@langwatch/scenario 0.2.1 → 0.2.6

package/README.md

@@ -222,10 +222,6 @@ export default defineConfig({
     model: openai("gpt-4o-mini"),
     temperature: 0.1,
   },
-
-  // Configure the LangWatch reporting endpoint and API key
-  langwatchEndpoint: "https://app.langwatch.ai",
-  langwatchApiKey: process.env.LANGWATCH_API_KEY,
 });
 ```
 
@@ -239,8 +235,6 @@ The following configuration options are all optional. You can specify any combin
   - `model`: **(Required if `defaultModel` is set)** An instance of a language model from a provider like `@ai-sdk/openai`.
   - `temperature` _(Optional)_: The default temperature for the model (e.g., `0.1`).
   - `maxTokens` _(Optional)_: The default maximum number of tokens for the model to generate.
-- `langwatchEndpoint` _(Optional)_: The endpoint for the LangWatch reporting service. If not specified, it defaults to the `LANGWATCH_ENDPOINT` environment variable, or `https://app.langwatch.ai`.
-- `langwatchApiKey` _(Optional)_: Your LangWatch API key for authenticating with the reporting service. If not specified, it defaults to the `LANGWATCH_API_KEY` environment variable.
 
 ### Environment Variables
 

package/dist/{chunk-ZMHTHRDR.mjs → chunk-MOOKAYIE.mjs}

@@ -4,14 +4,13 @@ import {
 
 // src/domain/core/config.ts
 import { z } from "zod";
+var DEFAULT_TEMPERATURE = 0;
 var scenarioProjectConfigSchema = z.object({
   defaultModel: z.object({
     model: z.custom(),
-    temperature: z.number().min(0).max(1).optional().default(0),
+    temperature: z.number().min(0).max(1).optional().default(DEFAULT_TEMPERATURE),
     maxTokens: z.number().optional()
-  }).optional(),
-  langwatchEndpoint: z.string().optional(),
-  langwatchApiKey: z.string().optional()
+  }).optional()
 }).strict();
 function defineConfig(config2) {
   return config2;
@@ -44,11 +43,18 @@ var JudgeAgentAdapter = class {
   }
 };
 
+// src/domain/scenarios/index.ts
+var DEFAULT_MAX_TURNS = 10;
+var DEFAULT_VERBOSE = false;
+
 // src/domain/index.ts
 var domain_exports = {};
 __export(domain_exports, {
   AgentAdapter: () => AgentAdapter,
   AgentRole: () => AgentRole,
+  DEFAULT_MAX_TURNS: () => DEFAULT_MAX_TURNS,
+  DEFAULT_TEMPERATURE: () => DEFAULT_TEMPERATURE,
+  DEFAULT_VERBOSE: () => DEFAULT_VERBOSE,
   JudgeAgentAdapter: () => JudgeAgentAdapter,
   UserSimulatorAgentAdapter: () => UserSimulatorAgentAdapter,
   allAgentRoles: () => allAgentRoles,
@@ -231,7 +237,7 @@ async function loadProjectConfig() {
   configLoadPromise = (async () => {
     try {
       config = await loadScenarioProjectConfig();
-      logger.info("loaded scenario project config", { config });
+      logger.debug("loaded scenario project config", { config });
     } catch (error) {
       logger.error("error loading scenario project config", { error });
     } finally {
@@ -425,17 +431,20 @@ var EventReporter = class {
   eventsEndpoint;
   eventAlertMessageLogger;
   logger = new Logger("scenario.events.EventReporter");
+  isEnabled;
   constructor(config2) {
     this.apiKey = config2.apiKey ?? "";
     this.eventsEndpoint = new URL("/api/scenario-events", config2.endpoint);
     this.eventAlertMessageLogger = new EventAlertMessageLogger();
     this.eventAlertMessageLogger.handleGreeting();
+    this.isEnabled = this.apiKey.length > 0 && this.eventsEndpoint.href.length > 0;
   }
   /**
    * Posts an event to the configured endpoint.
    * Logs success/failure but doesn't throw - event posting shouldn't break scenario execution.
    */
   async postEvent(event) {
+    if (!this.isEnabled) return {};
     const result = {};
     this.logger.debug(`[${event.type}] Posting event`, { event });
     const processedEvent = this.processEventForApi(event);
@@ -598,6 +607,7 @@ var EventBus = class _EventBus {
 };
 
 export {
+  DEFAULT_TEMPERATURE,
   scenarioProjectConfigSchema,
   defineConfig,
   AgentRole,
@@ -605,9 +615,11 @@ export {
   AgentAdapter,
   UserSimulatorAgentAdapter,
   JudgeAgentAdapter,
+  DEFAULT_MAX_TURNS,
+  DEFAULT_VERBOSE,
   domain_exports,
-  loadScenarioProjectConfig,
   Logger,
+  env,
   getProjectConfig,
   generateThreadId,
   generateScenarioRunId,

package/dist/index.d.mts

@@ -1,7 +1,6 @@
-import * as zod from 'zod';
-import { z } from 'zod';
 import * as ai from 'ai';
 import { CoreMessage, CoreToolMessage, LanguageModel } from 'ai';
+import { z } from 'zod';
 import { Observable } from 'rxjs';
 
 declare enum AgentRole {
@@ -121,6 +120,8 @@ declare abstract class JudgeAgentAdapter implements AgentAdapter {
     abstract call(input: AgentInput): Promise<AgentReturnTypes>;
 }
 
+declare const DEFAULT_MAX_TURNS = 10;
+declare const DEFAULT_VERBOSE = false;
 /**
  * Configuration for a scenario.
  */
@@ -147,11 +148,19 @@ interface ScenarioConfig {
      */
     script?: ScriptStep[];
     /**
-     * Whether to output verbose logging. Defaults to false.
+     * Whether to output verbose logging.
+     *
+     * If no value is provided, this defaults to {@link DEFAULT_VERBOSE}.
+     *
+     * @default {@link DEFAULT_VERBOSE}
      */
     verbose?: boolean;
     /**
-     * The maximum number of turns to execute. Defaults to 20.
+     * The maximum number of turns to execute.
+     *
+     * If no value is provided, this defaults to {@link DEFAULT_MAX_TURNS}.
