@empiricalrun/test-gen 0.79.3 → 0.79.5

package/CHANGELOG.md

@@ -1,5 +1,22 @@
 # @empiricalrun/test-gen
 
+## 0.79.5
+
+### Patch Changes
+
+- b58777e: refactor: extract CUA agent into standalone @empiricalrun/cua package
+- fda9bc3: feat: extract DashboardAPIClient into standalone @empiricalrun/dashboard-client package
+- Updated dependencies [b58777e]
+- Updated dependencies [fda9bc3]
+  - @empiricalrun/cua@0.2.0
+  - @empiricalrun/dashboard-client@0.2.0
+
+## 0.79.4
+
+### Patch Changes
+
+- dc2e6c9: feat: dashboard api client upgrade
+
 ## 0.79.3
 
 ### Patch Changes

package/dist/agent/browsing/run.d.ts

@@ -1,5 +1,5 @@
+import { BrowserAgentResult } from "@empiricalrun/cua";
 import { Usage } from "@empiricalrun/shared-types/chat-agent";
-import { BrowserAgentResult } from "../cua";
 type GenerateTestsType = {
     testCaseName: string;
     testCaseSuites: string[];

package/dist/agent/browsing/run.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"run.d.ts","sourceRoot":"","sources":["../../../src/agent/browsing/run.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,EAAE,MAAM,uCAAuC,CAAC;AAa9D,OAAO,EAAE,kBAAkB,EAAE,MAAM,QAAQ,CAAC;AAG5C,KAAK,iBAAiB,GAAG;IACvB,YAAY,EAAE,MAAM,CAAC;IACrB,cAAc,EAAE,MAAM,EAAE,CAAC;IACzB,YAAY,EAAE,MAAM,CAAC;IACrB,gBAAgB,EAAE,MAAM,CAAC;IACzB,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,CAAC;IACrB,OAAO,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,yBAAyB,EAAE,OAAO,CAAC;IACnC,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CACtC,CAAC;AAEF,wBAAsB,8BAA8B,CAAC,EACnD,gBAAgB,EAChB,OAAO,EACP,YAAY,GACb,EAAE;IACD,gBAAgB,EAAE,MAAM,EAAE,CAAC;IAC3B,OAAO,EAAE,MAAM,CAAC;IAChB,YAAY,EAAE,MAAM,CAAC;CACtB,GAAG,OAAO,CAAC,MAAM,CAAC,CAQlB;AAED,wBAAsB,gBAAgB,CAAC,EACrC,YAAY,EACZ,cAAc,EACd,YAAY,EACZ,gBAAgB,EAChB,WAAW,EACX,YAAY,EACZ,OAAO,EACP,OAAO,EACP,yBAAyB,EACzB,YAAY,GACb,EAAE,iBAAiB,GAAG,OAAO,CAAC;IAC7B,OAAO,EAAE,OAAO,CAAC;IACjB,KAAK,EAAE,MAAM,CAAC;IACd,MAAM,EAAE,kBAAkB,GAAG,SAAS,CAAC;IACvC,KAAK,EAAE,KAAK,GAAG,SAAS,CAAC;CAC1B,CAAC,CA+FD"}
+{"version":3,"file":"run.d.ts","sourceRoot":"","sources":["../../../src/agent/browsing/run.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,MAAM,mBAAmB,CAAC;AACvD,OAAO,EAAE,KAAK,EAAE,MAAM,uCAAuC,CAAC;AAc9D,KAAK,iBAAiB,GAAG;IACvB,YAAY,EAAE,MAAM,CAAC;IACrB,cAAc,EAAE,MAAM,EAAE,CAAC;IACzB,YAAY,EAAE,MAAM,CAAC;IACrB,gBAAgB,EAAE,MAAM,CAAC;IACzB,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,CAAC;IACrB,OAAO,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,yBAAyB,EAAE,OAAO,CAAC;IACnC,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CACtC,CAAC;AAEF,wBAAsB,8BAA8B,CAAC,EACnD,gBAAgB,EAChB,OAAO,EACP,YAAY,GACb,EAAE;IACD,gBAAgB,EAAE,MAAM,EAAE,CAAC;IAC3B,OAAO,EAAE,MAAM,CAAC;IAChB,YAAY,EAAE,MAAM,CAAC;CACtB,GAAG,OAAO,CAAC,MAAM,CAAC,CAQlB;AAED,wBAAsB,gBAAgB,CAAC,EACrC,YAAY,EACZ,cAAc,EACd,YAAY,EACZ,gBAAgB,EAChB,WAAW,EACX,YAAY,EACZ,OAAO,EACP,OAAO,EACP,yBAAyB,EACzB,YAAY,GACb,EAAE,iBAAiB,GAAG,OAAO,CAAC;IAC7B,OAAO,EAAE,OAAO,CAAC;IACjB,KAAK,EAAE,MAAM,CAAC;IACd,MAAM,EAAE,kBAAkB,GAAG,SAAS,CAAC;IACvC,KAAK,EAAE,KAAK,GAAG,SAAS,CAAC;CAC1B,CAAC,CA+FD"}

package/dist/agent/browsing/utils.d.ts

@@ -1,21 +1,14 @@
 import { Page } from "playwright";
 import { PlaywrightTestConfig } from "playwright/test";
 export declare function isRegExp(obj: any): obj is RegExp;
-export declare function prepareBrowsingAgentTask(steps: string[]): string;
 export declare function addImportForMethod(testFilePath: string, methodName: string): void;
 export declare function replaceTodoWithCreateTest(testFilePath: string, repoDir: string): string | undefined;
-export declare function markTestAsOnly({ testCaseName, testCaseSuites, specPath, }: {
-    testCaseName: string;
-    testCaseSuites: string[];
-    specPath: string;
-}): Promise<void>;
 export declare function injectPwLocatorGenerator(page: Page): Promise<void>;
 /**
  * function to read playwright config from the source repo
  * @return {*}  {Promise<PlaywrightTestConfig>}
  */
 export declare function readPlaywrightConfig(repoDir: string): Promise<PlaywrightTestConfig>;
-export declare function getValidProjectNames(playwrightConfig: PlaywrightTestConfig): Promise<string[]>;
 /**
  * detect the project name for the given file in playwright test repo
  * if project and test file path for running test don't match, then playwright throws error

