@empiricalrun/test-gen 0.75.0 → 0.76.0

package/dist/agent/chat/index.js

@@ -1,204 +1,137 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.fetchEnvironmentVariables = fetchEnvironmentVariables;
-exports.runChatAgentForCLI = runChatAgentForCLI;
-const llm_1 = require("@empiricalrun/llm");
-const chat_1 = require("@empiricalrun/llm/chat");
-const picocolors_1 = require("picocolors");
-const client_1 = require("../../dashboard/client");
-const reader_1 = require("../../file-info/adapters/file-system/reader");
-const human_in_the_loop_1 = require("../../human-in-the-loop");
-const validation_1 = require("../../recorder/validation");
-const executor_1 = require("../../tools/executor");
-const git_1 = require("../../tools/executor/utils/git");
+exports.ChatAgent = void 0;
+const tools_1 = require("../../tools");
 const base_1 = require("../base");
-const filesystem_cache_1 = require("./filesystem-cache");
-const state_1 = require("./state");
-const utils_1 = require("./utils");
-function stopCriteria(userPrompt) {
-    return userPrompt?.toLowerCase() === "stop";
-}
-function concludeAgent(chatModel, useDiskForChatState, selectedModel, error) {
-    console.log(`\n${(0, picocolors_1.gray)("Usage summary -> " + (0, state_1.getUsageSummary)(chatModel))}`);
-    if (useDiskForChatState) {
-        (0, state_1.saveToDisk)(chatModel.messages, selectedModel, chatModel.askUserForInput, error);
-    }
-}
-async function fetchEnvironmentVariables() {
-    const projectApiKey = process.env.EMPIRICALRUN_API_KEY;
-    let apiClient;
-    let queryParams = {};
-    if (projectApiKey) {
-        apiClient = new client_1.DashboardAPIClient({
-            authType: "project-api-key",
-            projectApiKey: process.env.EMPIRICALRUN_API_KEY,
-        });
-    }
-    else {
-        apiClient = new client_1.DashboardAPIClient({
-            authType: "user-access-token",
-        });
-        const repoName = await (0, validation_1.validatePackageJson)(process.cwd());
-        queryParams = { project_repo_name: repoName };
-    }
-    const data = await apiClient.request("/api/environment-variables", { method: "GET", params: queryParams });
-    if (!data.data || !data.data.environment_variables) {
-        console.error("Failed to fetch environment variables:", data);
-        throw new Error("Failed to fetch environment variables");
-    }
-    const envVars = data.data.environment_variables.reduce((acc, envVar) => {
-        acc[envVar.name] = envVar.value;
-        return acc;
-    }, {});
-    return envVars;
-}
-async function runChatAgentForCLI({ useDiskForChatState, selectedModel, initialPromptContent, useTriage, resetChat, useFSCache, }) {
-    let chatState;
-    const enableStreaming = !useFSCache;
-    const cache = useFSCache ? new filesystem_cache_1.FilesystemLLMCache() : undefined;
-    if (resetChat) {
-        (0, state_1.clearChatState)();
-    }
-    if (useDiskForChatState) {
-        chatState = (0, state_1.loadChatState)();
-    }
-    // TODO: Store branch name in chat state so that we don't recreate it every time
-    const randomId = crypto.randomUUID().substring(0, 8);
-    const branchName = `branch-${randomId}`;
-    await (0, git_1.checkoutBranch)(branchName, process.cwd());
-    let messagesLoadedFromDisk = chatState?.messages || [];
-    let chatModel = (0, chat_1.createChatModel)(messagesLoadedFromDisk, selectedModel, undefined, cache);
-    chatModel.validateEnvVarsForAuth();
-    if (initialPromptContent && chatModel.messages.length === 0) {
-        chatModel.pushUserMessage(initialPromptContent, []);
-    }
-    else if (initialPromptContent && chatModel.messages.length > 0) {
-        console.warn(`Ignoring initial prompt because we have existing messages.`);
-    }
-    if (chatModel.askUserForInput) {
-        // Show last message to the user for context when we loaded from disk
-        const latest = chatModel.getHumanReadableLatestMessage();
-        if (latest) {
-            console.log(`${(0, picocolors_1.blue)(latest.role)}: ${latest.textMessage}`);
+const pw_utils_docs_1 = require("./prompt/pw-utils-docs");
+class ChatAgent extends base_1.BaseAgent {
+    getTools() {
+        const tools = [
+            ...tools_1.commonTools,
+            ...tools_1.testGenerationTools,
+            // Model-specific tools
+            ...(0, tools_1.textEditorToolsForModel)(this.selectedModel),
+        ];
+        // Feature flag: video analysis (if not already included)
+        if (this.featureFlags?.includes("useVideoAnalysis") &&
+            !tools.some((tool) => tool.schema.name === "fetchVideoAnalysis")) {
+            tools.push(tools_1.fetchVideoAnalysis);
         }
+        return tools;
     }
-    if (chatState && chatState.error) {
-        // Reset error state as we are attempting a retry
-        chatState.error = null;
-    }
-    const handleSigInt = () => {
-        concludeAgent(chatModel, useDiskForChatState, selectedModel, null);
-        process.exit(0);
-    };
-    process.once("SIGINT", handleSigInt);
-    process.once("SIGTERM", handleSigInt);
-    let userPrompt;
-    let reporterFunc = async (chatState, latest) => {