+     *
+     * @default {@link DEFAULT_MAX_TURNS}
      */
     maxTurns?: number;
     /**
@@ -337,6 +346,8 @@ interface ScenarioExecutionStateLike {
     hasToolCall(toolName: string): boolean;
 }
 
+/** Default temperature for language model inference */
+declare const DEFAULT_TEMPERATURE = 0;
 declare const scenarioProjectConfigSchema: z.ZodObject<{
     defaultModel: z.ZodOptional<z.ZodObject<{
         model: z.ZodType<ai.LanguageModelV1, z.ZodTypeDef, ai.LanguageModelV1>;
@@ -351,28 +362,49 @@ declare const scenarioProjectConfigSchema: z.ZodObject<{
         temperature?: number | undefined;
         maxTokens?: number | undefined;
     }>>;
-    langwatchEndpoint: z.ZodOptional<z.ZodString>;
-    langwatchApiKey: z.ZodOptional<z.ZodString>;
 }, "strict", z.ZodTypeAny, {
     defaultModel?: {
         model: ai.LanguageModelV1;
         temperature: number;
         maxTokens?: number | undefined;
     } | undefined;
-    langwatchEndpoint?: string | undefined;
-    langwatchApiKey?: string | undefined;
 }, {
     defaultModel?: {
         model: ai.LanguageModelV1;
         temperature?: number | undefined;
         maxTokens?: number | undefined;
     } | undefined;
-    langwatchEndpoint?: string | undefined;
-    langwatchApiKey?: string | undefined;
 }>;
 type ScenarioProjectConfig = z.infer<typeof scenarioProjectConfigSchema>;
 declare function defineConfig(config: ScenarioProjectConfig): ScenarioProjectConfig;
 
+type domain_AgentAdapter = AgentAdapter;
+declare const domain_AgentAdapter: typeof AgentAdapter;
+type domain_AgentInput = AgentInput;
+type domain_AgentReturnTypes = AgentReturnTypes;
+type domain_AgentRole = AgentRole;
+declare const domain_AgentRole: typeof AgentRole;
+declare const domain_DEFAULT_MAX_TURNS: typeof DEFAULT_MAX_TURNS;
+declare const domain_DEFAULT_TEMPERATURE: typeof DEFAULT_TEMPERATURE;
+declare const domain_DEFAULT_VERBOSE: typeof DEFAULT_VERBOSE;
+type domain_JudgeAgentAdapter = JudgeAgentAdapter;
+declare const domain_JudgeAgentAdapter: typeof JudgeAgentAdapter;
+type domain_ScenarioConfig = ScenarioConfig;
+type domain_ScenarioConfigFinal = ScenarioConfigFinal;
+type domain_ScenarioExecutionLike = ScenarioExecutionLike;
+type domain_ScenarioExecutionStateLike = ScenarioExecutionStateLike;
+type domain_ScenarioProjectConfig = ScenarioProjectConfig;
+type domain_ScenarioResult = ScenarioResult;
+type domain_ScriptStep = ScriptStep;
+type domain_UserSimulatorAgentAdapter = UserSimulatorAgentAdapter;
+declare const domain_UserSimulatorAgentAdapter: typeof UserSimulatorAgentAdapter;
+declare const domain_allAgentRoles: typeof allAgentRoles;
+declare const domain_defineConfig: typeof defineConfig;
+declare const domain_scenarioProjectConfigSchema: typeof scenarioProjectConfigSchema;
+declare namespace domain {
+  export { domain_AgentAdapter as AgentAdapter, type domain_AgentInput as AgentInput, type domain_AgentReturnTypes as AgentReturnTypes, domain_AgentRole as AgentRole, domain_DEFAULT_MAX_TURNS as DEFAULT_MAX_TURNS, domain_DEFAULT_TEMPERATURE as DEFAULT_TEMPERATURE, domain_DEFAULT_VERBOSE as DEFAULT_VERBOSE, domain_JudgeAgentAdapter as JudgeAgentAdapter, type domain_ScenarioConfig as ScenarioConfig, type domain_ScenarioConfigFinal as ScenarioConfigFinal, type domain_ScenarioExecutionLike as ScenarioExecutionLike, type domain_ScenarioExecutionStateLike as ScenarioExecutionStateLike, type domain_ScenarioProjectConfig as ScenarioProjectConfig, type domain_ScenarioResult as ScenarioResult, type domain_ScriptStep as ScriptStep, domain_UserSimulatorAgentAdapter as UserSimulatorAgentAdapter, domain_allAgentRoles as allAgentRoles, domain_defineConfig as defineConfig, domain_scenarioProjectConfigSchema as scenarioProjectConfigSchema };
+}
+
 /**
  * Configuration for the inference parameters of a testing agent.
  */
@@ -400,6 +432,14 @@ interface TestingAgentConfig extends TestingAgentInferenceConfig {
      * The name of the agent.
      */
     name?: string;
+    /**
+     * System prompt to use for the agent.
+     *
+     * Useful in more complex scenarios where you want to set the system prompt
+     * for the agent directly. If left blank, this will be automatically generated
+     * from the scenario description.
+     */
+    systemPrompt?: string;
 }
 /**
  * The arguments for finishing a test, used by the judge agent's tool.
@@ -502,8 +542,17 @@ declare const judgeAgent: (cfg: JudgeAgentConfig) => {
  *
  * @param config Optional configuration for the agent.
  * @param config.model The language model to use for generating responses.
- * @param config.temperature The temperature to use for the model.
+ *                     If not provided, a default model will be used.
+ * @param config.temperature The temperature for the language model (0.0-1.0).
+ *                          Lower values make responses more deterministic.
+ *                          Defaults to {@link DEFAULT_TEMPERATURE}.
  * @param config.maxTokens The maximum number of tokens to generate.
+ *                        If not provided, uses model defaults.
+ * @param config.name The name of the agent.
+ * @param config.systemPrompt Custom system prompt to override default user simulation behavior.
+ *                           Use this to create specialized user personas or behaviors.
+ *
+ * @throws {Error} If no model is configured either in parameters or global config.
  *
  * @example
  * ```typescript
@@ -517,7 +566,8 @@ declare const judgeAgent: (cfg: JudgeAgentConfig) => {
  * };
  *
  * async function main() {
- *   const result = await run({
+ *   // Basic user simulator with default behavior
+ *   const basicResult = await run({
  *     name: "User Simulator Test",
  *     description: "A simple test to see if the user simulator works.",
  *     agents: [myAgent, userSimulatorAgent()],
@@ -526,9 +576,50 @@ declare const judgeAgent: (cfg: JudgeAgentConfig) => {
  *       agent(),
  *     ],
  *   });
+ *
+ *   // Customized user simulator
+ *   const customResult = await run({
+ *     name: "Expert User Test",
+ *     description: "User seeks help with TypeScript programming",
+ *     agents: [
+ *       myAgent,
+ *       userSimulatorAgent({
+ *         model: openai("gpt-4"),
+ *         temperature: 0.3,
+ *         systemPrompt: "You are a technical user who asks detailed questions"
+ *       })
+ *     ],
+ *     script: [
+ *       user(),
+ *       agent(),
+ *     ],
+ *   });
+ *
+ *   // User simulator with custom persona
+ *   const expertResult = await run({
+ *     name: "Expert Developer Test",
+ *     description: "Testing with a technical expert user persona.",
+ *     agents: [
+ *       myAgent,
+ *       userSimulatorAgent({
+ *         systemPrompt: `
+ *           You are an expert software developer testing an AI coding assistant.
+ *           Ask challenging, technical questions and be demanding about code quality.
+ *           Use technical jargon and expect detailed, accurate responses.
+ *         `
+ *       })
+ *     ],
+ *     script: [
+ *       user(),
+ *       agent(),
+ *     ],
+ *   });
  * }
  * main();
  * ```
+ *
+ * @note
+ * - Uses role reversal internally to work around LLM biases toward assistant roles
  */
 declare const userSimulatorAgent: (config?: TestingAgentConfig) => {
     role: AgentRole.USER;
@@ -538,6 +629,16 @@ declare const userSimulatorAgent: (config?: TestingAgentConfig) => {
     }>;
 };
 
+type agents_FinishTestArgs = FinishTestArgs;
+type agents_JudgeAgentConfig = JudgeAgentConfig;
+type agents_TestingAgentConfig = TestingAgentConfig;
+type agents_TestingAgentInferenceConfig = TestingAgentInferenceConfig;
+declare const agents_judgeAgent: typeof judgeAgent;
+declare const agents_userSimulatorAgent: typeof userSimulatorAgent;
+declare namespace agents {
+  export { type agents_FinishTestArgs as FinishTestArgs, type agents_JudgeAgentConfig as JudgeAgentConfig, type agents_TestingAgentConfig as TestingAgentConfig, type agents_TestingAgentInferenceConfig as TestingAgentInferenceConfig, agents_judgeAgent as judgeAgent, agents_userSimulatorAgent as userSimulatorAgent };
+}
+
 /**
  * Verdict enum represents the possible outcomes of a test scenario
  */
@@ -1107,6 +1208,14 @@ declare class ScenarioExecutionState implements ScenarioExecutionStateLike {
     hasToolCall(toolName: string): boolean;
 }
 
+type execution_ScenarioExecution = ScenarioExecution;
+declare const execution_ScenarioExecution: typeof ScenarioExecution;
+type execution_ScenarioExecutionState = ScenarioExecutionState;
+declare const execution_ScenarioExecutionState: typeof ScenarioExecutionState;
+declare namespace execution {
+  export { execution_ScenarioExecution as ScenarioExecution, execution_ScenarioExecutionState as ScenarioExecutionState };
+}
+
 /**
  * High-level interface for running a scenario test.