package/dist/agent/browsing/utils.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../../../src/agent/browsing/utils.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAClC,OAAO,EAAE,oBAAoB,EAAE,MAAM,iBAAiB,CAAC;AAYvD,wBAAgB,QAAQ,CAAC,GAAG,EAAE,GAAG,GAAG,GAAG,IAAI,MAAM,CAKhD;AAED,wBAAgB,wBAAwB,CAAC,KAAK,EAAE,MAAM,EAAE,UAIvD;AAED,wBAAgB,kBAAkB,CAAC,YAAY,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,QAe1E;AAED,wBAAgB,yBAAyB,CACvC,YAAY,EAAE,MAAM,EACpB,OAAO,EAAE,MAAM,GACd,MAAM,GAAG,SAAS,CAsBpB;AAED,wBAAsB,cAAc,CAAC,EACnC,YAAY,EACZ,cAAc,EACd,QAAQ,GACT,EAAE;IACD,YAAY,EAAE,MAAM,CAAC;IACrB,cAAc,EAAE,MAAM,EAAE,CAAC;IACzB,QAAQ,EAAE,MAAM,CAAC;CAClB,iBAoBA;AAyBD,wBAAsB,wBAAwB,CAAC,IAAI,EAAE,IAAI,iBA6HxD;AAED;;;GAGG;AACH,wBAAsB,oBAAoB,CACxC,OAAO,EAAE,MAAM,GACd,OAAO,CAAC,oBAAoB,CAAC,CA2B/B;AAWD,wBAAsB,oBAAoB,CACxC,gBAAgB,EAAE,oBAAoB,GACrC,OAAO,CAAC,MAAM,EAAE,CAAC,CAQnB;AAED;;;;;GAKG;AACH,wBAAsB,iBAAiB,CACrC,YAAY,EAAE,MAAM,EACpB,gBAAgB,EAAE,oBAAoB,EACtC,gBAAgB,GAAE,MAAM,EAAU,GACjC,OAAO,CAAC,MAAM,CAAC,CA+CjB"}
+{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../../../src/agent/browsing/utils.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAClC,OAAO,EAAE,oBAAoB,EAAE,MAAM,iBAAiB,CAAC;AAOvD,wBAAgB,QAAQ,CAAC,GAAG,EAAE,GAAG,GAAG,GAAG,IAAI,MAAM,CAKhD;AAED,wBAAgB,kBAAkB,CAAC,YAAY,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,QAe1E;AAED,wBAAgB,yBAAyB,CACvC,YAAY,EAAE,MAAM,EACpB,OAAO,EAAE,MAAM,GACd,MAAM,GAAG,SAAS,CAsBpB;AAED,wBAAsB,wBAAwB,CAAC,IAAI,EAAE,IAAI,iBA0GxD;AAED;;;GAGG;AACH,wBAAsB,oBAAoB,CACxC,OAAO,EAAE,MAAM,GACd,OAAO,CAAC,oBAAoB,CAAC,CA2B/B;AAWD;;;;;GAKG;AACH,wBAAsB,iBAAiB,CACrC,YAAY,EAAE,MAAM,EACpB,gBAAgB,EAAE,oBAAoB,EACtC,gBAAgB,GAAE,MAAM,EAAU,GACjC,OAAO,CAAC,MAAM,CAAC,CA+CjB"}

package/dist/agent/browsing/utils.js

@@ -4,13 +4,10 @@ var __importDefault = (this && this.__importDefault) || function (mod) {
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.isRegExp = isRegExp;
-exports.prepareBrowsingAgentTask = prepareBrowsingAgentTask;
 exports.addImportForMethod = addImportForMethod;
 exports.replaceTodoWithCreateTest = replaceTodoWithCreateTest;
-exports.markTestAsOnly = markTestAsOnly;
 exports.injectPwLocatorGenerator = injectPwLocatorGenerator;
 exports.readPlaywrightConfig = readPlaywrightConfig;
-exports.getValidProjectNames = getValidProjectNames;
 exports.detectProjectName = detectProjectName;
 const fs_1 = __importDefault(require("fs"));
 const minimatch_1 = require("minimatch");
@@ -22,11 +19,6 @@ function isRegExp(obj) {
     return (obj instanceof RegExp ||
         Object.prototype.toString.call(obj) === "[object RegExp]");
 }
-function prepareBrowsingAgentTask(steps) {
-    const sanitizedSteps = steps.map((step) => step.replace(/`/g, "\\`"));
-    const task = `${sanitizedSteps.join("\n")}\n`;
-    return task;
-}
 function addImportForMethod(testFilePath, methodName) {
     // Instead of using "@empiricalrun/test-gen", we use the local dist file
     // This is to avoid assuming that the test-gen package is installed in the project
@@ -50,50 +42,12 @@ function replaceTodoWithCreateTest(testFilePath, repoDir) {
     addImportForMethod(absoluteTestFilePath, "createTest");
     return todoContent;
 }
-async function markTestAsOnly({ testCaseName, testCaseSuites, specPath, }) {
-    const testFileContent = fs_1.default.readFileSync(specPath, "utf-8");
-    const { testBlock, testNode } = (0, web_1.getTypescriptTestBlock)({
-        scenarioName: testCaseName,
-        content: testFileContent,
-        suites: testCaseSuites,
-    });
-    const parentDescribe = (0, web_1.findFirstSerialDescribeBlock)(testNode);
-    const isFileMarkedSerial = await (0, web_1.hasTopLevelDescribeConfigureWithSerialMode)(specPath);
-    // add test.only / describe.only to the spec file so that only that block is executed
-    if (!isFileMarkedSerial) {
-        const updatedTestFileContent = newContentsWithTestOnly(testFileContent, testBlock, testBlock, parentDescribe?.getText() || "");
-        fs_1.default.writeFileSync(specPath, updatedTestFileContent);
-    }
-}
-function newContentsWithTestOnly(existingContents, originalTestBlock, updatedTestBlock, parentDescribeBlock) {
-    if (!parentDescribeBlock) {
-        const testMarkedAsOnly = updatedTestBlock.replace("test(", "test.only(");
-        return existingContents.replace(originalTestBlock, testMarkedAsOnly);
-    }
-    else {
-        const updatedDescribeBlock = parentDescribeBlock.replace(originalTestBlock, updatedTestBlock);
-        // TODO: this should only happen when the describe block has "serial" execution
-        const describeMarkedAsOnly = updatedDescribeBlock.replace("test.describe(", "test.describe.only(");
-        return existingContents.replace(parentDescribeBlock, describeMarkedAsOnly);
-    }
-}
 async function injectPwLocatorGenerator(page) {
-    let pathToInstalledTestGen = require.resolve(".").split("dist")[0];
-    if (process.env.RUNNING_BROWSER_TESTS_FOR_TEST_GEN) {
-        pathToInstalledTestGen = process.cwd();
-    }
-    const annotateElementPath = path_1.default.join(pathToInstalledTestGen, "dist", "browser-injected-scripts", "annotate-elements.js");
-    if (!fs_1.default.existsSync(annotateElementPath)) {
-        throw new Error(`annotate-elements.js not found at path: ${annotateElementPath}`);
-    }
     const remoteScriptResponses = await Promise.all([
         "https://assets-test.empirical.run/pw-selector.js",
         "https://code.jquery.com/jquery-3.7.1.min.js",
     ].map((url) => fetch(url)));
-    const scripts = await Promise.all([
-        ...remoteScriptResponses.map((r) => r.text()),
-        fs_1.default.readFileSync(annotateElementPath, "utf-8"),
-    ]);
+    const scripts = await Promise.all(remoteScriptResponses.map((r) => r.text()));
     page.on("load", async () => {
         try {
             await Promise.all(scripts.map((s) => page.addScriptTag({ content: s })));
@@ -219,15 +173,6 @@ function matchAgainstPattern(pattern, filePathToTest) {
         return (0, minimatch_1.minimatch)(filePathToTest, pattern);
     }
 }
-async function getValidProjectNames(playwrightConfig) {
-    if (!playwrightConfig.projects) {
-        return [];
-    }
-    const filteredProjectNames = playwrightConfig.projects
-        .map((p) => p.name)
-        .filter((p) => !!p);
-    return filteredProjectNames;
-}
 /**
  * detect the project name for the given file in playwright test repo
  * if project and test file path for running test don't match, then playwright throws error