-        if (useDiskForChatState) {
-            (0, state_1.saveToDisk)(chatState.messages, selectedModel, chatState.askUserForInput, chatState.error);
-        }
-        if (latest) {
-            if (!enableStreaming) {
-                console.log(`${(0, picocolors_1.blue)(latest.role)}: ${latest.textMessage}`);
-            }
-            else {
-                process.stdout.write(`\n`);
-            }
-        }
-    };
-    const trace = (0, llm_1.createLangfuseTrace)({
-        name: "chat_agent",
-        input: initialPromptContent || "",
-        tags: [selectedModel, "chat_agent"],
-    });
-    if (trace) {
-        const traceUrl = trace.getTraceUrl();
-        console.log(`Starting ${selectedModel}: ${traceUrl}`);
-    }
-    let authType = "user-access-token";
-    if (process.env.EMPIRICALRUN_API_KEY) {
-        authType = "project-api-key";
-    }
-    const apiClient = new client_1.DashboardAPIClient({
-        authType,
-    });
-    const fileInfo = await (0, reader_1.getFileInfoFromFS)(process.cwd());
-    const agentParams = { featureFlags: [], repoInfo: fileInfo, selectedModel };
-    const agent = useTriage
-        ? new base_1.TriageAgent(agentParams)
-        : new base_1.ChatAgent(agentParams);
-    while (!stopCriteria(userPrompt)) {
-        if (chatModel.askUserForInput) {
-            try {
-                userPrompt = await human_in_the_loop_1.humanLoop.getFeedback({
-                    message: "User:",
-                });
-            }
-            catch (e) {
-                // https://github.com/SBoudrias/Inquirer.js/issues/1502#issuecomment-2275991680
-                if (e instanceof Error && e.name === "ExitPromptError") {
-                    concludeAgent(chatModel, useDiskForChatState, selectedModel, null);
-                    process.exit(0);
-                }
-                concludeAgent(chatModel, useDiskForChatState, selectedModel, {
-                    message: e.message,
-                    stack: e.stack || "Stack trace not available",
-                    timestamp: new Date().toISOString(),
-                });
-                throw e;
-            }
-            if (!stopCriteria(userPrompt)) {
-                const { text, attachments } = (0, utils_1.extractAttachments)(userPrompt);
-                chatModel.pushUserMessage(text, attachments);
-            }
-        }
-        else {
-            const toolExecutor = new executor_1.ToolExecutor({
-                chatSession: null,
-                branchName,
-                repoPath: process.cwd(),
-                apiClient,
-                trace,
-                featureFlags: [],
-                environmentOverrides: await fetchEnvironmentVariables(),
-            });
-            await agent.runLoop({
-                chatModel,
-                reporter: reporterFunc,
-                streamingMessageReporter: (() => {
-                    if (!enableStreaming) {
-                        return;
-                    }
-                    let hasStarted = false;
-                    let startedRole = undefined;
-                    return async (delta, snapshot, thinking) => {
-                        if (delta) {
-                            const role = thinking ? "Thinking" : "Assistant";
-                            if (!hasStarted) {
-                                process.stdout.write(`${(0, picocolors_1.blue)(role)}: `);
-                                hasStarted = true;
-                                startedRole = role;
-                            }
-                            else if (hasStarted && role !== startedRole) {
-                                // Changing from thinking -> text block
-                                process.stdout.write("\n");
-                                process.stdout.write(`${(0, picocolors_1.blue)(role)}: `);
-                                startedRole = role;
-                            }
-                            process.stdout.write(delta);
-                        }
-                    };
-                })(),
-                trace,
-                onPendingToolCall: async (toolCalls) => {
-                    const toolResults = await toolExecutor.execute(toolCalls);
-                    chatModel.pushToolResultsMessage(toolCalls, toolResults);
-                },
-            });
+    async buildSystemPrompt(repoContext) {
+        if (!repoContext) {
+            throw new Error(`Chat agent needs repo context`);
         }
+        const preamble = `
+You are a helpful assistant that can answer questions and help with tasks related to writing and maintaining Playwright tests.
+
+You are working on a test code repository that contains Playwright tests and other related files. Your working directory
+has been checked out on a git branch, and your actions that edit files will be automatically committed. If you make any
+file edits, you must create a pull request after your work is done, to get your work reviewed and merged by a human.
+
+# Your capabilities
+
+- Adding new Playwright tests or helper methods
+- Going through test reports and identifying app issues versus test issues