  *
@@ -1158,6 +1267,11 @@ declare class ScenarioExecutionState implements ScenarioExecutionStateLike {
  */
 declare function run(cfg: ScenarioConfig): Promise<ScenarioResult>;
 
+declare const runner_run: typeof run;
+declare namespace runner {
+  export { runner_run as run };
+}
+
 /**
  * Scenario script DSL (Domain Specific Language) module.
  *
@@ -1247,74 +1361,18 @@ declare const succeed: (reasoning?: string) => ScriptStep;
  */
 declare const fail: (reasoning?: string) => ScriptStep;
 
-declare const scenario: {
-    message: (message: ai.CoreMessage) => ScriptStep;
-    agent: (content?: string | ai.CoreMessage) => ScriptStep;
-    judge: (content?: string | ai.CoreMessage) => ScriptStep;
-    user: (content?: string | ai.CoreMessage) => ScriptStep;
-    proceed: (turns?: number, onTurn?: (state: ScenarioExecutionStateLike) => void | Promise<void>, onStep?: (state: ScenarioExecutionStateLike) => void | Promise<void>) => ScriptStep;
-    succeed: (reasoning?: string) => ScriptStep;
-    fail: (reasoning?: string) => ScriptStep;
-    run(cfg: ScenarioConfig): Promise<ScenarioResult>;
-    ScenarioExecution: typeof ScenarioExecution;
-    ScenarioExecutionState: typeof ScenarioExecutionState;
-    defineConfig(config: ScenarioProjectConfig): ScenarioProjectConfig;
-    scenarioProjectConfigSchema: zod.ZodObject<{
-        defaultModel: zod.ZodOptional<zod.ZodObject<{
-            model: zod.ZodType<ai.LanguageModelV1, zod.ZodTypeDef, ai.LanguageModelV1>;
-            temperature: zod.ZodDefault<zod.ZodOptional<zod.ZodNumber>>;
-            maxTokens: zod.ZodOptional<zod.ZodNumber>;
-        }, "strip", zod.ZodTypeAny, {
-            model: ai.LanguageModelV1;
-            temperature: number;
-            maxTokens?: number | undefined;
-        }, {
-            model: ai.LanguageModelV1;
-            temperature?: number | undefined;
-            maxTokens?: number | undefined;
-        }>>;
-        langwatchEndpoint: zod.ZodOptional<zod.ZodString>;
-        langwatchApiKey: zod.ZodOptional<zod.ZodString>;
-    }, "strict", zod.ZodTypeAny, {
-        defaultModel?: {
-            model: ai.LanguageModelV1;
-            temperature: number;
-            maxTokens?: number | undefined;
-        } | undefined;
-        langwatchEndpoint?: string | undefined;
-        langwatchApiKey?: string | undefined;
-    }, {
-        defaultModel?: {
-            model: ai.LanguageModelV1;
-            temperature?: number | undefined;
-            maxTokens?: number | undefined;
-        } | undefined;
-        langwatchEndpoint?: string | undefined;
-        langwatchApiKey?: string | undefined;
-    }>;
-    AgentRole: typeof AgentRole;
-    allAgentRoles: readonly [AgentRole.USER, AgentRole.AGENT, AgentRole.JUDGE];
-    AgentAdapter: typeof AgentAdapter;
-    UserSimulatorAgentAdapter: typeof UserSimulatorAgentAdapter;
-    JudgeAgentAdapter: typeof JudgeAgentAdapter;
-    judgeAgent: (cfg: JudgeAgentConfig) => {
-        role: AgentRole.JUDGE;
-        criteria: string[];
-        call: (input: AgentInput) => Promise<never[] | {
-            success: boolean;
-            messages: ai.CoreMessage[];
-            reasoning: string;
-            metCriteria: string[];
-            unmetCriteria: string[];
-        }>;
-    };
-    userSimulatorAgent: (config?: TestingAgentConfig) => {
-        role: AgentRole.USER;
-        call: (input: AgentInput) => Promise<{
-            role: "user";
-            content: string;
-        }>;
-    };
-};
+declare const script_agent: typeof agent;
+declare const script_fail: typeof fail;
+declare const script_judge: typeof judge;
+declare const script_message: typeof message;
+declare const script_proceed: typeof proceed;
+declare const script_succeed: typeof succeed;
+declare const script_user: typeof user;
+declare namespace script {
+  export { script_agent as agent, script_fail as fail, script_judge as judge, script_message as message, script_proceed as proceed, script_succeed as succeed, script_user as user };
+}
+
+type ScenarioApi = typeof agents & typeof domain & typeof execution & typeof runner & typeof script;
+declare const scenario: ScenarioApi;
 
-export { AgentAdapter, type AgentInput, type AgentReturnTypes, AgentRole, type FinishTestArgs, JudgeAgentAdapter, type JudgeAgentConfig, type ScenarioConfig, type ScenarioConfigFinal, ScenarioExecution, type ScenarioExecutionLike, ScenarioExecutionState, type ScenarioExecutionStateLike, type ScenarioProjectConfig, type ScenarioResult, type ScriptStep, type TestingAgentConfig, type TestingAgentInferenceConfig, UserSimulatorAgentAdapter, agent, allAgentRoles, scenario as default, defineConfig, fail, judge, judgeAgent, message, proceed, run, scenario, scenarioProjectConfigSchema, succeed, user, userSimulatorAgent };
+export { AgentAdapter, type AgentInput, type AgentReturnTypes, AgentRole, DEFAULT_MAX_TURNS, DEFAULT_TEMPERATURE, DEFAULT_VERBOSE, type FinishTestArgs, JudgeAgentAdapter, type JudgeAgentConfig, type ScenarioConfig, type ScenarioConfigFinal, ScenarioExecution, type ScenarioExecutionLike, ScenarioExecutionState, type ScenarioExecutionStateLike, type ScenarioProjectConfig, type ScenarioResult, type ScriptStep, type TestingAgentConfig, type TestingAgentInferenceConfig, UserSimulatorAgentAdapter, agent, allAgentRoles, scenario as default, defineConfig, fail, judge, judgeAgent, message, proceed, run, scenario, scenarioProjectConfigSchema, succeed, user, userSimulatorAgent };

package/dist/index.d.ts

@@ -1,7 +1,6 @@
-import * as zod from 'zod';
-import { z } from 'zod';
 import * as ai from 'ai';
 import { CoreMessage, CoreToolMessage, LanguageModel } from 'ai';
+import { z } from 'zod';
 import { Observable } from 'rxjs';
 
 declare enum AgentRole {
@@ -121,6 +120,8 @@ declare abstract class JudgeAgentAdapter implements AgentAdapter {
     abstract call(input: AgentInput): Promise<AgentReturnTypes>;
 }
 
+declare const DEFAULT_MAX_TURNS = 10;
+declare const DEFAULT_VERBOSE = false;
 /**
  * Configuration for a scenario.
  */
@@ -147,11 +148,19 @@ interface ScenarioConfig {
      */
     script?: ScriptStep[];
     /**
-     * Whether to output verbose logging. Defaults to false.
+     * Whether to output verbose logging.
+     *
+     * If no value is provided, this defaults to {@link DEFAULT_VERBOSE}.
+     *
+     * @default {@link DEFAULT_VERBOSE}
      */
     verbose?: boolean;
     /**
-     * The maximum number of turns to execute. Defaults to 20.
+     * The maximum number of turns to execute.
+     *
+     * If no value is provided, this defaults to {@link DEFAULT_MAX_TURNS}.