package/dist/agent/chat/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/agent/chat/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uCAAuC,CAAC;AACzE,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,qCAAqC,CAAC;AAapE,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAKpC,qBAAa,SAAS,CAAC,CAAC,CAAE,SAAQ,SAAS,CAAC,CAAC,CAAC;IAC5C,SAAS,CAAC,QAAQ,IAAI,WAAW;IAiB3B,iBAAiB,CACrB,eAAe,EAAE,MAAM,OAAO,CAAC,QAAQ,CAAC,GACvC,OAAO,CAAC,MAAM,CAAC;CAsGnB"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/agent/chat/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uCAAuC,CAAC;AACzE,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,qCAAqC,CAAC;AAsBpE,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAMpC,qBAAa,SAAS,CAAC,CAAC,CAAE,SAAQ,SAAS,CAAC,CAAC,CAAC;IAC5C,SAAS,CAAC,QAAQ,IAAI,WAAW;IA0B3B,iBAAiB,CACrB,eAAe,EAAE,MAAM,OAAO,CAAC,QAAQ,CAAC,GACvC,OAAO,CAAC,MAAM,CAAC;CAyGnB"}

package/dist/agent/chat/index.js

@@ -3,32 +3,49 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.ChatAgent = void 0;
 const tools_1 = require("../../tools");
 const analyse_video_1 = require("../../tools/definitions/analyse-video");
+const download_build_1 = require("../../tools/definitions/download-build");
+const grep_1 = require("../../tools/definitions/grep");
+const list_tests_and_projects_1 = require("../../tools/definitions/list-tests-and-projects");
+const run_test_1 = require("../../tools/definitions/run-test");
 const diagnosis_fetcher_1 = require("../../tools/diagnosis-fetcher");
+const fetch_file_1 = require("../../tools/fetch-file");
+const list_environments_1 = require("../../tools/list-environments");
 const test_run_fetcher_1 = require("../../tools/test-run-fetcher");
 const base_1 = require("../base");
 const pw_utils_docs_1 = require("./prompt/pw-utils-docs");
 const repo_1 = require("./prompt/repo");
 const test_case_def_1 = require("./prompt/test-case-def");
+const trace_utils_docs_1 = require("./prompt/trace-utils-docs");
 class ChatAgent extends base_1.BaseAgent {
     getTools() {
         const custom = [
             analyse_video_1.analyseVideo,
-            ...tools_1.commonTools,
+            run_test_1.runTestTool,
+            grep_1.grepTool,
+            list_environments_1.listEnvironmentsTool,
+            download_build_1.downloadBuildTool,
+            fetch_file_1.fetchFileTool,
+            list_tests_and_projects_1.listProjectsTool,
+            list_tests_and_projects_1.listTestsForProjectTool,
             ...tools_1.testGenerationTools,
             test_run_fetcher_1.fetchTestRunDetailsTool,
             diagnosis_fetcher_1.fetchDiagnosisReportTool,
             ...(0, tools_1.textEditorToolsForModel)(this.selectedModel),
             tools_1.safeBashTool,
-            ...(this.featureFlags.includes("useScrapeHtml") ? [tools_1.scrapeHtmlTool] : []),
+            ...(!(0, tools_1.hasBuiltInWebFetch)(this.selectedModel) ? [tools_1.scrapeHtmlTool] : []),
         ];
         return {
             custom,
             builtInTextEditor: (0, tools_1.hasBuiltInTextEditor)(this.selectedModel),
+            builtInWebFetch: (0, tools_1.hasBuiltInWebFetch)(this.selectedModel)
+                ? { enabled: true, max_uses: 1 }
+                : undefined,
         };
     }
     async buildSystemPrompt(repoInfoBuilder) {
         const repoInfo = await repoInfoBuilder();
         const repoContext = await (0, repo_1.getRepoContextPrompt)(repoInfo);
+        const hasTraceUtils = this.featureFlags.includes("enableTraceUtilsViaBash");
         const preamble = `
 You are a helpful assistant that can answer questions and help with tasks related to writing and maintaining Playwright tests.
 
@@ -121,6 +138,8 @@ ${test_case_def_1.testCasesDefinitionPrompt}
 # Recipes
 ${pw_utils_docs_1.playwrightUtilsDocs}
 