+- Modifying existing tests to adapt to changes in the application
+- Modifying repo configuration (e.g. in playwright.config.ts) and dependencies (e.g. in package.json)
+
+# Going through test reports
+
+- App issues: app issues caught by test failures, like UI issues, API endpoint issues, etc. These are issues that
+will be reported to an app developer to investigate and fix.
+- Test issues: Playwright tests can become outdated when app code changes. These are issues that need to be
+fixed with modifications to the test code, and it is your job to do that.
+
+# Tools
+
+You are given a set of tools to help you fulfill the user's request. Read their descriptions to
+understand what each tool does.
+
+For example, if the user asks you to run a test, you could use the runTest tool.
+Once the test is run, you will receive the results in the form of a JSON object.
+Summarize the results in a few sentences.
+
+If the user provides a diagnosis URL, you can use the fetchDiagnosisDetails tool
+to get more information about the test case and its results.
+
+If the user provides a test run URL, you can use the fetchTestRunDetails tool
+to get detailed information about a specific test run.
+
+Or if the user asks you to modify a test, you could use the generateTestWithBrowserAgent tool. If you suspect
+that a UI selector needs to be updated, using the browser agent is a good idea.
+
+Before using generateTestWithBrowserAgent, you need to prepare the test code for the browser agent.
+You can do this by using the strReplaceEditor or the text editor tool to add a TODO comment to the test
+code. This comment explains to the browser agent what it needs to do.
+
+For example, if the expected modification is to click on a login button, you could add the following comment. 
+
+// TODO(agent): Click on the login button
+
+The position of the comment is important: the browser agent will look for this comment and replace it with 
+the actual code to click on the login button. If you are fixing a failing test, your comment should be 
+around the failing line of code, so that it can be replaced/modified.
+
+# Proactiveness
+
+1. You are allowed to be proactive, but ONLY for read-only tool calls: like searching for content, reading files, fetching data from tools, and
+running Playwright tests.
+2. For any read-write tool calls (e.g. modifying any file), you should share your plan and get the user's approval before proceeding.
+
+# Rules to follow
+
+You must follow these rules while adding new tests or modifying existing tests. There can be exceptions to these rules, but
+ONLY when explicitly asked for by the user.
+
+1. You can't delete some steps from the test to make it pass. The test needs to accomplish its objective (which is to validate a particular user scenario)
+2. Do not add any conditional logic or try catch blocks in a test. A good test deterministically tests a user scenario
+3. Trust Playwright's ability to auto-wait while taking actions on elements.
+   - Example 1: Do not add checks on locator.isVisible() before clicking on it: Playwright already waits for visibility on locator.click()
+   - Example 2: Do not add page.waitForLoadState after a page.goto: Playwright already waits for page "load" event in page.goto()
+4. Do not add waitForTimeout or waitForLoadState in a test. Playwright will automatically wait for the page to load.
+5. Try/catch blocks are a code smell for tests: you should not use them.
+6. Do not use then() or catch() syntax in a test. Use async/await only
+
+There are few exceptions to these rules. BEFORE applying any of the following exceptions, you MUST share your plan with the user and get their approval.
+
+## Exceptions for conditional logic
+
+There are few exceptions where you can add conditional logic to a test. If the application UI reveals some UI elements on certain conditions, we can add conditional logic.
+
+For example, a form view shows a "Save" button only when the form is dirty. In this case, we will have to check if the "Save" button is visible before clicking on it. To do this,
+follow this pattern:
+
+\`\`\`
+const saveButton = page.getByRole('button', { name: 'Save' });
+if (await saveButton.isVisible()) {
+  await saveButton.click();
+}
+\`\`\`
+
+Note that locator.isVisible() DOES NOT wait for the element to be visible. If the element in question shows up after a delay, we have no option but to add a waitForTimeout.
+
+\`\`\`
+const saveButton = page.getByRole('button', { name: 'Save' });
+await page.waitForTimeout(100); // Wait for the element to be visible -- only if necessary.
+if (await saveButton.isVisible()) {
+  await saveButton.click();
+}
+\`\`\`
+
+`;
+        return `${preamble}
+
+# Recipes
+${pw_utils_docs_1.playwrightUtilsDocs}
+
+# Repo context
+${repoContext}
+
+# Reference
+Today's date is ${new Date().toDateString()}
+`;
     }
-    trace?.update({ output: { messages: chatModel.messages } });
-    await llm_1.langfuseInstance?.flushAsync();
-    const usageSummary = (0, state_1.getUsageSummary)(chatModel);
-    console.log(`\n${(0, picocolors_1.gray)("Usage summary -> " + usageSummary)}`);
 }
+exports.ChatAgent = ChatAgent;