+     *
+     * @default {@link DEFAULT_MAX_TURNS}
      */
     maxTurns?: number;
     /**
@@ -337,6 +346,8 @@ interface ScenarioExecutionStateLike {
     hasToolCall(toolName: string): boolean;
 }
 
+/** Default temperature for language model inference */
+declare const DEFAULT_TEMPERATURE = 0;
 declare const scenarioProjectConfigSchema: z.ZodObject<{
     defaultModel: z.ZodOptional<z.ZodObject<{
         model: z.ZodType<ai.LanguageModelV1, z.ZodTypeDef, ai.LanguageModelV1>;
@@ -351,28 +362,49 @@ declare const scenarioProjectConfigSchema: z.ZodObject<{
         temperature?: number | undefined;
         maxTokens?: number | undefined;
     }>>;
-    langwatchEndpoint: z.ZodOptional<z.ZodString>;
-    langwatchApiKey: z.ZodOptional<z.ZodString>;
 }, "strict", z.ZodTypeAny, {
     defaultModel?: {
         model: ai.LanguageModelV1;
         temperature: number;
         maxTokens?: number | undefined;
     } | undefined;
-    langwatchEndpoint?: string | undefined;
-    langwatchApiKey?: string | undefined;
 }, {
     defaultModel?: {
         model: ai.LanguageModelV1;
         temperature?: number | undefined;
         maxTokens?: number | undefined;
     } | undefined;
-    langwatchEndpoint?: string | undefined;
-    langwatchApiKey?: string | undefined;
 }>;
 type ScenarioProjectConfig = z.infer<typeof scenarioProjectConfigSchema>;
 declare function defineConfig(config: ScenarioProjectConfig): ScenarioProjectConfig;
 
+type domain_AgentAdapter = AgentAdapter;
+declare const domain_AgentAdapter: typeof AgentAdapter;
+type domain_AgentInput = AgentInput;
+type domain_AgentReturnTypes = AgentReturnTypes;
+type domain_AgentRole = AgentRole;
+declare const domain_AgentRole: typeof AgentRole;
+declare const domain_DEFAULT_MAX_TURNS: typeof DEFAULT_MAX_TURNS;
+declare const domain_DEFAULT_TEMPERATURE: typeof DEFAULT_TEMPERATURE;
+declare const domain_DEFAULT_VERBOSE: typeof DEFAULT_VERBOSE;
+type domain_JudgeAgentAdapter = JudgeAgentAdapter;
+declare const domain_JudgeAgentAdapter: typeof JudgeAgentAdapter;
+type domain_ScenarioConfig = ScenarioConfig;
+type domain_ScenarioConfigFinal = ScenarioConfigFinal;
+type domain_ScenarioExecutionLike = ScenarioExecutionLike;
+type domain_ScenarioExecutionStateLike = ScenarioExecutionStateLike;
+type domain_ScenarioProjectConfig = ScenarioProjectConfig;
+type domain_ScenarioResult = ScenarioResult;
+type domain_ScriptStep = ScriptStep;
+type domain_UserSimulatorAgentAdapter = UserSimulatorAgentAdapter;
+declare const domain_UserSimulatorAgentAdapter: typeof UserSimulatorAgentAdapter;
+declare const domain_allAgentRoles: typeof allAgentRoles;
+declare const domain_defineConfig: typeof defineConfig;
+declare const domain_scenarioProjectConfigSchema: typeof scenarioProjectConfigSchema;
+declare namespace domain {
+  export { domain_AgentAdapter as AgentAdapter, type domain_AgentInput as AgentInput, type domain_AgentReturnTypes as AgentReturnTypes, domain_AgentRole as AgentRole, domain_DEFAULT_MAX_TURNS as DEFAULT_MAX_TURNS, domain_DEFAULT_TEMPERATURE as DEFAULT_TEMPERATURE, domain_DEFAULT_VERBOSE as DEFAULT_VERBOSE, domain_JudgeAgentAdapter as JudgeAgentAdapter, type domain_ScenarioConfig as ScenarioConfig, type domain_ScenarioConfigFinal as ScenarioConfigFinal, type domain_ScenarioExecutionLike as ScenarioExecutionLike, type domain_ScenarioExecutionStateLike as ScenarioExecutionStateLike, type domain_ScenarioProjectConfig as ScenarioProjectConfig, type domain_ScenarioResult as ScenarioResult, type domain_ScriptStep as ScriptStep, domain_UserSimulatorAgentAdapter as UserSimulatorAgentAdapter, domain_allAgentRoles as allAgentRoles, domain_defineConfig as defineConfig, domain_scenarioProjectConfigSchema as scenarioProjectConfigSchema };
+}
+
 /**
  * Configuration for the inference parameters of a testing agent.
  */
@@ -400,6 +432,14 @@ interface TestingAgentConfig extends TestingAgentInferenceConfig {
      * The name of the agent.
      */
     name?: string;
+    /**
+     * System prompt to use for the agent.
+     *
+     * Useful in more complex scenarios where you want to set the system prompt
+     * for the agent directly. If left blank, this will be automatically generated
+     * from the scenario description.
+     */
+    systemPrompt?: string;
 }
 /**
  * The arguments for finishing a test, used by the judge agent's tool.
@@ -502,8 +542,17 @@ declare const judgeAgent: (cfg: JudgeAgentConfig) => {
  *
  * @param config Optional configuration for the agent.
  * @param config.model The language model to use for generating responses.
- * @param config.temperature The temperature to use for the model.
+ *                     If not provided, a default model will be used.
+ * @param config.temperature The temperature for the language model (0.0-1.0).
+ *                          Lower values make responses more deterministic.
+ *                          Defaults to {@link DEFAULT_TEMPERATURE}.
  * @param config.maxTokens The maximum number of tokens to generate.
+ *                        If not provided, uses model defaults.
+ * @param config.name The name of the agent.
+ * @param config.systemPrompt Custom system prompt to override default user simulation behavior.
+ *                           Use this to create specialized user personas or behaviors.
+ *
+ * @throws {Error} If no model is configured either in parameters or global config.
  *
  * @example
  * ```typescript
@@ -517,7 +566,8 @@ declare const judgeAgent: (cfg: JudgeAgentConfig) => {
  * };
  *
  * async function main() {
- *   const result = await run({
+ *   // Basic user simulator with default behavior
+ *   const basicResult = await run({
  *     name: "User Simulator Test",
  *     description: "A simple test to see if the user simulator works.",
  *     agents: [myAgent, userSimulatorAgent()],
@@ -526,9 +576,50 @@ declare const judgeAgent: (cfg: JudgeAgentConfig) => {
  *       agent(),
  *     ],
  *   });
+ *
+ *   // Customized user simulator
+ *   const customResult = await run({
+ *     name: "Expert User Test",
+ *     description: "User seeks help with TypeScript programming",
+ *     agents: [
+ *       myAgent,
+ *       userSimulatorAgent({
+ *         model: openai("gpt-4"),
+ *         temperature: 0.3,
+ *         systemPrompt: "You are a technical user who asks detailed questions"
+ *       })
+ *     ],
+ *     script: [
+ *       user(),
+ *       agent(),
+ *     ],
+ *   });
+ *
+ *   // User simulator with custom persona
+ *   const expertResult = await run({
+ *     name: "Expert Developer Test",
+ *     description: "Testing with a technical expert user persona.",
+ *     agents: [
+ *       myAgent,
+ *       userSimulatorAgent({
+ *         systemPrompt: `
+ *           You are an expert software developer testing an AI coding assistant.
+ *           Ask challenging, technical questions and be demanding about code quality.
+ *           Use technical jargon and expect detailed, accurate responses.
