@minded-ai/mindedjs 1.0.0-ec2-beta-14 → 1.0.0-ec2-beta-16

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@minded-ai/mindedjs",
-  "version": "1.0.0-ec2-beta-14",
+  "version": "1.0.0-ec2-beta-16",
   "description": "MindedJS is a TypeScript library for building agents.",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package/src/agent.ts

@@ -510,7 +510,7 @@ export class Agent {
       // Add playbooks to messages
       const combinedPlaybooks = combinePlaybooks(this.playbooks);
       if (combinedPlaybooks) {
-        const compiledPrompt = compilePrompt(combinedPlaybooks, state);
+        const compiledPrompt = compilePrompt(combinedPlaybooks, { state: state, memory: state.memory, env: process.env });
         const systemMessage = new SystemMessage(compiledPrompt);
         if (state.messages.length === 0 || state.messages[0].getType() === 'system') {
           state.messages[0] = systemMessage;

package/src/browserTask/executeBrowserTask.py

@@ -172,12 +172,11 @@ def create_pydantic_model_from_schema(output_schema: Optional[List[Dict[str, Any
 
 
 async def main(session_id: str, cdp_url: str, task: str, output_schema_json: Optional[str] = None, 
-               otp_secret: Optional[str] = None, screenshot_config: Optional[Dict[str, Any]] = None):
+               otp_secret: Optional[str] = None, screenshot_config: Optional[Dict[str, Any]] = None, folder_path: str = None):
   llm = ChatOpenAI(
     model="gpt-4.1",
     api_key=os.getenv("OPENAI_API_KEY"),
   )
-  folder_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), f'downloads_{session_id}')
   
   # Create folder if it doesn't exist
   os.makedirs(folder_path, exist_ok=True)
@@ -299,11 +298,12 @@ if __name__ == '__main__':
     output_schema = payload.get('outputSchema')
     otp_secret = payload.get('otpSecret')
     screenshot_config = payload.get('screenshotConfig')
+    folder_path = payload.get('folderPath')
 
-    if not session_id or not cdp_url or not task:
-      raise SystemExit("Missing required fields in JSON payload: sessionId, cdpUrl, task")
+    if not session_id or not cdp_url or not task or not folder_path:
+      raise SystemExit("Missing required fields in JSON payload: sessionId, cdpUrl, task, folderPath")
 
     output_schema_json = json.dumps(output_schema) if output_schema is not None else None
-    asyncio.run(main(session_id, cdp_url, task, output_schema_json, otp_secret, screenshot_config))
+    asyncio.run(main(session_id, cdp_url, task, output_schema_json, otp_secret, screenshot_config, folder_path))
   else:
     raise SystemExit("Usage: uv run executeBrowserTask.py; send a JSON payload via stdin")

package/src/browserTask/executeBrowserTask.ts

@@ -2,6 +2,7 @@ import { spawn } from 'node:child_process';
 import { logger } from '../utils/logger';
 import { mindedConnection } from '../platform/mindedConnection';
 import path from 'path';
+import { existsSync, readdirSync } from 'fs';
 import {
   mindedConnectionSocketMessageType,
   CreateBrowserSessionResponse,
@@ -120,6 +121,9 @@ export const invokeBrowserTask = async (options: InvokeBrowserTaskOptions): Prom
       validateLocalOperatorSetup();
       const pythonScriptPath = path.resolve(__dirname, 'executeBrowserTask.py');
 
+      // Calculate folder path for downloads
+      const folderPath = path.join(__dirname, `downloads_${sessionId}`);
+
       const args = ['run', pythonScriptPath];
       logger.info({
         message: 'Spawning Python process',
@@ -146,6 +150,7 @@ export const invokeBrowserTask = async (options: InvokeBrowserTaskOptions): Prom
         outputSchema,
         otpSecret: process.env.TOTP_SECRET,
         screenshotConfig,
+        folderPath,
       };
 
       // Write JSON payload to stdin
@@ -167,14 +172,11 @@ export const invokeBrowserTask = async (options: InvokeBrowserTaskOptions): Prom
         process.stderr.write(text);
       });
 
-      let wasKilledByUs = false;
-
       const interval = setInterval(() => {
         // Check if chromium process stopped running
         if (!isLocalBrowserRunning()) {
           logger.error({ message: 'Local browser process stopped running, killing browser task' });
-          wasKilledByUs = true;
-          child.kill('SIGTERM');
+          child.kill();
           clearInterval(interval);
         }
       }, 1000);
@@ -191,11 +193,6 @@ export const invokeBrowserTask = async (options: InvokeBrowserTaskOptions): Prom
 
       clearInterval(interval);
 
