minovative-mind-cli 2.8.3 → 2.9.0

package/README.md

@@ -88,7 +88,8 @@ Hot-swap during a session using `/models`:
 
 | Model                     | Best for                                            |
 | ------------------------- | --------------------------------------------------- |
-| **Auto** (default)        | Automatically selects (3.6 Flash or 3.5 Flash Lite) |
+| **Auto** (default)        | Automatically selects (3.7 Flash or 3.5 Flash Lite) |
+| **Gemini 3.7 Flash**      | Next-gen performance, reasoning & fast execution   |
 | **Gemini 3.6 Flash**      | Everyday coding — fast and accurate                 |
 | **Gemini 3.1 Pro**        | Complex architectural changes                       |
 | **Gemini 3.5 Flash Lite** | Best for speed and cost efficiency                  |

package/dist/services/agent/slashCommands.js

@@ -9,9 +9,11 @@ import { toggleDebugMode, isDebugOn } from '../../utils/logger.js';
 import { changeLogger } from '../changeLogger.js';
 import { chatHistoryService } from '../chatHistoryService.js';
 import { printLogo, brandBg, brandFg } from '../../utils/logo.js';
+import { renderTerminalMarkdown } from '../../utils/terminal.js';
 import { readPaste } from '../../utils/paste.js';
 import { setApprovalMode, getApprovalMode, isSubAgentsEnabled, setSubAgentsEnabled } from '../agent-tools.js';
 import { ProxyClient, getAndResetTurnUsage } from '../proxyClient.js';
+import { checkByokSubscription } from '../auth.js';
 import { GEMINI_MODELS, isByokEnabled } from '../../utils/config.js';
 import { loadCredentials, updateCredentialField } from '../../utils/credentialStore.js';
 import { getGlobalActiveModel, setGlobalActiveModel, ProxyChatSession } from '../ai.js';
@@ -24,7 +26,7 @@ import { getGlobalActiveModel, setGlobalActiveModel, ProxyChatSession } from '..
  * - `/paste`       : Enter multi-line paste mode using EOF tracking (`Ctrl+D` submission).
  * - `/plan`        : Toggle AI step-by-step implementation planning mode.
  * - `/clear`       : Clear conversation history, wipe terminal screen, and reset CLI header.
- * - `/models`      : Hot-swap the active generative AI model (Gemini 3.1 Pro, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Auto routing).
+ * - `/models`      : Hot-swap the active generative AI model (Gemini 3.1 Pro, Gemini 3.7 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Auto routing).
  * - `/debug`       : Toggle internal agent telemetry and diagnostic logging.
  * - `/auto-approve`: Toggle automatic confirmation skipping for terminal shell execution commands.
  * - `/sub-agents`  : Toggle MMAAK Engine for parallel investigation sub-agent orchestration.
@@ -119,10 +121,15 @@ export async function handleSlashCommand(command, context) {
             //   label: 'Claude 5 Sonnet',
             //   hint: 'Best for raw speed and cost efficiency',
             // },
+            {
+                value: 'gemini-3.7-flash',
+                label: 'Gemini 3.7 Flash',
+                hint: 'Next-gen reasoning, speed & balanced performance',
+            },
             {
                 value: 'gemini-3.6-flash',
                 label: 'Gemini 3.6 Flash',
-                hint: 'Balanced performance & fast',
+                hint: 'Everyday coding — fast and accurate',
             },
             {
                 value: 'gemini-3.5-flash-lite',
@@ -132,9 +139,9 @@ export async function handleSlashCommand(command, context) {
             {
                 value: 'auto',
                 label: 'Auto (Flash-Lite / Flash)',
-                hint: 'Dynamically routes between Gemini 3.5 Flash-Lite and Gemini 3.6 based on prompt complexity',
+                hint: 'Dynamically routes between Gemini 3.5 Flash-Lite and Gemini 3.7 Flash based on prompt complexity',
             },
-        ].filter(o => !(byokEnabled && o.value.includes('claude')));
+        ].filter((o) => !(byokEnabled && o.value.includes('claude')));
         const selectedModel = await p['select']({
             message: `Select AI Model (Current: ${pc.cyan(currentModel)})`,
             initialValue: currentModel,
@@ -622,9 +629,7 @@ export async function handleSlashCommand(command, context) {
                             .join('');
                         if (textParts.trim()) {
                             console.log(`\n${pc.blue('◆')} ${pc.bold('Minovative Mind')} ${pc.dim('(History)')}\n`);
-                            const { marked } = await import('marked');
-                            const cleanText = textParts.replace(/\n([ \t]*\n){2,}/g, '\n\n');
-                            console.log(marked.parse(cleanText));
+                            console.log(renderTerminalMarkdown(textParts));
                         }
                     }
                 }
@@ -1047,12 +1052,29 @@ ${diffOut}
             if (!byokEnabled && !creds.geminiApiKey) {
                 p.log.error('You must set an API key before enabling BYOK mode.');
             }
+            else if (!byokEnabled) {
+                const subCheck = await checkByokSubscription();
+                if (!subCheck.active) {
+                    p.log.error(subCheck.message ||
+                        'A $3.99/month BYOK Subscription is required. Visit https://www.minovativemind.dev/pricing');
+                }
+                else {
+                    await updateCredentialField('useByok', true);
+                    p.log.success(`BYOK mode is now ${pc.green('Enabled')}`);
+                }
+            }
             else {
-                await updateCredentialField('useByok', !byokEnabled);
-                p.log.success(`BYOK mode is now ${!byokEnabled ? pc.green('Enabled') : pc.yellow('Disabled')}`);
+                await updateCredentialField('useByok', false);
+                p.log.success(`BYOK mode is now ${pc.yellow('Disabled')}`);
             }
         }
         else if (action === 'set') {
+            const subCheck = await checkByokSubscription();
+            if (!subCheck.active) {
+                p.log.error(subCheck.message ||
+                    'A $3.99/month BYOK Subscription is required to use your own API key. Please visit https://www.minovativemind.dev/pricing to subscribe. The $3.99 is to cover account maintance for you.');
+                return { shouldContinue: true };
+            }
             const key = await p['password']({
                 message: 'Enter your Google AI Studio API Key:',
                 validate: (v) => (!v ? 'API key is required' : undefined),