+${trace_utils_docs_1.traceUtilsDoc}
+
 # Repo context
 ${repoContext}
 

package/dist/agent/chat/prompt/pw-utils-docs.d.ts

@@ -1,2 +1,2 @@
-export declare const playwrightUtilsDocs = "\nYou can refer to the following recipes to learn how to write tests for different scenarios.\n\n<email-automation>\n\n# Email automation\n\n## Example usage\n\n### Dynamic email\n\nThis dynamically generates a random email address that can \nbe used for the test (e.g. invite a new user).\n\n```ts\nimport { EmailClient } from \"@empiricalrun/playwright-utils\";\nimport { expect } from \"@playwright/test\";\n\nconst client = new EmailClient();\nconst address = client.getAddress();\n\n// Input the `address` in the application\n// that sends the email.\n\n// Get email received on the `address`\nconst email = await client.waitForEmail();\nexpect(\n  email.links.find((l) => l.text === \"Join your team\")\n).toBeTruthy();\n```\n\n### Static email\n\nThis uses a known (static) email that can be used to login\ninto an application.\n\nThis needs an email id (e.g. `test-login-user`). The email id\nis appended with the domain (managed internally) to get the full\nemail address.\n\n```ts\nimport { EmailClient } from \"@empiricalrun/playwright-utils\";\n\nconst emailId = `test-login-user`;\n\nconst client = new EmailClient({ emailId });\nconst address = client.getAddress(); // Returns full address with domain\n\n// Get email received on the `address`\nconst email = await client.waitForEmail();\n\n// Get login OTP\nconst loginCode = email.codes[0];\n```\n\n</email-automation>\n\n<fixtures>\n\nThe playwright-utils package provides fixtures that wrap around Playwright's built-in\n`page`, `context` fixtures to provide a mouse highlighter (which makes it easier to\nsee actions taken in a video).\n\nTo use this, you can use the `baseTestFixture` and `extendExpect` imports\nin your fixtures file.\n\n```ts\nimport { test as base, expect as baseExpect } from \"@playwright/test\";\nimport { baseTestFixture, extendExpect } from \"@empiricalrun/playwright-utils/test\";\n\nexport const test = baseTestFixture(base);\nexport const expect = extendExpect(baseExpect);\n```\n\n### Get a new browser context\n\nThis package provides a fixture `customContextPageProvider` which is a good way to create\na fresh, new browser context, and a page inside it.\n\nThere are two benefits of using this to create contexts or pages:\n1. Videos get recorded and attached to the test report\n2. Mouse highlights are available\n\n```ts\nimport { test, expect } from \"./fixtures\";\n\ntest(\"Example test\", async ({ page: builtInPage, customContextPageProvider }) => {\n  // builtInPage is from default browser context\n  const { page: newPage, context } = await customContextPageProvider();\n  // newPage is from this other browser context\n});\n```\n\ncustomContextPageProvider can accept options to customize the browser context.\n\n```\n  customContextPageProvider: (\n    options?: BrowserContextOptions,\n  ) => Promise<{ context: BrowserContext; page: Page }>;\n```\n\nFor example, pass { storageState: undefined } to create a new browser context without\nthe auth state of the current browser context. This is useful for multi-user scenarios.\n\n</fixtures>\n\n<video-labels>\n\n# Video Labels\n\nPages generate video recordings after test execution, with 1 page generating 1 video file (webm). \n\nIf your test case relies on multiple pages (e.g. for multi-user or multi-app flows), it can get difficult to\nknow which page does \"video-1.webm\" belong to.\n\nTo solve this, you should set video labels for pages. This will enable you to identify videos faster.\n\n## Usage\n\n```typescript\nimport { setVideoLabel } from '@empiricalrun/playwright-utils/test';\n\ntest('my test', async ({ page }) => {\n  setVideoLabel(page, 'checkout-flow');\n  // Video will be saved as 'checkout-flow.webm'\n});\n```\n\n## Multiple Contexts\n\n```typescript\ntest('multi-user scenario', async ({ page, customContextPageProvider }) => {\n  setVideoLabel(page, 'host-page');\n\n  const { page: guestPage } = await customContextPageProvider({ storageState: undefined });\n  setVideoLabel(guestPage, 'guest-page');\n  // Videos saved as 'guest-page.webm' and 'host-page.webm'\n});\n```\n\n## Notes\n\n- The default behavior is to label videos for multiple pages as: `video-0.webm`, `video-1.webm`, etc.\n- If setVideoLabel is called twice for the same page, the last label will be set\n\n</video-labels>\n\n";
+export declare const playwrightUtilsDocs = "\nYou can refer to the following recipes to learn how to write or diagnose tests for different scenarios.\n\n<email-automation>\n\n# Email automation\n\n## Example usage\n\n### Dynamic email\n\nThis dynamically generates a random email address that can \nbe used for the test (e.g. invite a new user).\n\n```ts\nimport { EmailClient } from \"@empiricalrun/playwright-utils\";\nimport { expect } from \"@playwright/test\";\n\nconst client = new EmailClient();\nconst address = client.getAddress();\n\n// Input the `address` in the application\n// that sends the email.\n\n// Get email received on the `address`\nconst email = await client.waitForEmail();\nexpect(\n  email.links.find((l) => l.text === \"Join your team\")\n).toBeTruthy();\n```\n\n### Static email\n\nThis uses a known (static) email that can be used to login\ninto an application.\n\nThis needs an email id (e.g. `test-login-user`). The email id\nis appended with the domain (managed internally) to get the full\nemail address.\n\n```ts\nimport { EmailClient } from \"@empiricalrun/playwright-utils\";\n\nconst emailId = `test-login-user`;\n\nconst client = new EmailClient({ emailId });\nconst address = client.getAddress(); // Returns full address with domain\n\n// Get email received on the `address`\nconst email = await client.waitForEmail();\n\n// Get login OTP\nconst loginCode = email.codes[0];\n```\n\n</email-automation>\n\n<fixtures>\n\nThe playwright-utils package provides fixtures that wrap around Playwright's built-in\n`page`, `context` fixtures to provide a mouse highlighter (which makes it easier to\nsee actions taken in a video).\n\nTo use this, you can use the `baseTestFixture` and `extendExpect` imports\nin your fixtures file.\n\n```ts\nimport { test as base, expect as baseExpect } from \"@playwright/test\";\nimport { baseTestFixture, extendExpect } from \"@empiricalrun/playwright-utils/test\";\n\nexport const test = baseTestFixture(base);\nexport const expect = extendExpect(baseExpect);\n```\n\n### Get a new browser context\n\nThis package provides a fixture `customContextPageProvider` which is a good way to create\na fresh, new browser context, and a page inside it.\n\nThere are two benefits of using this to create contexts or pages:\n1. Videos get recorded and attached to the test report\n2. Mouse highlights are available\n\n```ts\nimport { test, expect } from \"./fixtures\";\n\ntest(\"Example test\", async ({ page: builtInPage, customContextPageProvider }) => {\n  // builtInPage is from default browser context\n  const { page: newPage, context } = await customContextPageProvider();\n  // newPage is from this other browser context\n});\n```\n\ncustomContextPageProvider can accept options to customize the browser context.\n\n```\n  customContextPageProvider: (\n    options?: BrowserContextOptions,\n  ) => Promise<{ context: BrowserContext; page: Page }>;\n```\n\nFor example, pass { storageState: undefined } to create a new browser context without\nthe auth state of the current browser context. This is useful for multi-user scenarios.\n\n</fixtures>\n\n<video-labels>\n\n# Video Labels\n\nPages generate video recordings after test execution, with 1 page generating 1 video file (webm). \n\nIf your test case relies on multiple pages (e.g. for multi-user or multi-app flows), it can get difficult to\nknow which page does \"video-1.webm\" belong to.\n\nTo solve this, you should set video labels for pages. This will enable you to identify videos faster.\n\n## Usage\n\n```typescript\nimport { setVideoLabel } from '@empiricalrun/playwright-utils/test';\n\ntest('my test', async ({ page }) => {\n  setVideoLabel(page, 'checkout-flow');\n  // Video will be saved as 'checkout-flow.webm'\n});\n```\n\n## Multiple Contexts\n\n```typescript\ntest('multi-user scenario', async ({ page, customContextPageProvider }) => {\n  setVideoLabel(page, 'host-page');\n\n  const { page: guestPage } = await customContextPageProvider({ storageState: undefined });\n  setVideoLabel(guestPage, 'guest-page');\n  // Videos saved as 'guest-page.webm' and 'host-page.webm'\n});\n```\n\n## Notes\n\n- The default behavior is to label videos for multiple pages as: `video-0.webm`, `video-1.webm`, etc.\n- If setVideoLabel is called twice for the same page, the last label will be set\n\n</video-labels>\n\n";
 //# sourceMappingURL=pw-utils-docs.d.ts.map

package/dist/agent/chat/prompt/pw-utils-docs.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"pw-utils-docs.d.ts","sourceRoot":"","sources":["../../../../src/agent/chat/prompt/pw-utils-docs.ts"],"names":[],"mappings":"AA4IA,eAAO,MAAM,mBAAmB,qsIAe/B,CAAC"}
+{"version":3,"file":"pw-utils-docs.d.ts","sourceRoot":"","sources":["../../../../src/agent/chat/prompt/pw-utils-docs.ts"],"names":[],"mappings":"AA4IA,eAAO,MAAM,mBAAmB,itIAe/B,CAAC"}

package/dist/agent/chat/prompt/pw-utils-docs.js

@@ -139,7 +139,7 @@ test('multi-user scenario', async ({ page, customContextPageProvider }) => {
 - If setVideoLabel is called twice for the same page, the last label will be set
 `;
 exports.playwrightUtilsDocs = `
-You can refer to the following recipes to learn how to write tests for different scenarios.
+You can refer to the following recipes to learn how to write or diagnose tests for different scenarios.
 