-      if (wasKilledByUs) {
-        logger.error({ message: 'Browser task was killed by user', stderr: stderrBuffer });
-        throw new Error('Task was stopped by user');
-      }
-
       if (exitCode !== 0) {
         logger.error({ message: 'Operator failed', exitCode, stderr: stderrBuffer });
         throw new Error(`Local browser task failed with exit code ${exitCode}`);
@@ -212,10 +209,23 @@ export const invokeBrowserTask = async (options: InvokeBrowserTaskOptions): Prom
         }
       }
 
+      // Read downloaded files from the folder
+      let downloadedFiles: string[] = [];
+      if (existsSync(folderPath)) {
+        try {
+          const files = readdirSync(folderPath);
+          downloadedFiles = files.map((file) => path.join(folderPath, file));
+          logger.debug({ message: 'Found downloaded files', count: downloadedFiles.length, files: downloadedFiles });
+        } catch (error) {
+          logger.error({ message: 'Failed to read downloads folder', error });
+        }
+      }
+
       return {
         result,
         steps: [],
         recordings: [],
+        downloadedFiles,
       };
     }
 

package/src/edges/createPromptRouter.ts

@@ -107,9 +107,12 @@ export const createPromptRouter = ({
     const promptTemplate = includeReasoning ? ROUTER_PROMPT : ROUTER_PROMPT_WITHOUT_REASONING;
 
     // Compile prompt with EJS and parameters
-    const routerPrompt = compilePrompt(promptTemplate, state, {
+    const routerPrompt = compilePrompt(promptTemplate, {
       steps: stepsStr,
       messages: messagesStr,
+      memory: state.memory,
+      state: state,
+      env: process.env,
     });
 
     // Define response schema

package/src/internalTools/appActionRunnerTool.ts

@@ -47,7 +47,7 @@ const appActionRunnerToolCreator = (schema: ActionInputParam[], nodeTitle: strin
     name: nodeTitle,
     description,
     input: zodSchema,
-    execute: async ({ input }: ToolExecuteInput<typeof zodSchema>) => {
+    execute: async ({ input, state }: ToolExecuteInput<typeof zodSchema>) => {
       const response = await mindedConnection.awaitEmit(
         mindedConnectionSocketMessageType.OnAppAction,
         {
@@ -57,6 +57,7 @@ const appActionRunnerToolCreator = (schema: ActionInputParam[], nodeTitle: strin
         20000,
       );
       return {
+        state,
         result: response as { result?: any },
       };
     },

package/src/nodes/addAppToolNode.ts

@@ -39,12 +39,13 @@ export const addAppToolNode = async ({
 
     // Compile parameters with variable injection support
     const compiledParameters: Record<string, any> = {};
+    const compileContext = { state, memory: state.memory, env: process.env };
 
     for (const [key, value] of Object.entries(node.parameters || {})) {
       if (value !== '') {
         // If the value is a string, compile it to allow variable injection
         if (typeof value === 'string') {
-          compiledParameters[key] = compilePrompt(value, state);
+          compiledParameters[key] = compilePrompt(value, compileContext);
         } else {
           compiledParameters[key] = value;
         }
@@ -69,7 +70,7 @@ export const addAppToolNode = async ({
     const combinedPlaybooks = combinePlaybooks(agent.playbooks) || '';
 
     // Compile the prompt if it exists to allow variable injection
-    const compiledNodePrompt = node.prompt ? compilePrompt(node.prompt, state) : null;
+    const compiledNodePrompt = node.prompt ? compilePrompt(node.prompt, compileContext) : null;
 
     // Check if any compiled parameter is too long (>1000 characters)
     const hasLongParameters = Object.values(compiledParameters).some((value) => typeof value === 'string' && value.length > 1000);
@@ -106,7 +107,7 @@ export const addAppToolNode = async ({
     User instructions for choosing tool parameters are:
         ${compiledNodePrompt ? compiledNodePrompt : 'no instructions set by the user'}`;
 