package/dist/services/agent/syntaxAgent.d.ts

@@ -8,12 +8,14 @@
  * @param error Optional error message from local validation to guide the AI.
  * @returns The fixed content, or undefined if it was completely valid or couldn't be fixed.
  */
-export declare function validateAndFixSyntax(content: string, filePath: string, error?: string): Promise<string | undefined>;
+export declare function validateAndFixSyntax(content: string, filePath: string, error?: string, abortSignal?: AbortSignal): Promise<string | undefined>;
 export interface AIFuzzyMatchOptions {
     /** Optional custom Gemini model override. Defaults to GEMINI_MODELS.FLASH. */
     model?: string;
     /** Whether to validate and fix syntax on the updated content. Defaults to true. */
     validateSyntax?: boolean;
+    /** Optional AbortSignal to cancel execution immediately. */
+    abortSignal?: AbortSignal;
 }
 export interface AIFuzzyMatchResult {
     /** Indicates whether the fuzzy search matching and edit application succeeded. */

package/dist/services/agent/syntaxAgent.js

@@ -68,7 +68,9 @@ function cleanOutput(raw) {
  * @param error Optional error message from local validation to guide the AI.
  * @returns The fixed content, or undefined if it was completely valid or couldn't be fixed.
  */
-export async function validateAndFixSyntax(content, filePath, error) {
+export async function validateAndFixSyntax(content, filePath, error, abortSignal) {
+    if (abortSignal?.aborted)
+        return undefined;
     // 1. First run fast local validation to check if content is already valid or gather error details
     const localResult = localValidate(filePath, content);
     if (localResult.isValid && !error) {
@@ -108,7 +110,7 @@ ${win.snippet}
 