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

@@ -1,5 +1,6 @@
-import { FileInfo } from "@empiricalrun/shared-types";
-export declare const getSystemPrompt: (repoInfo: FileInfo, mode?: string) => Promise<string>;
-export declare function buildTriageSystemPrompt(repoContext: string): Promise<string>;
-export declare function buildSystemPrompt(repoContext: string): Promise<string>;
+export declare const AGENT_TO_SYSTEM_PROMPT_MAP: {
+    chat: (repoContext: string) => Promise<string>;
+    video: (repoContext: string) => Promise<string>;
+    "code-review": (repoContext: string) => Promise<string>;
+};
 //# sourceMappingURL=index.d.ts.map

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

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/agent/chat/prompt/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,4BAA4B,CAAC;AAKtD,eAAO,MAAM,eAAe,GAAU,UAAU,QAAQ,EAAE,OAAO,MAAM,oBAMtE,CAAC;AAEF,wBAAsB,uBAAuB,CAAC,WAAW,EAAE,MAAM,mBA8DhE;AAED,wBAAsB,iBAAiB,CAAC,WAAW,EAAE,MAAM,mBA8G1D"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../src/agent/chat/prompt/index.ts"],"names":[],"mappings":"AAEA,eAAO,MAAM,0BAA0B;wBACjB,MAAM;yBACL,MAAM;iCACE,MAAM;CAEpC,CAAC"}

package/dist/agent/chat/prompt/index.js

@@ -1,82 +1,93 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.getSystemPrompt = void 0;
-exports.buildTriageSystemPrompt = buildTriageSystemPrompt;
-exports.buildSystemPrompt = buildSystemPrompt;
+exports.AGENT_TO_SYSTEM_PROMPT_MAP = void 0;
 const pw_utils_docs_1 = require("./pw-utils-docs");
-const repo_1 = require("./repo");
-const getSystemPrompt = async (repoInfo, mode) => {
-    const repoContext = await (0, repo_1.getRepoInfoPrompt)(repoInfo);
-    if (mode === "triage") {
-        return await buildTriageSystemPrompt(repoContext);
-    }
-    return await buildSystemPrompt(repoContext);
+exports.AGENT_TO_SYSTEM_PROMPT_MAP = {
+    chat: (repoContext) => buildChatSystemPrompt(repoContext),
+    video: (repoContext) => buildVideoAnalysisSystemPrompt(repoContext),
+    "code-review": (repoContext) => buildCodeReviewAgentSystemPrompt(repoContext),
 };
-exports.getSystemPrompt = getSystemPrompt;
-async function buildTriageSystemPrompt(repoContext) {
+async function buildCodeReviewAgentSystemPrompt(repoContext) {
+    console.warn("not using repoContext in code review agent prompt for now", typeof repoContext);
     return `
-You are a helpful assistant that help with analysis of Playwright test reports. Your goal is to help the user analyse a test report and identify the root cause of the test failures, and log the unique failuers as issues so that the user can keep a track and fix them.
-
-You are working on a test code repository that contains Playwright tests and other related files. Your working directory has been checked out on a git branch.
-
-# Your capabilities
-
-When provided with a test report URL, you can use these capabilities to triage the test failures in the report:
-
-## Fetch and view the test report
-
-- Use viewFailedTestRunReportTool tool to get more information about all tests that failed in the run
-
-## Analyze each test case
-
-You are provided with multiple tools to help you understand each failing test case better. Understanding each test case allows you to identify the root cause and create more accurate issues. These tools can also be called in parallel.
-
-- Each test case generates artifacts: images, videos, playwright trace zip file. With your tools, you can fetch image, analyze the video frames and trace.zip to find out failing network requests and console logs
-  - Each video represents one browser tab of the test case (so multiple videos implies the test had multiple tabs or browser windows)
-- Read the error stack and test file to understand what the test is doing
-- Fetch the last successful run of the test case to understand the earlier flow. This report will contain image and video URLs that can also be analyzed with your available tools.
-- If you think the issue is explained by a timing or intermittent issue, you can also re-run the test case
-
-## Listing, updating and creating issues
-
-- Test failures will become issues that can be assigned to developers to fix the app or update the test. Similar test failures should be grouped into one issue to avoid duplicates.
-- Before you create a new issue, you MUST list existing issues that have been created for this repo, to avoid creating duplicate issues.
-- If you find duplicates, use the update issue tool to update the existing issue with new information from the test report
-- When you are creating a new issue, use the description and title to clearly call out the error reason (share error stack, error message, relevant lines of code, etc.) so that a follow-up triaging session can match the issue against a new failure and avoid duplicate issues.
-- What makes a good issue: accurate classification between app or test issue, accurate grouping, and a good auto-fix prompt - see more about this below
-
-## Classify tests as app or test issues
-- An app issue is an issue in the application that is being tested. This often shows up as a network failure, or error message in the console log, or an error toast in the UI. Use the last successful run artifacts to compare the app state between the successful and failed run.
-- A test issue is an issue in the test code. If the application has changed the UI, a selector in the test may no longer work. Or if the application has changed the flow, the test may need to be updated to reflect the new flow.
+You are an expert code reviewer that specializes in reviewing Playwright test code. You are
+provided with tools to fetch diff for a code review, where a test has been added, test modified,
+or some configuration has changed.
+
+# Your goals
+- Identify code smells in test code - see below
+- Call out test data assumptions or lack of clean up
+
+# Output format
+- You are expected to return two sections in your response: describe_code_change and code_review_comments
+- describe_code_change: A brief summary of what the code change is doing. This should be 4-6 sentences in a bullet list.
+- code_review_comments: A bulleted list of code review comments that catch for any of the specific bits below or other
+  red flags you might see in the code. Each comment should be 1-2 sentences.
+
+Return these as XML tags with markdown inside them
+
+<describe_code_change>
+- ...
+</describe_code_change>
+
+<code_review_comments>
+- ...
+</code_review_comments>
+
+# Specific bits to catch in the code review
+
+## Code smells to look for
+- Any form of try-catch or exception handling is a code smell in test code. If there's an
+  exception, the test should fail
+- Any conditionals (if, switch, ternary) in test code is a code smell. Tests are expected to be
+  deterministic. If you see conditionals, check if there's a comment explaining why it's needed.
+  Critically review the comment -- if it's not convincing, call it out as a code smell.
+
+## Ensure Playwright best practices
+- Use locators instead of selectors: waitForSelector, $, $$ are bad - use locators instead (e.g. locator.waitFor)
+- If the test relies on some Playwright APIs that do not auto-wait (e.g. isVisible(), count()), we need to ensure 
+  they are used AFTER some action that ensures the page has loaded. If nothing, at least it should have a waitForTimeout
+- Don't use waitForLoadState or networkidle - these are not required since Playwright auto-waits after navigations
+
+## Call out test data assumptions
+- If new test data is created (e.g. creating a new entity in the app, doing some actions on it) - it should be cleaned up
+  at the end of the test. If not, call it out.
+- If the test data cannot be cleaned up, are we using some random names to ensure no conflicts in future test runs?
+- If the test assumes some data exists (e.g. a user with a specific email) - call it out. It might fail across other
+  environments.
+- No hard coded URLs - use relative URLs instead - that can work across environments.
+- Dependency on static data that can change across environments (e.g. number of rows in a table) should be avoided.
+
+## Remove debug artifacts
+- If there are console.logs or page.screenshot usage, call it out. They should be removed before merging.
+`;
+}
+async function buildVideoAnalysisSystemPrompt(repoContext) {
+    console.warn("not using repoContext in video analysis prompt for now", typeof repoContext);
+    return `
+You are a video analysis agent specialized in analyzing screen recordings and user interface interactions.
 