-    const compiledPrompt = compilePrompt(message, state);
+    const compiledPrompt = compilePrompt(message, { state: state, memory: state.memory, env: process.env });
     const systemMessage = new SystemMessage(compiledPrompt);
     if (state.messages.length === 0 || state.messages[0].getType() === 'system') {
       state.messages[0] = systemMessage;

package/src/nodes/addBrowserTaskNode.ts

@@ -39,7 +39,7 @@ export const addBrowserTaskNode = async ({ graph, node, agent, llm }: AddBrowser
 
     const combinedPlaybooks = combinePlaybooks(agent.playbooks);
     if (combinedPlaybooks) {
-      const compiledPrompt = compilePrompt(combinedPlaybooks, state);
+      const compiledPrompt = compilePrompt(combinedPlaybooks, { state: state, memory: state.memory });
       const systemMessage = new SystemMessage(compiledPrompt);
       if (state.messages.length === 0 || state.messages[0].getType() === 'system') {
         state.messages[0] = systemMessage;
@@ -59,8 +59,16 @@ export const addBrowserTaskNode = async ({ graph, node, agent, llm }: AddBrowser
       const toolCall = AIToolCallMessage.tool_calls[0];
       const inputParams = toolCall.args || {};
 
+      // Prepare parameters for prompt compilation
+      const promptParams = {
+        input: inputParams,
+        memory: state.memory,
+        state: state,
+        system: { currentTime: new Date().toISOString() },
+      };
+
       // Compile the prompt with parameters
-      const compiledPrompt = compilePrompt(node.prompt, state, { input: inputParams });
+      const compiledPrompt = compilePrompt(node.prompt, { ...promptParams });
 
       // Build the full prompt with compiled content
       const fullPrompt = `

package/src/nodes/addBrowserTaskRunNode.ts

@@ -53,7 +53,7 @@ export const addBrowserTaskRunNode = async ({ graph, browserTaskNode, attachedTo
         }));
 
       // We compile the env variables to avoid having to pass them to the platform in the tool input
-      const promptCompiledWithEnv = compilePrompt(prompt, state);
+      const promptCompiledWithEnv = compilePrompt(prompt, { env: process.env });
 
       const { browserTaskMode } = getConfig();
 
@@ -111,6 +111,7 @@ export const addBrowserTaskRunNode = async ({ graph, browserTaskNode, attachedTo
         hasResult: !!result.result,
         stepCount: result.steps?.length || 0,
         recordingCount: result.recordings?.length || 0,
+        downloadedFilesCount: result.downloadedFiles?.length || 0,
       });
 
       // Create tool message with the result
@@ -121,6 +122,7 @@ export const addBrowserTaskRunNode = async ({ graph, browserTaskNode, attachedTo
           steps: result.steps || [],
           recordings: result.recordings || [],
           inputParams: inputParams,
+          downloadedFiles: result.downloadedFiles || [],
         }),
         name: 'browser-task',
         tool_call_id: toolCall.id,
@@ -133,6 +135,7 @@ export const addBrowserTaskRunNode = async ({ graph, browserTaskNode, attachedTo
             steps: result.steps,
             recordings: result.recordings,
             inputParams: inputParams,
+            downloadedFiles: result.downloadedFiles || [],
           },
         },
       });
@@ -149,6 +152,7 @@ export const addBrowserTaskRunNode = async ({ graph, browserTaskNode, attachedTo
             nodeDisplayName: browserTaskNode.displayName,
             steps: result.steps,
             recordings: result.recordings,
+            downloadedFiles: result.downloadedFiles || [],
             status: 'completed',
           },
           update: true, // This triggers the message reducer to update the existing message

package/src/nodes/addPromptNode.ts

@@ -51,7 +51,7 @@ export const addPromptNode = async ({ graph, node, llm, tools, emit, agent }: Ad
     if (combinedPlaybooks) {
       finalMessage = combinedPlaybooks + '\n\n' + currentPromptNode;
     }
-    const compiledPrompt = compilePrompt(finalMessage, state);
+    const compiledPrompt = compilePrompt(finalMessage, { state: state, memory: state.memory, env: process.env });
     const systemMessage = new SystemMessage(compiledPrompt);
     if (state.messages.length === 0 || state.messages[0].getType() === 'system') {
       state.messages[0] = systemMessage;

package/src/nodes/addToolNode.ts

@@ -52,7 +52,7 @@ export const addToolNode = async ({
       }
     }
     if (finalMessage) {
-      const compiledPrompt = compilePrompt(finalMessage, state);
+      const compiledPrompt = compilePrompt(finalMessage, { state: state, memory: state.memory, env: process.env });
       const systemMessage = new SystemMessage(compiledPrompt);
       if (state.messages.length === 0 || state.messages[0].getType() === 'system') {
         state.messages[0] = systemMessage;

package/src/nodes/compilePrompt.ts

@@ -1,16 +1,13 @@
 import * as ejs from 'ejs';
 import { logger } from '../utils/logger';
-import { stateAnnotation } from '../types/LangGraph.types';
+
 /**
  * Compile prompt with parameters using EJS and placeholder replacement
  */
-export function compilePrompt(prompt: string, state: typeof stateAnnotation.State, params: Record<string, any> = {}): string {
+export function compilePrompt(prompt: string, params: Record<string, any> = {}): string {
   try {
-    // Always-included params
+    // Define system parameters
     params.system = { currentTime: new Date().toISOString() };
-    params.env = process.env;
-    params.state = state;
-    params.memory = state.memory;
 