 Please fix the syntax error in the snippet above and return ONLY the raw repaired code snippet:`;
             try {
-                const result = await snippetChat.sendMessage(snippetPrompt);
+                const result = await snippetChat.sendMessage(snippetPrompt, undefined, abortSignal);
                 const repairedSnippet = cleanOutput(result.response.text());
                 if (repairedSnippet && repairedSnippet !== win.snippet) {
                     const candidateContent = content.slice(0, win.startPos) + repairedSnippet + content.slice(win.endPos);
@@ -123,6 +125,8 @@ Please fix the syntax error in the snippet above and return ONLY the raw repaire
             }
         }
     }
+    if (abortSignal?.aborted)
+        return undefined;
     // 3. Full file repair strategy (for smaller files or when snippet repair was insufficient)
     const fullSystemInstruction = `
 <identity>
@@ -150,7 +154,7 @@ ${effectiveError}
 Content:
 ${content}`;
     try {
-        const result = await fullChat.sendMessage(fullPrompt);
+        const result = await fullChat.sendMessage(fullPrompt, undefined, abortSignal);
         const output = cleanOutput(result.response.text());
         if (output === 'VALID' || output === '' || output === content) {
             return undefined;
@@ -185,6 +189,12 @@ export async function aiFuzzyMatch(fileContent, searchContent, replaceContent, f
             error: 'File content and search content must not be empty.',
         };
     }
+    if (options.abortSignal?.aborted) {
+        return {
+            success: false,
+            error: 'Operation aborted',
+        };
+    }
     const model = options.model || GEMINI_MODELS.FLASH;
     const shouldValidateSyntax = options.validateSyntax !== false;
     const systemInstruction = `
@@ -217,7 +227,7 @@ ${replaceContent}
 File Content:
 ${fileContent}`;
     try {
-        const response = await chat.sendMessage(prompt);
+        const response = await chat.sendMessage(prompt, undefined, options.abortSignal);
         const rawOutput = response.response.text();
         const output = cleanOutput(rawOutput);
         if (output === 'UNMATCHED' || !output || output === fileContent) {
@@ -231,7 +241,7 @@ ${fileContent}`;
             const valResult = localValidate(filePath, finalContent);
             if (!valResult.isValid) {
                 // Attempt automatic syntax repair on AI output if local validation detected syntax error
-                const repaired = await validateAndFixSyntax(finalContent, filePath, valResult.error);
+                const repaired = await validateAndFixSyntax(finalContent, filePath, valResult.error, options.abortSignal);
                 if (repaired) {
                     finalContent = repaired;
                 }

package/dist/services/agent/toolLoop.js

@@ -264,7 +264,7 @@ export async function processResponse(chat, result, workspaceRoot, inputHandler,
             additionalText = `[USER INTERRUPTION] The user sent the following message during your execution:\n"${queuedMsg}"\n\nPlease incorporate this feedback into your ongoing work. Address the user's message, but DO NOT lose track of your original overall plan or focus. After addressing this interruption, continue with your broader objective.`;
             p.log.info(pc.cyan(`Sending queued message to AI...`));
             // Dynamically upgrade agent permissions/intent to EXECUTE mode if the interrupted instruction requires system modifications
-            const newIntent = await routeIntent(queuedMsg);
+            const newIntent = await routeIntent(queuedMsg, '', abortSignal);
             if (agentState.targetAgent === 'CHAT') {
                 if (newIntent.targetAgent === 'EXECUTE' || newIntent.needsContext) {
                     agentState.targetAgent = 'EXECUTE';

package/dist/services/agent.d.ts

@@ -125,4 +125,4 @@ export declare const HISTORY_SUMMARIZATION_THRESHOLD = 50;
  * @param chat - The active ProxyChatSession instance.
  * @returns A promise resolving to true if history was summarized and updated; false otherwise.
  */
-export declare function summarizeHistoryIfNeeded(chat: any): Promise<boolean>;
+export declare function summarizeHistoryIfNeeded(chat: any, abortSignal?: AbortSignal): Promise<boolean>;