-## Grouping test failures
-- Before you create issues, group the failures together so that we create useful issues
-- What makes a good group: failures that have the same root cause - because of similar error stacks - and can be fixed with the same change to the app or test
-  - Both "type of failure" and "proposed fix" are important to determine if two failures belong to the same group
-  - Example: if two tests fail with strict mode violations, but for 2 different selectors, they are different groups because the proposed fixes are different
-- What does not make a good group: the location of the test or the name of the test. Two tests that are located in the same file or have similar names should ONLY be grouped together if the root cause of failures is same
+Available Tools:
+- extract_frames: Extract frames from videos for detailed visual analysis
+- fetch_video_analysis: Get comprehensive video analysis summary (legacy)
 
-## Crafting a good auto-fix prompt for test issues
-- When you create issues with type "test", you are expected to share a test_issue_prompt which is your proposed change to the test to adapt to the new app state.
-- This prompt is handed over to another agent to update the test code, and your prompt is the ONLY context that the agent has to update the test.
-- Therefore, your prompt must contain:
-  - Which test cases to be updated - with test and describe block names, file name
-  - What failed in the test - error message, error stack, relevant lines of code, or bits to locate the failure
-  - Your suggested change to the test
-  - Emphasis to re-run the test after making the change, to ensure that the change works
+When analyzing videos:
+1. Use extract_frames to get individual frames for detailed analysis
+2. Analyze each frame for UI elements, user actions, and state changes  
+3. Provide specific observations about what's happening in each frame
+4. The Summary should be in a bullet point format
+5. Reference frame IDs when discussing specific moments: "In frame_abc123_001, I can see..."
 
-## Conclusion
-- After you are done with triaging and creating issues, summarize the work done with a list of created issues for the user to review. Don't be too verbose - a bullet list of issues created or updated, with a small description is enough.
-- It is important to show proof that you have gone through all of the failures in the test run report, so use numbers to call out 1. total failures, and 2. failures associated with each issue.
+Your analysis should be:
+- Detailed and specific about UI elements and interactions
+- Sequential, following the flow of actions in the video
 
-# Repo context
-${repoContext}
+After the final summary you need to include the key frame IDs that best represent the important moments in the video.
 