     // First, render with EJS
     let compiledPrompt = ejs.render(prompt, params);

package/src/nodes/rpaStepsExecutor.ts

@@ -15,11 +15,17 @@ export async function executeRpaStep(
   state: typeof stateAnnotation.State,
   llm: (typeof LLMProviders)[keyof typeof LLMProviders],
 ): Promise<any> {
+  // Create params object with memory and secrets
+  const params = {
+    env: process.env,
+    state: state,
+  };
+
   switch (step.type) {
     case RpaActionType.CLICK: {
       // Compile selectors in case they contain placeholders
-      const compiledXpath = step.xpath ? compilePrompt(step.xpath, state) : undefined;
-      const compiledSelector = step.selector ? compilePrompt(step.selector, state) : undefined;
+      const compiledXpath = step.xpath ? compilePrompt(step.xpath, params) : undefined;
+      const compiledSelector = step.selector ? compilePrompt(step.selector, params) : undefined;
 
       if (compiledXpath) {
         await page.locator(`xpath=${compiledXpath}`).click({ timeout: 15000 });
@@ -31,9 +37,9 @@ export async function executeRpaStep(
 
     case RpaActionType.TYPE: {
       // Compile text and selectors
-      const compiledText = step.text ? compilePrompt(step.text, state) : '';
-      const compiledTypeXpath = step.xpath ? compilePrompt(step.xpath, state) : undefined;
-      const compiledTypeSelector = step.selector ? compilePrompt(step.selector, state) : undefined;
+      const compiledText = step.text ? compilePrompt(step.text, params) : '';
+      const compiledTypeXpath = step.xpath ? compilePrompt(step.xpath, params) : undefined;
+      const compiledTypeSelector = step.selector ? compilePrompt(step.selector, params) : undefined;
 
       if (step.shouldReplaceExistingText) {
         if (compiledTypeXpath) {
@@ -57,23 +63,23 @@ export async function executeRpaStep(
 
     case RpaActionType.GOTO: {
       // Compile URL
-      const compiledUrl = step.url ? compilePrompt(step.url, state) : '';
+      const compiledUrl = step.url ? compilePrompt(step.url, params) : '';
       await page.goto(compiledUrl, { waitUntil: 'load', timeout: 15000 });
       return { action: 'navigated', url: compiledUrl };
     }
 
     case RpaActionType.PRESS: {
       // Compile key
-      const compiledKey = step.key ? compilePrompt(step.key, state) : 'Enter';
+      const compiledKey = step.key ? compilePrompt(step.key, params) : 'Enter';
       await page.keyboard.press(compiledKey);
       return { action: 'pressed', key: compiledKey };
     }
 
     case RpaActionType.SELECT: {
       // Compile value and selectors
-      const compiledValue = step.value ? compilePrompt(step.value, state) : '';
-      const compiledSelectXpath = step.xpath ? compilePrompt(step.xpath, state) : undefined;
-      const compiledSelectSelector = step.selector ? compilePrompt(step.selector, state) : undefined;
+      const compiledValue = step.value ? compilePrompt(step.value, params) : '';
+      const compiledSelectXpath = step.xpath ? compilePrompt(step.xpath, params) : undefined;
+      const compiledSelectSelector = step.selector ? compilePrompt(step.selector, params) : undefined;
 
       if (compiledSelectXpath) {
         await page.locator(`xpath=${compiledSelectXpath}`).selectOption(compiledValue, { timeout: 15000 });
@@ -86,7 +92,7 @@ export async function executeRpaStep(
     case RpaActionType.SCREENSHOT: {
       const screenshot = await page.screenshot({ type: 'png' });
       // Compile description if provided
-      const compiledDescription = step.description ? compilePrompt(step.description, state) : undefined;
+      const compiledDescription = step.description ? compilePrompt(step.description, params) : undefined;
       return {
         action: 'screenshot',
         description: compiledDescription,

package/src/platform/mindedConnectionTypes.ts

@@ -471,6 +471,7 @@ export interface InvokeBrowserTaskResponse extends BaseSdkConnectionSocketMessag
   result?: any;
   steps?: any[];
   recordings?: any[];
+  downloadedFiles?: string[];
 }
 
 export interface ExecuteToolRequest extends BasemindedConnectionSocketMessage {