package/dist/services/agent.js

@@ -21,8 +21,7 @@ import path from 'node:path';
 import * as crypto from 'node:crypto';
 import { exec } from 'node:child_process';
 import { promisify } from 'node:util';
-import { marked } from 'marked';
-import { markedTerminal } from 'marked-terminal';
+import { renderTerminalMarkdown } from '../utils/terminal.js';
 const execAsync = promisify(exec);
 import { debugLog, isDebugOn } from '../utils/logger.js';
 import { ensureProjectStorage, ensureIgnored, readCache, writeCache, invalidateCacheForDependents, } from '../utils/projectStorage.js';
@@ -44,7 +43,6 @@ import { getMetricCollector } from './metrics.js';
 import { Orchestrator } from './orchestration/orchestrator.js';
 import { isSubAgentsEnabled, getApprovalMode } from './agent-tools.js';
 import { runWithAgentId } from '../utils/asyncContext.js';
-marked.use(markedTerminal({ reflowText: false }));
 // Export submodules for potential external uses if required
 export { AsyncInputHandler } from './agent/inputHandler.js';
 /**
@@ -317,15 +315,19 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
             inputHandler.start(spinner);
             changeLogger.startChangeSet(userInput);
             let finalInput = userInput;
+            if (ac.signal.aborted)
+                return;
             // Inter-turn cooling-off delay to allow API token buckets to settle before running intent routing
             await new Promise((resolve) => setTimeout(resolve, TPM_COOLING_DELAYS.INTER_TURN_MS));
+            if (ac.signal.aborted)
+                return;
             // Stage 0: Summarize chat history if length threshold is met
             if (!cachedContextResult) {
                 const isDebug = isDebugOn();
                 if (!inputHandler.isCurrentlyPrompting() && !isDebug) {
                     spinner.start('Checking conversation history...');
                 }
-                const historySummarized = await summarizeHistoryIfNeeded(chat);
+                const historySummarized = await summarizeHistoryIfNeeded(chat, ac.signal);
                 if (historySummarized) {
                     debugLog('Chat history was summarized and compressed prior to context investigation.');
                 }
@@ -333,6 +335,8 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                     spinner.stop();
                 }
             }
+            if (ac.signal.aborted)
+                return;
             // Stage 1: Gather Workspace Context and route intentions
             spinner.start('🔍 Investigating workspace...');
             let gatherRes = { targetAgent: 'EXECUTE', chainedMessages: [], contextResult: cachedContextResult || null };
@@ -342,19 +346,34 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                     collector.startTimer('contextGather');
                 const chatHistory = chat.getRecentHistory(3);
                 const toolLogs = [];
-                gatherRes = await gatherContext(workspaceRoot, userInput, chatHistory, inputHandler, ac.signal, (msg) => {
-                    if (!inputHandler.isCurrentlyPrompting()) {
-                        spinner.message(`🔍 Investigating workspace... ${pc.dim(msg)}`);
-                    }
-                }, (toolMsg, label) => {
-                    const formattedLog = `${pc.dim(`[${label}]`)} ${toolMsg}`;
-                    toolLogs.push(formattedLog);
-                    if (!inputHandler.isCurrentlyPrompting()) {
-                        spinner.message(`🔍 Investigating workspace... ${formattedLog}`);
+                try {
+                    gatherRes = await gatherContext(workspaceRoot, userInput, chatHistory, inputHandler, ac.signal, (msg) => {
+                        if (!inputHandler.isCurrentlyPrompting()) {
+                            spinner.message(`🔍 Investigating workspace... ${pc.dim(msg)}`);
+                        }
+                    }, (toolMsg, label) => {
+                        const formattedLog = `${pc.dim(`[${label}]`)} ${toolMsg}`;
+                        toolLogs.push(formattedLog);
+                        if (!inputHandler.isCurrentlyPrompting()) {
+                            spinner.message(`🔍 Investigating workspace... ${formattedLog}`);
+                        }
+                    });
+                }
+                catch (gatherErr) {
+                    spinner.stop();
+                    process.stdout.write('\x1b[2K\r');
+                    if (ac.signal.aborted || gatherErr?.name === 'AbortError' || gatherErr?.message?.includes('abort')) {
+                        return;
                     }
-                });
+                    throw gatherErr;
+                }