+ *         `
+ *       })
+ *     ],
+ *     script: [
+ *       user(),
+ *       agent(),
+ *     ],
+ *   });
  * }
  * main();
  * ```
+ *
+ * @note
+ * - Uses role reversal internally to work around LLM biases toward assistant roles
  */
 declare const userSimulatorAgent: (config?: TestingAgentConfig) => {
     role: AgentRole.USER;
@@ -538,6 +629,16 @@ declare const userSimulatorAgent: (config?: TestingAgentConfig) => {
     }>;
 };
 
+type agents_FinishTestArgs = FinishTestArgs;
+type agents_JudgeAgentConfig = JudgeAgentConfig;
+type agents_TestingAgentConfig = TestingAgentConfig;
+type agents_TestingAgentInferenceConfig = TestingAgentInferenceConfig;
+declare const agents_judgeAgent: typeof judgeAgent;
+declare const agents_userSimulatorAgent: typeof userSimulatorAgent;
+declare namespace agents {
+  export { type agents_FinishTestArgs as FinishTestArgs, type agents_JudgeAgentConfig as JudgeAgentConfig, type agents_TestingAgentConfig as TestingAgentConfig, type agents_TestingAgentInferenceConfig as TestingAgentInferenceConfig, agents_judgeAgent as judgeAgent, agents_userSimulatorAgent as userSimulatorAgent };
+}
+
 /**
  * Verdict enum represents the possible outcomes of a test scenario
  */
@@ -1107,6 +1208,14 @@ declare class ScenarioExecutionState implements ScenarioExecutionStateLike {
     hasToolCall(toolName: string): boolean;
 }
 
+type execution_ScenarioExecution = ScenarioExecution;
+declare const execution_ScenarioExecution: typeof ScenarioExecution;
+type execution_ScenarioExecutionState = ScenarioExecutionState;
+declare const execution_ScenarioExecutionState: typeof ScenarioExecutionState;
+declare namespace execution {
+  export { execution_ScenarioExecution as ScenarioExecution, execution_ScenarioExecutionState as ScenarioExecutionState };
+}
+
 /**
  * High-level interface for running a scenario test.
  *
@@ -1158,6 +1267,11 @@ declare class ScenarioExecutionState implements ScenarioExecutionStateLike {
  */
 declare function run(cfg: ScenarioConfig): Promise<ScenarioResult>;
 
+declare const runner_run: typeof run;
+declare namespace runner {
+  export { runner_run as run };
+}
+
 /**
  * Scenario script DSL (Domain Specific Language) module.
  *
@@ -1247,74 +1361,18 @@ declare const succeed: (reasoning?: string) => ScriptStep;
  */
 declare const fail: (reasoning?: string) => ScriptStep;
 
-declare const scenario: {
-    message: (message: ai.CoreMessage) => ScriptStep;
-    agent: (content?: string | ai.CoreMessage) => ScriptStep;
-    judge: (content?: string | ai.CoreMessage) => ScriptStep;
-    user: (content?: string | ai.CoreMessage) => ScriptStep;
-    proceed: (turns?: number, onTurn?: (state: ScenarioExecutionStateLike) => void | Promise<void>, onStep?: (state: ScenarioExecutionStateLike) => void | Promise<void>) => ScriptStep;
-    succeed: (reasoning?: string) => ScriptStep;
-    fail: (reasoning?: string) => ScriptStep;
-    run(cfg: ScenarioConfig): Promise<ScenarioResult>;
-    ScenarioExecution: typeof ScenarioExecution;
-    ScenarioExecutionState: typeof ScenarioExecutionState;
-    defineConfig(config: ScenarioProjectConfig): ScenarioProjectConfig;
-    scenarioProjectConfigSchema: zod.ZodObject<{
-        defaultModel: zod.ZodOptional<zod.ZodObject<{
-            model: zod.ZodType<ai.LanguageModelV1, zod.ZodTypeDef, ai.LanguageModelV1>;
-            temperature: zod.ZodDefault<zod.ZodOptional<zod.ZodNumber>>;
-            maxTokens: zod.ZodOptional<zod.ZodNumber>;
-        }, "strip", zod.ZodTypeAny, {
-            model: ai.LanguageModelV1;
-            temperature: number;
-            maxTokens?: number | undefined;
-        }, {
-            model: ai.LanguageModelV1;
-            temperature?: number | undefined;
-            maxTokens?: number | undefined;
-        }>>;
-        langwatchEndpoint: zod.ZodOptional<zod.ZodString>;
-        langwatchApiKey: zod.ZodOptional<zod.ZodString>;
-    }, "strict", zod.ZodTypeAny, {
-        defaultModel?: {
-            model: ai.LanguageModelV1;
-            temperature: number;
-            maxTokens?: number | undefined;
-        } | undefined;
-        langwatchEndpoint?: string | undefined;
-        langwatchApiKey?: string | undefined;
-    }, {
-        defaultModel?: {
-            model: ai.LanguageModelV1;
-            temperature?: number | undefined;
-            maxTokens?: number | undefined;
-        } | undefined;
-        langwatchEndpoint?: string | undefined;
-        langwatchApiKey?: string | undefined;
-    }>;
-    AgentRole: typeof AgentRole;
-    allAgentRoles: readonly [AgentRole.USER, AgentRole.AGENT, AgentRole.JUDGE];
-    AgentAdapter: typeof AgentAdapter;
-    UserSimulatorAgentAdapter: typeof UserSimulatorAgentAdapter;
-    JudgeAgentAdapter: typeof JudgeAgentAdapter;
-    judgeAgent: (cfg: JudgeAgentConfig) => {
-        role: AgentRole.JUDGE;
-        criteria: string[];
-        call: (input: AgentInput) => Promise<never[] | {
-            success: boolean;
-            messages: ai.CoreMessage[];
-            reasoning: string;
-            metCriteria: string[];
-            unmetCriteria: string[];
-        }>;
-    };
-    userSimulatorAgent: (config?: TestingAgentConfig) => {
-        role: AgentRole.USER;
-        call: (input: AgentInput) => Promise<{
-            role: "user";
-            content: string;
-        }>;
-    };
-};
+declare const script_agent: typeof agent;
+declare const script_fail: typeof fail;
+declare const script_judge: typeof judge;
+declare const script_message: typeof message;
+declare const script_proceed: typeof proceed;
+declare const script_succeed: typeof succeed;
+declare const script_user: typeof user;
+declare namespace script {
+  export { script_agent as agent, script_fail as fail, script_judge as judge, script_message as message, script_proceed as proceed, script_succeed as succeed, script_user as user };
+}
+
+type ScenarioApi = typeof agents & typeof domain & typeof execution & typeof runner & typeof script;
+declare const scenario: ScenarioApi;
 
-export { AgentAdapter, type AgentInput, type AgentReturnTypes, AgentRole, type FinishTestArgs, JudgeAgentAdapter, type JudgeAgentConfig, type ScenarioConfig, type ScenarioConfigFinal, ScenarioExecution, type ScenarioExecutionLike, ScenarioExecutionState, type ScenarioExecutionStateLike, type ScenarioProjectConfig, type ScenarioResult, type ScriptStep, type TestingAgentConfig, type TestingAgentInferenceConfig, UserSimulatorAgentAdapter, agent, allAgentRoles, scenario as default, defineConfig, fail, judge, judgeAgent, message, proceed, run, scenario, scenarioProjectConfigSchema, succeed, user, userSimulatorAgent };
+export { AgentAdapter, type AgentInput, type AgentReturnTypes, AgentRole, DEFAULT_MAX_TURNS, DEFAULT_TEMPERATURE, DEFAULT_VERBOSE, type FinishTestArgs, JudgeAgentAdapter, type JudgeAgentConfig, type ScenarioConfig, type ScenarioConfigFinal, ScenarioExecution, type ScenarioExecutionLike, ScenarioExecutionState, type ScenarioExecutionStateLike, type ScenarioProjectConfig, type ScenarioResult, type ScriptStep, type TestingAgentConfig, type TestingAgentInferenceConfig, UserSimulatorAgentAdapter, agent, allAgentRoles, scenario as default, defineConfig, fail, judge, judgeAgent, message, proceed, run, scenario, scenarioProjectConfigSchema, succeed, user, userSimulatorAgent };

package/dist/index.js

@@ -32,6 +32,9 @@ var index_exports = {};
 __export(index_exports, {
   AgentAdapter: () => AgentAdapter,
   AgentRole: () => AgentRole,
+  DEFAULT_MAX_TURNS: () => DEFAULT_MAX_TURNS,
+  DEFAULT_TEMPERATURE: () => DEFAULT_TEMPERATURE,
+  DEFAULT_VERBOSE: () => DEFAULT_VERBOSE,
   JudgeAgentAdapter: () => JudgeAgentAdapter,
   ScenarioExecution: () => ScenarioExecution,
   ScenarioExecutionState: () => ScenarioExecutionState,
@@ -70,6 +73,9 @@ var domain_exports = {};
 __export(domain_exports, {
   AgentAdapter: () => AgentAdapter,
   AgentRole: () => AgentRole,
+  DEFAULT_MAX_TURNS: () => DEFAULT_MAX_TURNS,
+  DEFAULT_TEMPERATURE: () => DEFAULT_TEMPERATURE,
+  DEFAULT_VERBOSE: () => DEFAULT_VERBOSE,
   JudgeAgentAdapter: () => JudgeAgentAdapter,
   UserSimulatorAgentAdapter: () => UserSimulatorAgentAdapter,
   allAgentRoles: () => allAgentRoles,
@@ -79,14 +85,13 @@ __export(domain_exports, {
 
 // src/domain/core/config.ts
 var import_zod = require("zod");
+var DEFAULT_TEMPERATURE = 0;
 var scenarioProjectConfigSchema = import_zod.z.object({
   defaultModel: import_zod.z.object({
     model: import_zod.z.custom(),
-    temperature: import_zod.z.number().min(0).max(1).optional().default(0),
+    temperature: import_zod.z.number().min(0).max(1).optional().default(DEFAULT_TEMPERATURE),
     maxTokens: import_zod.z.number().optional()
-  }).optional(),
-  langwatchEndpoint: import_zod.z.string().optional(),
-  langwatchApiKey: import_zod.z.string().optional()
