" && params.skillNames !== void 0) {
     return {
@@ -245,7 +259,10 @@ function resolveSystemAssertion(assertionId, params) {
         type: "llm_judge",
         prompt: params?.prompt ?? "",
         systemPrompt: params?.systemPrompt,
-        minScore: params?.minScore
+        minScore: params?.minScore,
+        model: params?.model,
+        maxTokens: params?.maxTokens,
+        temperature: params?.temperature
       };
       break;
     default:
@@ -381,7 +398,7 @@ async function fetchEvaluationData(api, projectId2, evalRunId2) {
 // src/run-scenario/index.ts
 import {
   AssertionResultStatus,
-  AVAILABLE_MODEL_IDS
+  DEFAULT_JUDGE_MODEL
 } from "@wix/evalforge-types";
 import {
   evaluateAssertions as evaluateAssertionsBase
@@ -514,7 +531,7 @@ import { randomUUID as randomUUID2 } from "crypto";
 // src/run-scenario/agents/registry.ts
 var AgentAdapterRegistry = class {
   /**
-   * Map of command strings to their registered adapters.
+   * Map of run commands to their registered adapters.
    * Multiple commands can map to the same adapter.
    */
   adapters = /* @__PURE__ */ new Map();
@@ -543,9 +560,9 @@ var AgentAdapterRegistry = class {
     }
   }
   /**
-   * Get an adapter by command string.
+   * Get an adapter by run command.
    *
-   * @param runCommand - The command string to look up (e.g., 'claude', 'cursor')
+   * @param runCommand - The run command to look up
    * @returns The registered adapter, or undefined if not found
    */
   get(runCommand) {
@@ -554,7 +571,7 @@ var AgentAdapterRegistry = class {
   /**
    * Check if a command has a registered adapter.
    *
-   * @param runCommand - The command string to check
+   * @param runCommand - The run command to check
    * @returns True if an adapter is registered for this command
    */
   has(runCommand) {
@@ -571,7 +588,7 @@ var AgentAdapterRegistry = class {
   /**
    * Get all supported commands.
    *
-   * @returns Array of all registered command strings
+   * @returns Array of all registered run commands
    */
   getSupportedCommands() {
     return Array.from(this.adapters.keys());
@@ -621,6 +638,9 @@ function getAdapter(runCommand) {
   return adapter;
 }
 
+// src/run-scenario/agents/claude-code/claude-code-adapter.ts
+import { AgentRunCommand } from "@wix/evalforge-types";
+
 // src/run-scenario/agents/claude-code/execute.ts
 import {
   ClaudeModel,
@@ -1642,7 +1662,7 @@ function buildLLMTraceFromSteps(steps, totalDurationMs, usage, model) {
 var ClaudeCodeAdapter = class {
   id = "claude-code";
   name = "Claude Code";
-  supportedCommands = ["claude"];
+  supportedCommands = [AgentRunCommand.CLAUDE];
   /**
    * Execute a skill using the Claude Code SDK.
    *
@@ -2423,7 +2443,8 @@ function extractTemplateFiles(before, after) {
 }
 
 // src/run-scenario/run-agent-with-context.ts
-var DEFAULT_AGENT_COMMAND = "claude";
+import { AgentRunCommand as AgentRunCommand2 } from "@wix/evalforge-types";
+var DEFAULT_AGENT_COMMAND = AgentRunCommand2.CLAUDE;
 async function runAgentWithContext(config, evalRunId2, scenario, evalData, workDir) {
   const skillsGroupId = evalData.evalRun.skillsGroupId;
   if (!skillsGroupId) {
@@ -2510,7 +2531,7 @@ async function runScenario(config, evalRunId2, scenario, evalData, template, res
     }))
   };
   const { "x-wix-ai-gateway-stream": _stream, ...judgeHeaders } = config.aiGatewayHeaders;
-  const defaultJudgeModel = AVAILABLE_MODEL_IDS[0];
+  const defaultJudgeModel = DEFAULT_JUDGE_MODEL;
   const assertionContext = {
     workDir,
     defaultJudgeModel,