-# Reference
-Today's date is ${new Date().toDateString()}
-`;
+Exmple Frame Id Reference: <frame_abc123_001>
+  `;
 }
-async function buildSystemPrompt(repoContext) {
+async function buildChatSystemPrompt(repoContext) {
     const preamble = `
 You are a helpful assistant that can answer questions and help with tasks related to writing and maintaining Playwright tests.
 

package/dist/agent/chat/state.d.ts

@@ -17,9 +17,8 @@ export declare function createChatStateForMessages({ messages, selectedModel, as
     askUserForInput: boolean;
     error: ChatStateError | null;
 }): ChatState;
-export declare function chatStateFromModel<T>({ chatModel, selectedModel, error, }: {
+export declare function chatStateFromModel<T>({ chatModel, error, }: {
     chatModel: IChatModel<T>;
-    selectedModel: SupportedChatModels;
     error: ChatStateError | null;
 }): ChatState;
 export declare function getUsageSummary(chatModel: IChatModel<CanonicalMessage>): string;

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

@@ -1 +1 @@
-{"version":3,"file":"state.d.ts","sourceRoot":"","sources":["../../../src/agent/chat/state.ts"],"names":[],"mappings":"AAAA,OAAO,EAIL,UAAU,EACX,MAAM,wBAAwB,CAAC;AAChC,OAAO,EACL,UAAU,EACV,gBAAgB,EAChB,SAAS,EACT,cAAc,EACd,eAAe,EACf,mBAAmB,EAEnB,UAAU,EACX,MAAM,4BAA4B,CAAC;AA+BpC,eAAO,MAAM,kCAAkC,EAAE,MAAM,CACrD,MAAM,EACN,CAAC,KAAK,EAAE,GAAG,KAAK,GAAG,CAIpB,CAAC;AAEF,eAAO,MAAM,yBAAyB,QAAQ,CAAC;AAE/C,eAAO,MAAM,eAAe,QAI3B,CAAC;AAEF,wBAAgB,cAAc,SAQ7B;AAED,wBAAgB,eAAe,CAAC,EAC9B,UAAU,EACV,WAAW,EACX,aAAa,EACb,aAAa,EACb,KAAK,GACN,EAAE;IACD,UAAU,EAAE,MAAM,GAAG,SAAS,CAAC;IAC/B,WAAW,EAAE,UAAU,EAAE,CAAC;IAC1B,aAAa,EAAE,SAAS,GAAG,SAAS,CAAC;IACrC,aAAa,EAAE,mBAAmB,CAAC;IACnC,KAAK,EAAE,cAAc,GAAG,IAAI,CAAC;CAC9B,aAYA;AAED,wBAAgB,0BAA0B,CAAC,EACzC,QAAQ,EACR,aAAa,EACb,eAAe,EACf,KAAK,GACN,EAAE;IACD,QAAQ,EAAE,gBAAgB,EAAE,CAAC;IAC7B,aAAa,EAAE,mBAAmB,CAAC;IACnC,eAAe,EAAE,OAAO,CAAC;IACzB,KAAK,EAAE,cAAc,GAAG,IAAI,CAAC;CAC9B,GAAG,SAAS,CASZ;AAED,wBAAgB,kBAAkB,CAAC,CAAC,EAAE,EACpC,SAAS,EACT,aAAa,EACb,KAAK,GACN,EAAE;IACD,SAAS,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC;IACzB,aAAa,EAAE,mBAAmB,CAAC;IACnC,KAAK,EAAE,cAAc,GAAG,IAAI,CAAC;CAC9B,aAOA;AAED,wBAAgB,eAAe,CAC7B,SAAS,EAAE,UAAU,CAAC,gBAAgB,CAAC,GACtC,MAAM,CA0BR;AAED,wBAAgB,aAAa,IAAI,SAAS,GAAG,SAAS,CAarD;AAED;;;GAGG;AACH,wBAAgB,gBAAgB,CAAC,QAAQ,EAAE,GAAG,GAAG,SAAS,CAqBzD;AAED,wBAAgB,UAAU,CACxB,QAAQ,EAAE,KAAK,CAAC,gBAAgB,CAAC,EACjC,aAAa,EAAE,mBAAmB,EAClC,eAAe,EAAE,OAAO,EACxB,KAAK,EAAE,cAAc,GAAG,IAAI,QAgB7B;AA2BD,wBAAgB,yBAAyB,CACvC,QAAQ,EAAE,gBAAgB,EAAE,GAC3B,MAAM,GAAG,IAAI,CAef;AAED,wBAAgB,mBAAmB,CACjC,UAAU,EAAE,MAAM,EAClB,QAAQ,EAAE,gBAAgB,EAAE,GAC3B;IACD,gBAAgB,EAAE,IAAI,GAAG,SAAS,CAAC;IACnC,QAAQ,EAAE,eAAe,GAAG,SAAS,CAAC;IACtC,UAAU,EAAE,UAAU,GAAG,SAAS,CAAC;CACpC,CA0BA"}
+{"version":3,"file":"state.d.ts","sourceRoot":"","sources":["../../../src/agent/chat/state.ts"],"names":[],"mappings":"AAAA,OAAO,EAIL,UAAU,EACX,MAAM,wBAAwB,CAAC;AAChC,OAAO,EACL,UAAU,EACV,gBAAgB,EAChB,SAAS,EACT,cAAc,EACd,eAAe,EACf,mBAAmB,EAEnB,UAAU,EACX,MAAM,4BAA4B,CAAC;AA+BpC,eAAO,MAAM,kCAAkC,EAAE,MAAM,CACrD,MAAM,EACN,CAAC,KAAK,EAAE,GAAG,KAAK,GAAG,CAIpB,CAAC;AAEF,eAAO,MAAM,yBAAyB,QAAQ,CAAC;AAE/C,eAAO,MAAM,eAAe,QAI3B,CAAC;AAEF,wBAAgB,cAAc,SAQ7B;AAED,wBAAgB,eAAe,CAAC,EAC9B,UAAU,EACV,WAAW,EACX,aAAa,EACb,aAAa,EACb,KAAK,GACN,EAAE;IACD,UAAU,EAAE,MAAM,GAAG,SAAS,CAAC;IAC/B,WAAW,EAAE,UAAU,EAAE,CAAC;IAC1B,aAAa,EAAE,SAAS,GAAG,SAAS,CAAC;IACrC,aAAa,EAAE,mBAAmB,CAAC;IACnC,KAAK,EAAE,cAAc,GAAG,IAAI,CAAC;CAC9B,aAYA;AAED,wBAAgB,0BAA0B,CAAC,EACzC,QAAQ,EACR,aAAa,EACb,eAAe,EACf,KAAK,GACN,EAAE;IACD,QAAQ,EAAE,gBAAgB,EAAE,CAAC;IAC7B,aAAa,EAAE,mBAAmB,CAAC;IACnC,eAAe,EAAE,OAAO,CAAC;IACzB,KAAK,EAAE,cAAc,GAAG,IAAI,CAAC;CAC9B,GAAG,SAAS,CASZ;AAED,wBAAgB,kBAAkB,CAAC,CAAC,EAAE,EACpC,SAAS,EACT,KAAK,GACN,EAAE;IACD,SAAS,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC;IACzB,KAAK,EAAE,cAAc,GAAG,IAAI,CAAC;CAC9B,aAOA;AAED,wBAAgB,eAAe,CAC7B,SAAS,EAAE,UAAU,CAAC,gBAAgB,CAAC,GACtC,MAAM,CA0BR;AAED,wBAAgB,aAAa,IAAI,SAAS,GAAG,SAAS,CAarD;AAED;;;GAGG;AACH,wBAAgB,gBAAgB,CAAC,QAAQ,EAAE,GAAG,GAAG,SAAS,CAqBzD;AAED,wBAAgB,UAAU,CACxB,QAAQ,EAAE,KAAK,CAAC,gBAAgB,CAAC,EACjC,aAAa,EAAE,mBAAmB,EAClC,eAAe,EAAE,OAAO,EACxB,KAAK,EAAE,cAAc,GAAG,IAAI,QAgB7B;AA2BD,wBAAgB,yBAAyB,CACvC,QAAQ,EAAE,gBAAgB,EAAE,GAC3B,MAAM,GAAG,IAAI,CAef;AAED,wBAAgB,mBAAmB,CACjC,UAAU,EAAE,MAAM,EAClB,QAAQ,EAAE,gBAAgB,EAAE,GAC3B;IACD,gBAAgB,EAAE,IAAI,GAAG,SAAS,CAAC;IACnC,QAAQ,EAAE,eAAe,GAAG,SAAS,CAAC;IACtC,UAAU,EAAE,UAAU,GAAG,SAAS,CAAC;CACpC,CA0BA"}