+  }).optional()
 }).strict();
 function defineConfig(config2) {
   return config2;
@@ -119,6 +124,10 @@ var JudgeAgentAdapter = class {
   }
 };
 
+// src/domain/scenarios/index.ts
+var DEFAULT_MAX_TURNS = 10;
+var DEFAULT_VERBOSE = false;
+
 // src/agents/utils.ts
 var toolMessageRole = "tool";
 var assistantMessageRole = "assistant";
@@ -339,7 +348,7 @@ async function loadProjectConfig() {
   configLoadPromise = (async () => {
     try {
       config = await loadScenarioProjectConfig();
-      logger.info("loaded scenario project config", { config });
+      logger.debug("loaded scenario project config", { config });
     } catch (error) {
       logger.error("error loading scenario project config", { error });
     } finally {
@@ -534,7 +543,7 @@ var userSimulatorAgent = (config2) => {
   return {
     role: "User" /* USER */,
     call: async (input) => {
-      const systemPrompt = buildSystemPrompt2(input.scenarioConfig.description);
+      const systemPrompt = (config2 == null ? void 0 : config2.systemPrompt) ?? buildSystemPrompt2(input.scenarioConfig.description);
       const messages = [
         { role: "system", content: systemPrompt },
         { role: "assistant", content: "Hello, how can I help you today" },
@@ -549,7 +558,7 @@ var userSimulatorAgent = (config2) => {
       const completion = await (0, import_ai2.generateText)({
         model: mergedConfig.model,
         messages: reversedMessages,
-        temperature: mergedConfig.temperature ?? 0,
+        temperature: mergedConfig.temperature ?? DEFAULT_TEMPERATURE,
         maxTokens: mergedConfig.maxTokens
       });
       const messageContent = completion.text;
@@ -735,6 +744,77 @@ var stateSchema = import_zod4.z.object({
 var runsSchema = import_zod4.z.object({ runs: import_zod4.z.array(import_zod4.z.string()) });
 var eventsSchema = import_zod4.z.object({ events: import_zod4.z.array(scenarioEventSchema) });
 
+// src/utils/message-conversion.ts
+function convertCoreMessagesToAguiMessages(coreMessages) {
+  const aguiMessages = [];
+  for (const msg of coreMessages) {
+    const id = "id" in msg && typeof msg.id === "string" ? msg.id : generateMessageId();
+    switch (true) {
+      case msg.role === "system":
+        aguiMessages.push({
+          id,
+          role: "system",
+          content: msg.content
+        });
+        break;
+      case (msg.role === "user" && typeof msg.content === "string"):
+        aguiMessages.push({
+          id,
+          role: "user",
+          content: msg.content
+        });
+        break;
+      // Handle any other user message content format
+      case (msg.role === "user" && Array.isArray(msg.content)):
+        aguiMessages.push({
+          id,
+          role: "user",
+          content: JSON.stringify(msg.content)
+        });
+        break;
+      case (msg.role === "assistant" && typeof msg.content === "string"):
+        aguiMessages.push({
+          id,
+          role: "assistant",
+          content: msg.content
+        });
+        break;
+      case (msg.role === "assistant" && Array.isArray(msg.content)): {
+        const toolCalls = msg.content.filter((p) => p.type === "tool-call");
+        const nonToolCalls = msg.content.filter((p) => p.type !== "tool-call");
+        aguiMessages.push({
+          id,
+          role: "assistant",
+          content: JSON.stringify(nonToolCalls),
+          toolCalls: toolCalls.map((c) => ({
+            id: c.toolCallId,
+            type: "function",
+            function: {
+              name: c.toolName,
+              arguments: JSON.stringify(c.args)
+            }
+          }))
+        });
+        break;
+      }
+      case msg.role === "tool":
+        msg.content.map((p, i) => {
+          aguiMessages.push({
+            id: `${id}-${i}`,
+            role: "tool",
+            toolCallId: p.toolCallId,
+            content: JSON.stringify(p.result)
+          });
+        });
+        break;
+      default:
+        throw new Error(`Unsupported message role: ${msg.role}`);
+    }
+  }
+  return aguiMessages;
+}
+var message_conversion_default = convertCoreMessagesToAguiMessages;
+
 // src/execution/scenario-execution.ts
 var batchRunId = getBatchRunId();
 var ScenarioExecution = class {
@@ -766,8 +846,8 @@ var ScenarioExecution = class {
       description: config2.description,
       agents: config2.agents,
       script,
-      verbose: config2.verbose ?? false,
-      maxTurns: config2.maxTurns ?? 10,
+      verbose: config2.verbose ?? DEFAULT_VERBOSE,
+      maxTurns: config2.maxTurns ?? DEFAULT_MAX_TURNS,
       threadId: config2.threadId ?? generateThreadId(),
       setId: config2.setId
     };
@@ -819,12 +899,14 @@ var ScenarioExecution = class {
         }
       }
       this.emitRunFinished({ scenarioRunId, status: "FAILED" /* FAILED */ });
-      return this.reachedMaxTurns([
-        "Reached end of script without conclusion, add one of the following to the end of the script:",
-        "- `Scenario.proceed()` to let the simulation continue to play out",
-        "- `Scenario.judge()` to force criteria judgement",
-        "- `Scenario.succeed()` or `Scenario.fail()` to end the test with an explicit result"
-      ].join("\n"));
+      return this.reachedMaxTurns(
+        [
+          "Reached end of script without conclusion, add one of the following to the end of the script:",
+          "- `Scenario.proceed()` to let the simulation continue to play out",
+          "- `Scenario.judge()` to force criteria judgement",
+          "- `Scenario.succeed()` or `Scenario.fail()` to end the test with an explicit result"
+        ].join("\n")
+      );
     } catch (error) {
       const errorResult = {
         success: false,
@@ -959,8 +1041,7 @@ var ScenarioExecution = class {
     while (true) {
       const goToNextTurn = turns === void 0 || initialTurn === null || this.state.currentTurn != null && this.state.currentTurn + 1 < initialTurn + turns;
       const nextMessage = await this._step(goToNextTurn, onTurn);
-      if (initialTurn === null)
-        initialTurn = this.state.currentTurn;
+      if (initialTurn === null) initialTurn = this.state.currentTurn;
       if (nextMessage === null) {
         return null;
       }
@@ -1046,7 +1127,10 @@ var ScenarioExecution = class {
     agent2 = nextAgent.agent;
     this.removePendingAgent(agent2);
     if (content) {
-      const message2 = typeof content === "string" ? { role: role === "User" /* USER */ ? "user" : "assistant", content } : content;
+      const message2 = typeof content === "string" ? {
+        role: role === "User" /* USER */ ? "user" : "assistant",
+        content
+      } : content;
       this.state.addMessage(message2);
       this.broadcastMessage(message2, index);
       return null;
@@ -1119,7 +1203,9 @@ var ScenarioExecution = class {
   reachedMaxTurns(errorMessage) {
     var _a;
     const agentRoleAgentsIdx = this.agents.map((agent2, i) => ({ agent: agent2, idx: i })).filter(({ agent: agent2 }) => agent2.role === "Agent" /* AGENT */).map(({ idx }) => idx);
-    const agentTimes = agentRoleAgentsIdx.map((i) => this.agentTimes.get(i) || 0);
+    const agentTimes = agentRoleAgentsIdx.map(
+      (i) => this.agentTimes.get(i) || 0
+    );
     const totalAgentTime = agentTimes.reduce((sum, time) => sum + time, 0);
     return {
       success: false,
@@ -1174,7 +1260,7 @@ var ScenarioExecution = class {
     this.emitEvent({
       ...this.makeBaseEvent({ scenarioRunId }),
       type: "SCENARIO_MESSAGE_SNAPSHOT" /* MESSAGE_SNAPSHOT */,
-      messages: this.state.messages
+      messages: message_conversion_default(this.state.messages)
       // Add any other required fields from MessagesSnapshotEventSchema
     });
   }
@@ -1221,10 +1307,8 @@ var ScenarioExecution = class {
 function convertAgentReturnTypesToMessages(response, role) {
   if (typeof response === "string")
     return [{ role, content: response }];
-  if (Array.isArray(response))
-    return response;
-  if (typeof response === "object" && "role" in response)
-    return [response];
+  if (Array.isArray(response)) return response;
+  if (typeof response === "object" && "role" in response) return [response];
   return [];
 }
 
@@ -1324,17 +1408,20 @@ var EventReporter = class {
   eventsEndpoint;
   eventAlertMessageLogger;
   logger = new Logger("scenario.events.EventReporter");
+  isEnabled;
   constructor(config2) {
     this.apiKey = config2.apiKey ?? "";
     this.eventsEndpoint = new URL("/api/scenario-events", config2.endpoint);
     this.eventAlertMessageLogger = new EventAlertMessageLogger();
     this.eventAlertMessageLogger.handleGreeting();
+    this.isEnabled = this.apiKey.length > 0 && this.eventsEndpoint.href.length > 0;
   }
   /**
    * Posts an event to the configured endpoint.