 <email-automation>
 ${emailRecipe}

package/dist/agent/chat/prompt/trace-utils-docs.d.ts

@@ -0,0 +1,2 @@
+export declare const traceUtilsDoc = "\n<trace-utils-cli>\n# Trace Utils\n\nCLI and library for inspecting Playwright `trace.zip` files.\n\n## Usage\n\nVia sandbox-cli:\n```bash\nsandbox-cli trace-utils <command> [options]\n```\n\n---\n\n## `steps`\n\nList all test steps from a trace.\n\n```bash\nsandbox-cli trace-utils steps --file <path> [--json]\n```\n\n| Option | Description | Default |\n|--------|-------------|---------|\n| `--file` | Path or URL to trace.zip | *required* |\n| `--json` | Output as JSON | `false` |\n\n---\n\n## `screenshots`\n\nExtract screenshots from a trace. Use either `--step` or `--start`/`--end` (one is required).\n\n```bash\n# By step ID\nsandbox-cli trace-utils screenshots --file <path> --step <id> [options]\n\n# By time range\nsandbox-cli trace-utils screenshots --file <path> --start <time> --end <time> [options]\n```\n\n| Option | Description | Default |\n|--------|-------------|---------|\n| `--file` | Path or URL to trace.zip | *required* |\n| `--step` | Step callId to extract around | - |\n| `--start` | Start time (HH:MM:SS.mmm) | - |\n| `--end` | End time (HH:MM:SS.mmm) | - |\n| `--page` | Page ID to filter by (required for multi-page traces with time-based extraction) | - |\n| `--before` | Milliseconds before step (only with --step) | `2000` |\n| `--after` | Milliseconds after step (only with --step) | `2000` |\n| `--max` | Maximum screenshots | `10` |\n| `--output` | Output directory | auto-generated in `/tmp` |\n\n**Examples:**\n\n```bash\n# 1. List steps to find callId or timestamps\nsandbox-cli trace-utils steps --file ./trace.zip\n\n# Output:\n# [test.step@36] locator.click(getByRole('button', { name: 'Create' })) (page@abc123)\n#   00:02:06.434 \u2192 00:02:06.543 (110ms)\n\n# 2a. Extract by step ID\nsandbox-cli trace-utils screenshots --file ./trace.zip --step \"test.step@36\"\n\n# 2b. Extract by time range (for multi-page traces, --page is required)\nsandbox-cli trace-utils screenshots --file ./trace.zip --start 00:02:06.000 --end 00:02:07.000 --page \"page@abc123\"\n```\n\n---\n\n## `network`\n\nSearch network requests. For multi-page traces, either `--step` or `--page` is required.\n\n```bash\nsandbox-cli trace-utils network --file <path> [filters] [options]\n```\n\n| Option | Description | Default |\n|--------|-------------|---------|\n| `--file` | Path or URL to trace.zip | *required* |\n| `--step <callId>` | Filter to requests during a specific step | - |\n| `--page` | Page ID to filter by (required for multi-page traces without --step) | - |\n| `--before <ms>` | Time before step to include (with --step) | `2000` |\n| `--after <ms>` | Time after step to include (with --step) | `2000` |\n| `--errors` | Show only failed requests (4xx, 5xx, no response) | - |\n| `--url <pattern>` | Filter by URL pattern (regex) | - |\n| `--method <method>` | Filter by HTTP method | - |\n| `--status <codes>` | Filter by status codes (e.g., `404,500`) | - |\n| `--json` | Output as JSON | `false` |\n\n**Examples:**\n\n```bash\n# Show all failed requests (for multi-page traces, specify --page)\nsandbox-cli trace-utils network --file ./trace.zip --errors --page \"page@abc123\"\n\n# Search for specific API calls\nsandbox-cli trace-utils network --file ./trace.zip --url \"/api/v1/users\" --page \"page@abc123\"\n\n# Failed POST requests\nsandbox-cli trace-utils network --file ./trace.zip --errors --method POST --page \"page@abc123\"\n\n# Network requests during a specific step (--page not required)\nsandbox-cli trace-utils network --file ./trace.zip --step \"pw:api@32\"\n```\n\n---\n\n## `console`\n\nView browser console output. For multi-page traces, `--page` is required.\n\n```bash\nsandbox-cli trace-utils console --file <path> [options]\n```\n\n| Option | Description | Default |\n|--------|-------------|---------|\n| `--file` | Path or URL to trace.zip | *required* |\n| `--page` | Page ID to filter by (required for multi-page traces) | - |\n| `--level` | Minimum level: `error`, `warning`, `info`, `debug` | `error` |\n| `--json` | Output as JSON | `false` |\n| `--search` | Filter logs containing text | - |\n\n---\n\n## `dom-snapshot`\n\nExtract DOM snapshot from a trace. Use either `--step` or `--time` (one is required).\n\n```bash\n# By step ID\nsandbox-cli trace-utils dom-snapshot --file <path> --step <id> [options]\n\n# By time (for multi-page traces, --page is required)\nsandbox-cli trace-utils dom-snapshot --file <path> --time <time> --page <pageId> [options]\n```\n\n| Option | Description | Default |\n|--------|-------------|---------|\n| `--file` | Path or URL to trace.zip | *required* |\n| `--step` | Step callId to extract snapshot for | - |\n| `--time` | Time to extract snapshot at (HH:MM:SS.mmm) | - |\n| `--page` | Page ID to filter by (required for multi-page traces with --time) | - |\n| `--type` | Snapshot type: `before`, `after`, `input` (only with --step) | `before` |\n| `--output` | Output file path | - |\n| `--json` | Output as JSON | `false` |\n\n**Snapshot types:**\n\n- `before` - DOM state when the step starts (default)\n- `after` - DOM state when the step completes\n- `input` - DOM state at the moment of user input (for click/fill actions)\n\n> **Tip for assertions:** For `expect` steps like `toBeVisible`, use `--type after` to see the DOM state when the assertion passed. The `before` snapshot may show a loading state if the test was waiting for an element to appear.\n\n**Examples:**\n\n```bash\n# 1. List steps to find callId\nsandbox-cli trace-utils steps --file ./trace.zip\n\n# 2. Extract DOM snapshot before a click step\nsandbox-cli trace-utils dom-snapshot --file ./trace.zip --step \"pw:api@32\" --type before\n\n# 3. Extract DOM snapshot after a step and save to file\nsandbox-cli trace-utils dom-snapshot --file ./trace.zip --step \"pw:api@32\" --type after --output ./snapshot.html\n\n# 4. For assertions, use --type after to see the DOM when assertion passed\nsandbox-cli trace-utils dom-snapshot --file ./trace.zip --step \"expect@43\" --type after\n\n# 5. Extract DOM snapshot at a specific time\nsandbox-cli trace-utils dom-snapshot --file ./trace.zip --time 00:02:06.434\n\n# 6. Get snapshot as JSON\nsandbox-cli trace-utils dom-snapshot --file ./trace.zip --step \"pw:api@32\" --json\n```\n\n**Output (default):**\n```\n# DOM Snapshot: locator.click\n# Step: pw:api@32\n# URL: https://example.com/dashboard\n# Viewport: 1280x720\n# Snapshot type: before\n\n<!DOCTYPE html>\n<html>\n  <head>\n    <title>Dashboard</title>\n    ...\n  </head>\n  <body>\n    ...\n  </body>\n</html>\n```\n\n</trace-utils-cli>\n";
+//# sourceMappingURL=trace-utils-docs.d.ts.map