                 if (collector)
                     collector.stopTimer('contextGather');
+                if (ac.signal.aborted) {
+                    spinner.stop();
+                    process.stdout.write('\x1b[2K\r');
+                    return;
+                }
                 // Print all collected logs at once to prevent flickering, while spinner is stopped
                 if (toolLogs.length > 0) {
                     spinner.stop();
@@ -363,6 +382,11 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                 }
                 // Post-investigation cooling-off delay to allow API Tokens-Per-Minute (TPM) sliding window to settle
                 await new Promise((resolve) => setTimeout(resolve, TPM_COOLING_DELAYS.POST_INVESTIGATION_MS));
+                if (ac.signal.aborted) {
+                    spinner.stop();
+                    process.stdout.write('\x1b[2K\r');
+                    return;
+                }
             }
             let latestUsage = undefined;
             // Collect any inputs that were queued while the Context Agent was investigating
@@ -374,7 +398,7 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
             let effectiveTargetAgent = gatherRes.targetAgent;
             if (gatherRes.chainedMessages.length > 0) {
                 const chainedContent = gatherRes.chainedMessages.join('\n');
-                const newIntent = await routeIntent(chainedContent);
+                const newIntent = await routeIntent(chainedContent, '', ac.signal);
                 if (gatherRes.contextResult !== null) {
                     // If the model previously needed search context, determine if followups change the mode
                     effectiveTargetAgent = newIntent.targetAgent;
@@ -394,6 +418,11 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                 }
                 finalInput += `\n\n[USER FOLLOW-UP INSTRUCTIONS SENT DURING INVESTIGATION]:\n${chainedContent}\n\nPlease incorporate these instructions into your work. Address them appropriately, but ensure you do not lose track of the original request's primary objective.`;
             }
+            if (ac.signal.aborted) {
+                spinner.stop();
+                process.stdout.write('\x1b[2K\r');
+                return;
+            }
             // Hot-swap the underlying LLM system instruction context depending on intent (conversational vs execution)
             debugLog(`Intent Router output: original targetAgent = ${gatherRes.targetAgent}, effective = ${effectiveTargetAgent}`);
             let config;
@@ -432,7 +461,7 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                 else {
                     spinner.stop(); // MUST clear the investigation spinner first to prevent leaking the setInterval
                     spinner.start(pc.blue('🧠 Evaluating execution complexity...'));
-                    const complexity = await evaluateExecutionComplexity(finalInput, gatherRes.contextResult?.summary, gatherRes.contextResult?.relevantFiles?.size || 0);
+                    const complexity = await evaluateExecutionComplexity(finalInput, gatherRes.contextResult?.summary, gatherRes.contextResult?.relevantFiles?.size || 0, '', ac.signal);
                     spinner.stop();
                     process.stdout.write('\x1b[2K\r');
                     selectedModel = complexity === 'EASY' ? GEMINI_MODELS.FLASH_LITE : GEMINI_MODELS.FLASH;
@@ -443,6 +472,11 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
             else {
                 chat.setModel(getGlobalActiveModel());
             }
+            if (ac.signal.aborted) {
+                spinner.stop();
+                process.stdout.write('\x1b[2K\r');
+                return;
+            }
             if (gatherRes.contextResult) {
                 if (!inputHandler.isCurrentlyPrompting()) {
                     spinner.stop(pc.green('🔍 Investigation complete.'));
@@ -513,8 +547,7 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                         });
                         // Print the summary text just like single-agent mode
                         console.log(`\n${pc.blue('◆')} ${pc.bold('Minovative Mind')} ${pc.dim(`(Orchestrator)`)}\n`);
-                        const cleanText = handledByOrchestrator.replace(/\n([ \t]*\n){2,}/g, '\n\n');
-                        console.log(marked.parse(cleanText));
+                        console.log(renderTerminalMarkdown(handledByOrchestrator));
                     }
                     // Print Usage Stats
                     const usage = getAndResetTurnUsage();
@@ -634,9 +667,7 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                 const cleanFinalText = finalText.replace(/\[TASK_FINISHED\]/g, '').trim();