    * Logs success/failure but doesn't throw - event posting shouldn't break scenario execution.
    */
   async postEvent(event) {
+    if (!this.isEnabled) return {};
     const result = {};
     this.logger.debug(`[${event.type}] Posting event`, { event });
     const processedEvent = this.processEventForApi(event);
@@ -1559,10 +1646,9 @@ async function run(cfg) {
   let eventBus = null;
   let subscription = null;
   try {
-    const projectConfig = await loadScenarioProjectConfig();
     eventBus = new EventBus({
-      endpoint: projectConfig.langwatchEndpoint ?? process.env.LANGWATCH_ENDPOINT ?? "https://app.langwatch.ai",
-      apiKey: projectConfig.langwatchApiKey ?? process.env.LANGWATCH_API_KEY
+      endpoint: env.LANGWATCH_ENDPOINT,
+      apiKey: env.LANGWATCH_API_KEY
     });
     eventBus.listen();
     subscription = eventBus.subscribeTo(execution.events$);
@@ -1638,6 +1724,9 @@ var index_default = scenario;
 0 && (module.exports = {
   AgentAdapter,
   AgentRole,
+  DEFAULT_MAX_TURNS,
+  DEFAULT_TEMPERATURE,
+  DEFAULT_VERBOSE,
   JudgeAgentAdapter,
   ScenarioExecution,
   ScenarioExecutionState,

package/dist/index.mjs

@@ -1,6 +1,9 @@
 import {
   AgentAdapter,
   AgentRole,
+  DEFAULT_MAX_TURNS,
+  DEFAULT_TEMPERATURE,
+  DEFAULT_VERBOSE,
   EventBus,
   JudgeAgentAdapter,
   Logger,
@@ -8,15 +11,15 @@ import {
   allAgentRoles,
   defineConfig,
   domain_exports,
+  env,
   generateMessageId,
   generateScenarioId,
   generateScenarioRunId,
   generateThreadId,
   getBatchRunId,
   getProjectConfig,
-  loadScenarioProjectConfig,
   scenarioProjectConfigSchema
-} from "./chunk-ZMHTHRDR.mjs";
+} from "./chunk-MOOKAYIE.mjs";
 import {
   __export
 } from "./chunk-7P6ASYW6.mjs";
@@ -268,7 +271,7 @@ var userSimulatorAgent = (config) => {
   return {
     role: "User" /* USER */,
     call: async (input) => {
-      const systemPrompt = buildSystemPrompt2(input.scenarioConfig.description);
+      const systemPrompt = (config == null ? void 0 : config.systemPrompt) ?? buildSystemPrompt2(input.scenarioConfig.description);
       const messages = [
         { role: "system", content: systemPrompt },
         { role: "assistant", content: "Hello, how can I help you today" },
@@ -283,7 +286,7 @@ var userSimulatorAgent = (config) => {
       const completion = await generateText2({
         model: mergedConfig.model,
         messages: reversedMessages,
-        temperature: mergedConfig.temperature ?? 0,
+        temperature: mergedConfig.temperature ?? DEFAULT_TEMPERATURE,
         maxTokens: mergedConfig.maxTokens
       });
       const messageContent = completion.text;
@@ -376,6 +379,77 @@ var ScenarioExecutionState = class {
   }
 };
 
+// src/utils/message-conversion.ts
+function convertCoreMessagesToAguiMessages(coreMessages) {
+  const aguiMessages = [];
+  for (const msg of coreMessages) {
+    const id = "id" in msg && typeof msg.id === "string" ? msg.id : generateMessageId();
+    switch (true) {
+      case msg.role === "system":
+        aguiMessages.push({
+          id,
+          role: "system",
+          content: msg.content
+        });
+        break;
+      case (msg.role === "user" && typeof msg.content === "string"):
+        aguiMessages.push({
+          id,
+          role: "user",
+          content: msg.content
+        });
+        break;
+      // Handle any other user message content format
+      case (msg.role === "user" && Array.isArray(msg.content)):
+        aguiMessages.push({
+          id,
+          role: "user",
+          content: JSON.stringify(msg.content)
+        });
+        break;
+      case (msg.role === "assistant" && typeof msg.content === "string"):
+        aguiMessages.push({
+          id,
+          role: "assistant",
+          content: msg.content
+        });
+        break;
+      case (msg.role === "assistant" && Array.isArray(msg.content)): {
+        const toolCalls = msg.content.filter((p) => p.type === "tool-call");
+        const nonToolCalls = msg.content.filter((p) => p.type !== "tool-call");
+        aguiMessages.push({
+          id,
+          role: "assistant",
+          content: JSON.stringify(nonToolCalls),
+          toolCalls: toolCalls.map((c) => ({
+            id: c.toolCallId,
+            type: "function",
+            function: {
+              name: c.toolName,
+              arguments: JSON.stringify(c.args)
+            }
+          }))
+        });
+        break;
+      }
+      case msg.role === "tool":
+        msg.content.map((p, i) => {
+          aguiMessages.push({
+            id: `${id}-${i}`,
+            role: "tool",
+            toolCallId: p.toolCallId,
+            content: JSON.stringify(p.result)
+          });
+        });
+        break;
+      default:
+        throw new Error(`Unsupported message role: ${msg.role}`);
+    }
+  }
+  return aguiMessages;
+}
+var message_conversion_default = convertCoreMessagesToAguiMessages;
+
 // src/execution/scenario-execution.ts
 var batchRunId = getBatchRunId();
 var ScenarioExecution = class {
@@ -407,8 +481,8 @@ var ScenarioExecution = class {
       description: config.description,
       agents: config.agents,
       script,
-      verbose: config.verbose ?? false,
-      maxTurns: config.maxTurns ?? 10,
+      verbose: config.verbose ?? DEFAULT_VERBOSE,
+      maxTurns: config.maxTurns ?? DEFAULT_MAX_TURNS,
       threadId: config.threadId ?? generateThreadId(),
       setId: config.setId
     };
@@ -460,12 +534,14 @@ var ScenarioExecution = class {
         }
       }
       this.emitRunFinished({ scenarioRunId, status: "FAILED" /* FAILED */ });
-      return this.reachedMaxTurns([
-        "Reached end of script without conclusion, add one of the following to the end of the script:",
-        "- `Scenario.proceed()` to let the simulation continue to play out",
-        "- `Scenario.judge()` to force criteria judgement",
-        "- `Scenario.succeed()` or `Scenario.fail()` to end the test with an explicit result"
-      ].join("\n"));
+      return this.reachedMaxTurns(
+        [
+          "Reached end of script without conclusion, add one of the following to the end of the script:",
+          "- `Scenario.proceed()` to let the simulation continue to play out",
+          "- `Scenario.judge()` to force criteria judgement",