package/dist/agent/chat/prompt/trace-utils-docs.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"trace-utils-docs.d.ts","sourceRoot":"","sources":["../../../../src/agent/chat/prompt/trace-utils-docs.ts"],"names":[],"mappings":"AACA,eAAO,MAAM,aAAa,w4MA2MzB,CAAC"}

package/dist/agent/chat/prompt/trace-utils-docs.js

@@ -0,0 +1,208 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.traceUtilsDoc = void 0;
+// TODO: Need a better way to keep this in sync with trace-utils/README.md
+exports.traceUtilsDoc = `
+<trace-utils-cli>
+# Trace Utils
+
+CLI and library for inspecting Playwright \`trace.zip\` files.
+
+## Usage
+
+Via sandbox-cli:
+\`\`\`bash
+sandbox-cli trace-utils <command> [options]
+\`\`\`
+
+---
+
+## \`steps\`
+
+List all test steps from a trace.
+
+\`\`\`bash
+sandbox-cli trace-utils steps --file <path> [--json]
+\`\`\`
+
+| Option | Description | Default |
+|--------|-------------|---------|
+| \`--file\` | Path or URL to trace.zip | *required* |
+| \`--json\` | Output as JSON | \`false\` |
+
+---
+
+## \`screenshots\`
+
+Extract screenshots from a trace. Use either \`--step\` or \`--start\`/\`--end\` (one is required).
+
+\`\`\`bash
+# By step ID
+sandbox-cli trace-utils screenshots --file <path> --step <id> [options]
+
+# By time range
+sandbox-cli trace-utils screenshots --file <path> --start <time> --end <time> [options]
+\`\`\`
+
+| Option | Description | Default |
+|--------|-------------|---------|
+| \`--file\` | Path or URL to trace.zip | *required* |
+| \`--step\` | Step callId to extract around | - |
+| \`--start\` | Start time (HH:MM:SS.mmm) | - |
+| \`--end\` | End time (HH:MM:SS.mmm) | - |
+| \`--page\` | Page ID to filter by (required for multi-page traces with time-based extraction) | - |
+| \`--before\` | Milliseconds before step (only with --step) | \`2000\` |
+| \`--after\` | Milliseconds after step (only with --step) | \`2000\` |
+| \`--max\` | Maximum screenshots | \`10\` |
+| \`--output\` | Output directory | auto-generated in \`/tmp\` |
+
+**Examples:**
+
+\`\`\`bash
+# 1. List steps to find callId or timestamps
+sandbox-cli trace-utils steps --file ./trace.zip
+
+# Output:
+# [test.step@36] locator.click(getByRole('button', { name: 'Create' })) (page@abc123)
+#   00:02:06.434 → 00:02:06.543 (110ms)
+
+# 2a. Extract by step ID
+sandbox-cli trace-utils screenshots --file ./trace.zip --step "test.step@36"
+
+# 2b. Extract by time range (for multi-page traces, --page is required)
+sandbox-cli trace-utils screenshots --file ./trace.zip --start 00:02:06.000 --end 00:02:07.000 --page "page@abc123"
+\`\`\`
+
+---
+
+## \`network\`
+
+Search network requests. For multi-page traces, either \`--step\` or \`--page\` is required.
+
+\`\`\`bash
+sandbox-cli trace-utils network --file <path> [filters] [options]
+\`\`\`
+
+| Option | Description | Default |
+|--------|-------------|---------|
+| \`--file\` | Path or URL to trace.zip | *required* |
+| \`--step <callId>\` | Filter to requests during a specific step | - |
+| \`--page\` | Page ID to filter by (required for multi-page traces without --step) | - |
+| \`--before <ms>\` | Time before step to include (with --step) | \`2000\` |
+| \`--after <ms>\` | Time after step to include (with --step) | \`2000\` |
+| \`--errors\` | Show only failed requests (4xx, 5xx, no response) | - |
+| \`--url <pattern>\` | Filter by URL pattern (regex) | - |
+| \`--method <method>\` | Filter by HTTP method | - |
+| \`--status <codes>\` | Filter by status codes (e.g., \`404,500\`) | - |
+| \`--json\` | Output as JSON | \`false\` |
+
+**Examples:**
+
+\`\`\`bash
+# Show all failed requests (for multi-page traces, specify --page)
+sandbox-cli trace-utils network --file ./trace.zip --errors --page "page@abc123"
+
+# Search for specific API calls
+sandbox-cli trace-utils network --file ./trace.zip --url "/api/v1/users" --page "page@abc123"
+
+# Failed POST requests
+sandbox-cli trace-utils network --file ./trace.zip --errors --method POST --page "page@abc123"
+
+# Network requests during a specific step (--page not required)
+sandbox-cli trace-utils network --file ./trace.zip --step "pw:api@32"
+\`\`\`
+
+---
+
+## \`console\`
+
+View browser console output. For multi-page traces, \`--page\` is required.
+
+\`\`\`bash
+sandbox-cli trace-utils console --file <path> [options]
+\`\`\`
+
+| Option | Description | Default |
+|--------|-------------|---------|
+| \`--file\` | Path or URL to trace.zip | *required* |
+| \`--page\` | Page ID to filter by (required for multi-page traces) | - |
+| \`--level\` | Minimum level: \`error\`, \`warning\`, \`info\`, \`debug\` | \`error\` |
+| \`--json\` | Output as JSON | \`false\` |
+| \`--search\` | Filter logs containing text | - |
+
+---
+
+## \`dom-snapshot\`
+
+Extract DOM snapshot from a trace. Use either \`--step\` or \`--time\` (one is required).
+
+\`\`\`bash
+# By step ID
+sandbox-cli trace-utils dom-snapshot --file <path> --step <id> [options]
+
+# By time (for multi-page traces, --page is required)
+sandbox-cli trace-utils dom-snapshot --file <path> --time <time> --page <pageId> [options]
+\`\`\`
+
+| Option | Description | Default |
+|--------|-------------|---------|
+| \`--file\` | Path or URL to trace.zip | *required* |
+| \`--step\` | Step callId to extract snapshot for | - |
+| \`--time\` | Time to extract snapshot at (HH:MM:SS.mmm) | - |
+| \`--page\` | Page ID to filter by (required for multi-page traces with --time) | - |
+| \`--type\` | Snapshot type: \`before\`, \`after\`, \`input\` (only with --step) | \`before\` |
+| \`--output\` | Output file path | - |
+| \`--json\` | Output as JSON | \`false\` |
+
+**Snapshot types:**
+
+- \`before\` - DOM state when the step starts (default)
+- \`after\` - DOM state when the step completes
+- \`input\` - DOM state at the moment of user input (for click/fill actions)
+
+> **Tip for assertions:** For \`expect\` steps like \`toBeVisible\`, use \`--type after\` to see the DOM state when the assertion passed. The \`before\` snapshot may show a loading state if the test was waiting for an element to appear.
+
+**Examples:**
+
+\`\`\`bash
+# 1. List steps to find callId
+sandbox-cli trace-utils steps --file ./trace.zip
+
+# 2. Extract DOM snapshot before a click step
+sandbox-cli trace-utils dom-snapshot --file ./trace.zip --step "pw:api@32" --type before
+
+# 3. Extract DOM snapshot after a step and save to file
+sandbox-cli trace-utils dom-snapshot --file ./trace.zip --step "pw:api@32" --type after --output ./snapshot.html
+
+# 4. For assertions, use --type after to see the DOM when assertion passed
+sandbox-cli trace-utils dom-snapshot --file ./trace.zip --step "expect@43" --type after
+
+# 5. Extract DOM snapshot at a specific time
+sandbox-cli trace-utils dom-snapshot --file ./trace.zip --time 00:02:06.434
+
+# 6. Get snapshot as JSON
+sandbox-cli trace-utils dom-snapshot --file ./trace.zip --step "pw:api@32" --json
+\`\`\`
+
+**Output (default):**
+\`\`\`
+# DOM Snapshot: locator.click
+# Step: pw:api@32
+# URL: https://example.com/dashboard
+# Viewport: 1280x720
+# Snapshot type: before
+
+<!DOCTYPE html>
+<html>
+  <head>
+    <title>Dashboard</title>
+    ...
+  </head>
+  <body>
+    ...
+  </body>
+</html>
+\`\`\`
+
+</trace-utils-cli>
+`;