package/dist/agent/chat/state.js

@@ -77,10 +77,10 @@ function createChatStateForMessages({ messages, selectedModel, askUserForInput,
         error: error,
     };
 }
-function chatStateFromModel({ chatModel, selectedModel, error, }) {
+function chatStateFromModel({ chatModel, error, }) {
     return createChatStateForMessages({
         messages: chatModel.messages,
-        selectedModel,
+        selectedModel: chatModel.selectedModel,
         askUserForInput: chatModel.askUserForInput,
         error,
     });

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

@@ -1,12 +1,11 @@
 import { TraceClient } from "@empiricalrun/llm";
 import { IChatModel } from "@empiricalrun/llm/chat";
-import { Attachment, ReporterFunction, SupportedChatModels } from "@empiricalrun/shared-types";
+import { Attachment, ReporterFunction } from "@empiricalrun/shared-types";
 export declare const log: (...args: any[]) => void;
 export declare function getModelName(model: string): string;
-export declare function handleAgentError({ error, chatModel, selectedModel, reporter, trace, }: {
+export declare function handleAgentError({ error, chatModel, reporter, trace, }: {
     error: unknown;
     chatModel: IChatModel<any>;
-    selectedModel: SupportedChatModels;
     reporter: ReporterFunction;
     trace?: TraceClient;
 }): Promise<void>;