+          "- `Scenario.succeed()` or `Scenario.fail()` to end the test with an explicit result"
+        ].join("\n")
+      );
     } catch (error) {
       const errorResult = {
         success: false,
@@ -600,8 +676,7 @@ var ScenarioExecution = class {
     while (true) {
       const goToNextTurn = turns === void 0 || initialTurn === null || this.state.currentTurn != null && this.state.currentTurn + 1 < initialTurn + turns;
       const nextMessage = await this._step(goToNextTurn, onTurn);
-      if (initialTurn === null)
-        initialTurn = this.state.currentTurn;
+      if (initialTurn === null) initialTurn = this.state.currentTurn;
       if (nextMessage === null) {
         return null;
       }
@@ -687,7 +762,10 @@ var ScenarioExecution = class {
     agent2 = nextAgent.agent;
     this.removePendingAgent(agent2);
     if (content) {
-      const message2 = typeof content === "string" ? { role: role === "User" /* USER */ ? "user" : "assistant", content } : content;
+      const message2 = typeof content === "string" ? {
+        role: role === "User" /* USER */ ? "user" : "assistant",
+        content
+      } : content;
       this.state.addMessage(message2);
       this.broadcastMessage(message2, index);
       return null;
@@ -760,7 +838,9 @@ var ScenarioExecution = class {
   reachedMaxTurns(errorMessage) {
     var _a;
     const agentRoleAgentsIdx = this.agents.map((agent2, i) => ({ agent: agent2, idx: i })).filter(({ agent: agent2 }) => agent2.role === "Agent" /* AGENT */).map(({ idx }) => idx);
-    const agentTimes = agentRoleAgentsIdx.map((i) => this.agentTimes.get(i) || 0);
+    const agentTimes = agentRoleAgentsIdx.map(
+      (i) => this.agentTimes.get(i) || 0
+    );
     const totalAgentTime = agentTimes.reduce((sum, time) => sum + time, 0);
     return {
       success: false,
@@ -815,7 +895,7 @@ var ScenarioExecution = class {
     this.emitEvent({
       ...this.makeBaseEvent({ scenarioRunId }),
       type: "SCENARIO_MESSAGE_SNAPSHOT" /* MESSAGE_SNAPSHOT */,
-      messages: this.state.messages
+      messages: message_conversion_default(this.state.messages)
       // Add any other required fields from MessagesSnapshotEventSchema
     });
   }
@@ -862,10 +942,8 @@ var ScenarioExecution = class {
 function convertAgentReturnTypesToMessages(response, role) {
   if (typeof response === "string")
     return [{ role, content: response }];
-  if (Array.isArray(response))
-    return response;
-  if (typeof response === "object" && "role" in response)
-    return [response];
+  if (Array.isArray(response)) return response;
+  if (typeof response === "object" && "role" in response) return [response];
   return [];
 }
 
@@ -938,10 +1016,9 @@ async function run(cfg) {
   let eventBus = null;
   let subscription = null;
   try {
-    const projectConfig = await loadScenarioProjectConfig();
     eventBus = new EventBus({
-      endpoint: projectConfig.langwatchEndpoint ?? process.env.LANGWATCH_ENDPOINT ?? "https://app.langwatch.ai",
-      apiKey: projectConfig.langwatchApiKey ?? process.env.LANGWATCH_API_KEY
+      endpoint: env.LANGWATCH_ENDPOINT,
+      apiKey: env.LANGWATCH_API_KEY
     });
     eventBus.listen();
     subscription = eventBus.subscribeTo(execution.events$);
@@ -1016,6 +1093,9 @@ var index_default = scenario;
 export {
   AgentAdapter,
   AgentRole,
+  DEFAULT_MAX_TURNS,
+  DEFAULT_TEMPERATURE,
+  DEFAULT_VERBOSE,
   JudgeAgentAdapter,
   ScenarioExecution,
   ScenarioExecutionState,

package/dist/integrations/vitest/setup.js

@@ -37,14 +37,13 @@ var import_node_url = require("url");
 
 // src/domain/core/config.ts
 var import_zod = require("zod");
+var DEFAULT_TEMPERATURE = 0;
 var scenarioProjectConfigSchema = import_zod.z.object({
   defaultModel: import_zod.z.object({
     model: import_zod.z.custom(),
-    temperature: import_zod.z.number().min(0).max(1).optional().default(0),
+    temperature: import_zod.z.number().min(0).max(1).optional().default(DEFAULT_TEMPERATURE),
     maxTokens: import_zod.z.number().optional()
-  }).optional(),
-  langwatchEndpoint: import_zod.z.string().optional(),
-  langwatchApiKey: import_zod.z.string().optional()
+  }).optional()
 }).strict();
 
 // src/utils/logger.ts
@@ -335,17 +334,20 @@ var EventReporter = class {
   eventsEndpoint;
   eventAlertMessageLogger;
   logger = new Logger("scenario.events.EventReporter");
+  isEnabled;
   constructor(config) {
     this.apiKey = config.apiKey ?? "";
     this.eventsEndpoint = new URL("/api/scenario-events", config.endpoint);
     this.eventAlertMessageLogger = new EventAlertMessageLogger();
     this.eventAlertMessageLogger.handleGreeting();
+    this.isEnabled = this.apiKey.length > 0 && this.eventsEndpoint.href.length > 0;
   }
   /**
    * Posts an event to the configured endpoint.
    * Logs success/failure but doesn't throw - event posting shouldn't break scenario execution.
    */
   async postEvent(event) {
+    if (!this.isEnabled) return {};
     const result = {};
     this.logger.debug(`[${event.type}] Posting event`, { event });
     const processedEvent = this.processEventForApi(event);

package/dist/integrations/vitest/setup.mjs

@@ -1,6 +1,6 @@
 import {
   EventBus
-} from "../../chunk-ZMHTHRDR.mjs";
+} from "../../chunk-MOOKAYIE.mjs";
 import "../../chunk-7P6ASYW6.mjs";
 
 // src/integrations/vitest/setup.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@langwatch/scenario",
-  "version": "0.2.1",
+  "version": "0.2.6",
   "description": "A TypeScript library for testing AI agents using scenarios",
   "main": "dist/index.js",
   "module": "dist/index.mjs",
@@ -84,7 +84,7 @@
     "test": "vitest",
     "test:ci": "vitest run",
     "lint": "eslint .",
-    "examples:vitest:run": "export SCENARIO_BATCH_ID=scenariobatch_$(uuidgen) && pnpm run buildpack && (cd examples/vitest && pnpm install) && pnpm -F vitest-example run test",
-    "generate:api-reference": "npx typedoc src --out api-reference-docs"
+    "examples:vitest:run": "export SCENARIO_BATCH_ID=scenariobatch_$(uuidgen) && (cd examples/vitest && pnpm install) && pnpm -F vitest-example run test",
+    "generate:api-reference": "npx typedoc src --out api-reference-docs && rm -rf ../docs/docs/public/reference/javascript/scenario && mv api-reference-docs ../docs/docs/public/reference/javascript/scenario"
   }
 }