package/dist/agent/code-review/index.d.ts

@@ -4,7 +4,7 @@ import { BaseAgent } from "../base";
 import { type CodeReviewResultV0, type CodeReviewResultV1, type CodeReviewResultV2, CodeReviewSeverity, CodeReviewVerdict } from "./types";
 export type { CodeReviewResultV1, CodeReviewResultV0, CodeReviewResultV2 };
 export { CodeReviewVerdict, CodeReviewSeverity };
-export { convertXmlToV2Format } from "./parser";
+export { convertXmlToV2Format } from "./xml-parser";
 export type CodeReviewVersionedResult = CodeReviewResultV0 | CodeReviewResultV1 | CodeReviewResultV2;
 export declare class CodeReviewAgent<T> extends BaseAgent<T> {
     protected getTools(): ToolsForLLM;

package/dist/agent/code-review/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/agent/code-review/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAEV,WAAW,EACZ,MAAM,uCAAuC,CAAC;AAC/C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,qCAAqC,CAAC;AAGpE,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAGpC,OAAO,EACL,KAAK,kBAAkB,EACvB,KAAK,kBAAkB,EACvB,KAAK,kBAAkB,EACvB,kBAAkB,EAClB,iBAAiB,EAClB,MAAM,SAAS,CAAC;AAEjB,YAAY,EAAE,kBAAkB,EAAE,kBAAkB,EAAE,kBAAkB,EAAE,CAAC;AAC3E,OAAO,EAAE,iBAAiB,EAAE,kBAAkB,EAAE,CAAC;AACjD,OAAO,EAAE,oBAAoB,EAAE,MAAM,UAAU,CAAC;AAEhD,MAAM,MAAM,yBAAyB,GACjC,kBAAkB,GAClB,kBAAkB,GAClB,kBAAkB,CAAC;AAEvB,qBAAa,eAAe,CAAC,CAAC,CAAE,SAAQ,SAAS,CAAC,CAAC,CAAC;IAClD,SAAS,CAAC,QAAQ,IAAI,WAAW;IAWjC,SAAS,IAAI,yBAAyB,GAAG,SAAS;cAmBlC,iBAAiB,CAC/B,eAAe,EAAE,MAAM,OAAO,CAAC,QAAQ,CAAC,GACvC,OAAO,CAAC,MAAM,CAAC;CAiInB"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/agent/code-review/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAEV,WAAW,EACZ,MAAM,uCAAuC,CAAC;AAC/C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,qCAAqC,CAAC;AAGpE,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAEpC,OAAO,EACL,KAAK,kBAAkB,EACvB,KAAK,kBAAkB,EACvB,KAAK,kBAAkB,EACvB,kBAAkB,EAClB,iBAAiB,EAClB,MAAM,SAAS,CAAC;AAGjB,YAAY,EAAE,kBAAkB,EAAE,kBAAkB,EAAE,kBAAkB,EAAE,CAAC;AAC3E,OAAO,EAAE,iBAAiB,EAAE,kBAAkB,EAAE,CAAC;AACjD,OAAO,EAAE,oBAAoB,EAAE,MAAM,cAAc,CAAC;AAEpD,MAAM,MAAM,yBAAyB,GACjC,kBAAkB,GAClB,kBAAkB,GAClB,kBAAkB,CAAC;AAEvB,qBAAa,eAAe,CAAC,CAAC,CAAE,SAAQ,SAAS,CAAC,CAAC,CAAC;IAClD,SAAS,CAAC,QAAQ,IAAI,WAAW;IAWjC,SAAS,IAAI,yBAAyB,GAAG,SAAS;cAmBlC,iBAAiB,CAC/B,eAAe,EAAE,MAAM,OAAO,CAAC,QAAQ,CAAC,GACvC,OAAO,CAAC,MAAM,CAAC;CAiInB"}