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

@@ -1 +1 @@
-{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../../../src/agent/chat/utils.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAChD,OAAO,EAAuB,UAAU,EAAE,MAAM,wBAAwB,CAAC;AACzE,OAAO,EACL,UAAU,EAEV,gBAAgB,EAChB,mBAAmB,EACpB,MAAM,4BAA4B,CAAC;AAKpC,eAAO,MAAM,GAAG,GAAI,GAAG,MAAM,GAAG,EAAE,SAEjC,CAAC;AAcF,wBAAgB,YAAY,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,CAMlD;AAED,wBAAsB,gBAAgB,CAAC,EACrC,KAAK,EACL,SAAS,EACT,aAAa,EACb,QAAQ,EACR,KAAK,GACN,EAAE;IACD,KAAK,EAAE,OAAO,CAAC;IACf,SAAS,EAAE,UAAU,CAAC,GAAG,CAAC,CAAC;IAC3B,aAAa,EAAE,mBAAmB,CAAC;IACnC,QAAQ,EAAE,gBAAgB,CAAC;IAC3B,KAAK,CAAC,EAAE,WAAW,CAAC;CACrB,iBAgBA;AAED,wBAAgB,kBAAkB,CAAC,UAAU,EAAE,MAAM,GAAG;IACtD,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,UAAU,EAAE,CAAC;CAC3B,CAsBA"}
+{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../../../src/agent/chat/utils.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAChD,OAAO,EAAuB,UAAU,EAAE,MAAM,wBAAwB,CAAC;AACzE,OAAO,EACL,UAAU,EAEV,gBAAgB,EACjB,MAAM,4BAA4B,CAAC;AAKpC,eAAO,MAAM,GAAG,GAAI,GAAG,MAAM,GAAG,EAAE,SAEjC,CAAC;AAcF,wBAAgB,YAAY,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,CAMlD;AAED,wBAAsB,gBAAgB,CAAC,EACrC,KAAK,EACL,SAAS,EACT,QAAQ,EACR,KAAK,GACN,EAAE;IACD,KAAK,EAAE,OAAO,CAAC;IACf,SAAS,EAAE,UAAU,CAAC,GAAG,CAAC,CAAC;IAC3B,QAAQ,EAAE,gBAAgB,CAAC;IAC3B,KAAK,CAAC,EAAE,WAAW,CAAC;CACrB,iBAeA;AAED,wBAAgB,kBAAkB,CAAC,UAAU,EAAE,MAAM,GAAG;IACtD,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,UAAU,EAAE,CAAC;CAC3B,CAsBA"}

package/dist/agent/chat/utils.js

@@ -29,7 +29,7 @@ function getModelName(model) {
         return model;
     return model;
 }
-async function handleAgentError({ error, chatModel, selectedModel, reporter, trace, }) {
+async function handleAgentError({ error, chatModel, reporter, trace, }) {
     const errorObject = {
         message: error.message,
         stack: error.stack || "Stack trace not available",
@@ -37,7 +37,6 @@ async function handleAgentError({ error, chatModel, selectedModel, reporter, tra
     };
     await reporter((0, state_1.chatStateFromModel)({
         chatModel,
-        selectedModel,
         error: errorObject,
     }), chatModel.getHumanReadableLatestMessage());
     trace?.update({ output: { error: errorObject } });

package/dist/agent/cli.d.ts

@@ -0,0 +1,11 @@
+import { AgentModeEnum, SupportedChatModels } from "@empiricalrun/shared-types";
+export declare function fetchEnvironmentVariables(): Promise<Record<string, string>>;
+export declare function runChatAgentForCLI({ useDiskForChatState, selectedModel, initialPromptContent, agentMode, resetChat, useFSCache, }: {
+    selectedModel: SupportedChatModels;
+    useDiskForChatState: boolean;
+    initialPromptContent: string | undefined;
+    agentMode: AgentModeEnum;
+    useFSCache: boolean;
+    resetChat: boolean;
+}): Promise<void>;
+//# sourceMappingURL=cli.d.ts.map

package/dist/agent/cli.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cli.d.ts","sourceRoot":"","sources":["../../src/agent/cli.ts"],"names":[],"mappings":"AAEA,OAAO,EACL,aAAa,EAMb,mBAAmB,EACpB,MAAM,4BAA4B,CAAC;AAwCpC,wBAAsB,yBAAyB,IAAI,OAAO,CACxD,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CACvB,CAkCA;AAED,wBAAsB,kBAAkB,CAAC,EACvC,mBAAmB,EACnB,aAAa,EACb,oBAAoB,EACpB,SAAS,EACT,SAAS,EACT,UAAU,GACX,EAAE;IACD,aAAa,EAAE,mBAAmB,CAAC;IACnC,mBAAmB,EAAE,OAAO,CAAC;IAC7B,oBAAoB,EAAE,MAAM,GAAG,SAAS,CAAC;IACzC,SAAS,EAAE,aAAa,CAAC;IACzB,UAAU,EAAE,OAAO,CAAC;IACpB,SAAS,EAAE,OAAO,CAAC;CACpB,iBAoLA"}