package/dist/agent/code-review/index.js

@@ -5,12 +5,12 @@ const tools_1 = require("../../tools");
 const fetch_session_diff_1 = require("../../tools/fetch-session-diff");
 const base_1 = require("../base");
 const repo_1 = require("../chat/prompt/repo");
-const parser_1 = require("./parser");
 const types_1 = require("./types");
 Object.defineProperty(exports, "CodeReviewSeverity", { enumerable: true, get: function () { return types_1.CodeReviewSeverity; } });
 Object.defineProperty(exports, "CodeReviewVerdict", { enumerable: true, get: function () { return types_1.CodeReviewVerdict; } });
-var parser_2 = require("./parser");
-Object.defineProperty(exports, "convertXmlToV2Format", { enumerable: true, get: function () { return parser_2.convertXmlToV2Format; } });
+const xml_parser_1 = require("./xml-parser");
+var xml_parser_2 = require("./xml-parser");
+Object.defineProperty(exports, "convertXmlToV2Format", { enumerable: true, get: function () { return xml_parser_2.convertXmlToV2Format; } });
 class CodeReviewAgent extends base_1.BaseAgent {
     getTools() {
         const custom = [
@@ -35,7 +35,7 @@ class CodeReviewAgent extends base_1.BaseAgent {
         if (!text) {
             return undefined;
         }
-        return (0, parser_1.convertXmlToV2Format)(text);
+        return (0, xml_parser_1.convertXmlToV2Format)(text);
     }
     async buildSystemPrompt(repoInfoBuilder) {
         const repoInfo = await repoInfoBuilder();

package/dist/agent/code-review/{parser.d.ts → xml-parser.d.ts}

@@ -2,4 +2,4 @@ import { type CodeReviewResultV2 } from "./types";
 export type { CodeReviewLineComment, CodeReviewResultV0, CodeReviewResultV1, CodeReviewResultV2, } from "./types";
 export { CodeReviewSeverity, CodeReviewVerdict } from "./types";
 export declare function convertXmlToV2Format(output: string): CodeReviewResultV2;
-//# sourceMappingURL=parser.d.ts.map
+//# sourceMappingURL=xml-parser.d.ts.map

package/dist/agent/code-review/xml-parser.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"xml-parser.d.ts","sourceRoot":"","sources":["../../../src/agent/code-review/xml-parser.ts"],"names":[],"mappings":"AAAA,OAAO,EAEL,KAAK,kBAAkB,EAGxB,MAAM,SAAS,CAAC;AAEjB,YAAY,EACV,qBAAqB,EACrB,kBAAkB,EAClB,kBAAkB,EAClB,kBAAkB,GACnB,MAAM,SAAS,CAAC;AAEjB,OAAO,EAAE,kBAAkB,EAAE,iBAAiB,EAAE,MAAM,SAAS,CAAC;AAchE,wBAAgB,oBAAoB,CAAC,MAAM,EAAE,MAAM,GAAG,kBAAkB,CAmFvE"}

package/dist/agent/code-review/{parser.js → xml-parser.js}

@@ -41,7 +41,7 @@ function convertXmlToV2Format(output) {
                             : null;
                     if (severity !== null) {
                         lineComments.push({
-                            file: fileMatch[1].trim(),
+                            file: fileMatch[1].trim().replace(/^\/repo\//, ""),
                             line_start: parseInt(lineStartMatch[1].trim(), 10),
                             line_end: parseInt(lineEndMatch[1].trim(), 10),
                             severity: severity,

package/dist/agent/fast-triage/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/agent/fast-triage/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uCAAuC,CAAC;AACzE,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,qCAAqC,CAAC;AASpE,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAEpC,qBAAa,eAAe,CAAC,CAAC,CAAE,SAAQ,SAAS,CAAC,CAAC,CAAC;IAClD,SAAS,CAAC,QAAQ,IAAI,WAAW;cAiBjB,iBAAiB,CAC/B,eAAe,EAAE,MAAM,OAAO,CAAC,QAAQ,CAAC,GACvC,OAAO,CAAC,MAAM,CAAC;CAqBnB"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/agent/fast-triage/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uCAAuC,CAAC;AACzE,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,qCAAqC,CAAC;AAapE,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAGpC,qBAAa,eAAe,CAAC,CAAC,CAAE,SAAQ,SAAS,CAAC,CAAC,CAAC;IAClD,SAAS,CAAC,QAAQ,IAAI,WAAW;cAkBjB,iBAAiB,CAC/B,eAAe,EAAE,MAAM,OAAO,CAAC,QAAQ,CAAC,GACvC,OAAO,CAAC,MAAM,CAAC;CA+CnB"}

package/dist/agent/fast-triage/index.js

@@ -2,13 +2,14 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.FastTriageAgent = void 0;
 const tools_1 = require("../../tools");
-// import { traceUtils } from "../../tools/definitions/trace-utils";
 // import { grepTool } from "../../tools/definitions/grep";
 const diagnosis_fetcher_1 = require("../../tools/diagnosis-fetcher");
 const fetch_file_1 = require("../../tools/fetch-file");
 const list_environments_1 = require("../../tools/list-environments");
+const publish_markdown_report_1 = require("../../tools/publish-markdown-report");
 const test_run_fetcher_1 = require("../../tools/test-run-fetcher");
 const base_1 = require("../base");
+const trace_utils_docs_1 = require("../chat/prompt/trace-utils-docs");
 class FastTriageAgent extends base_1.BaseAgent {
     getTools() {
         const tools = [
@@ -16,8 +17,9 @@ class FastTriageAgent extends base_1.BaseAgent {
             test_run_fetcher_1.fetchTestRunDetailsTool,
             list_environments_1.listEnvironmentsTool,
             fetch_file_1.fetchFileTool,
-            // TODO: Bring these back - once we have a remote tool executor for them
-            // traceUtils,
+            tools_1.safeBashTool,
+            publish_markdown_report_1.publishMarkdownReportTool,
+            // TODO: Do we need grep?
             // grepTool,
             ...(0, tools_1.textEditorToolsForModel)(this.selectedModel),
         ];
@@ -28,6 +30,7 @@ class FastTriageAgent extends base_1.BaseAgent {
     }
     async buildSystemPrompt(repoInfoBuilder) {
         const _repoInfo = await repoInfoBuilder();
+        // TODO: Use repo knowledge - can that also replace "issues" as memory?
         return `
 You are a fast triage agent that quickly analyzes test failures. [PLACEHOLDER - to be refined]
 
@@ -44,6 +47,31 @@ Provide a concise summary of:
 - Why it failed (root cause)
 - Whether this is an app issue or a test issue
 
+Use the bash tool to make use of trace-utils and fetch steps from traces,
+errors in network calls, and screenshots around failures.
+
+${trace_utils_docs_1.traceUtilsDoc}
+
+# Expected output
+
+A well-structured markdown report published via the publishMarkdownReport tool.
+
+Your report must include:
+- What failed and why (root cause analysis)
+- Whether this is an app issue or a test issue
+- Evidence from the failure: include screenshots (as file:// URLs), network logs, console logs, etc.
+- Evidence from the last passing run when available: include screenshots showing the expected behavior
+
+Use trace-utils to extract screenshots from both failing and passing runs. Reference them in your
+markdown using file:// URLs (e.g. file:///path/to/screenshot.png). The publishMarkdownReport tool
+will upload these files and replace them with public URLs automatically.
+
+As your final step, always call the publishMarkdownReport tool with your complete markdown report.
+
+# Things to keep in mind
+
+Follow the action, not just the assertion - The console showed deletePlayer error. I should have immediately looked at screenshots around the click that triggered it (test.step@126 - the "Yes" click) rather than the failing expect (expect@92).
+
 Today's date is ${new Date().toDateString()}
 `;
     }