                 if (cleanFinalText) {
                     console.log(`\n${pc.blue('◆')} ${pc.bold('Minovative Mind')} ${pc.dim(`(${chat.getModel()})`)}\n`);
-                    // Strip out excessive empty lines generated by LLMs to prevent huge visual gaps in marked-terminal
-                    const cleanText = cleanFinalText.replace(/\n([ \t]*\n){2,}/g, '\n\n');
-                    console.log(marked.parse(cleanText));
+                    console.log(renderTerminalMarkdown(cleanFinalText));
                 }
             }
             if (usage) {
@@ -715,7 +746,7 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                 if (!chatSessionState.title) {
                     chatSessionState.title = 'Generating title...';
                     try {
-                        const title = await generateChatTitle(userInput);
+                        const title = await generateChatTitle(userInput, ac.signal);
                         chatSessionState.title = title;
                         await chatHistoryService.saveSession({
                             id: chatSessionState.id,
@@ -804,10 +835,15 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                             //   label: 'Claude 5 Sonnet',
                             //   hint: 'Best for raw speed and cost efficiency',
                             // },
+                            {
+                                value: 'gemini-3.7-flash',
+                                label: 'Gemini 3.7 Flash',
+                                hint: 'Next-gen reasoning, speed & balanced performance',
+                            },
                             {
                                 value: 'gemini-3.6-flash',
                                 label: 'Gemini 3.6 Flash',
-                                hint: 'Balanced performance & fast',
+                                hint: 'Everyday coding — fast and accurate',
                             },
                             {
                                 value: 'gemini-3.5-flash-lite',
@@ -817,7 +853,7 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
                             {
                                 value: 'auto',
                                 label: 'Auto (Flash-Lite / Flash)',
-                                hint: 'Dynamically routes between Gemini 3.5 Flash-Lite and Gemini 3.6 based on prompt complexity',
+                                hint: 'Dynamically routes between Gemini 3.5 Flash-Lite and Gemini 3.7 Flash based on prompt complexity',
                             },
                         ].filter((o) => !(byokEnabled && o.value.includes('claude')));
                         const selectedModel = await p['select']({
@@ -847,6 +883,9 @@ export async function executeSingleTurn(workspaceRoot, userInput, chat, inputHan
             return { planModeReturn, contextResult: gatherRes.contextResult };
         }
         catch (err) {
+            if (ac.signal.aborted || err?.name === 'AbortError' || err?.message?.includes('abort')) {
+                return;
+            }
             const message = err instanceof Error ? err.message : String(err);
             p.log.error(`${pc.red('Error:')} ${message}`);
         }
@@ -960,7 +999,7 @@ export const HISTORY_SUMMARIZATION_THRESHOLD = 50;
  * @param chat - The active ProxyChatSession instance.
  * @returns A promise resolving to true if history was summarized and updated; false otherwise.
  */
-export async function summarizeHistoryIfNeeded(chat) {
+export async function summarizeHistoryIfNeeded(chat, abortSignal) {
     const history = chat.getRawHistory();
     if (!history || history.length < HISTORY_SUMMARIZATION_THRESHOLD) {
         return false;
@@ -971,7 +1010,7 @@ export async function summarizeHistoryIfNeeded(chat) {
         const olderHistory = history.slice(0, history.length - recentCount);
         const recentHistory = history.slice(history.length - recentCount);
         debugLog(`Summarizing ${olderHistory.length} older chat history entries...`);
-        const summaryText = await summarizeChatHistory(olderHistory);
+        const summaryText = await summarizeChatHistory(olderHistory, abortSignal);
         if (summaryText && summaryText.trim().length > 0) {
             const summaryContent = [
                 {
@@ -989,6 +1028,9 @@ export async function summarizeHistoryIfNeeded(chat) {
         }
     }
     catch (error) {
+        if (abortSignal?.aborted || error?.name === 'AbortError' || error?.message?.includes('abort')) {
+            return false;
+        }
         debugLog(`Failed to summarize chat history: ${error?.message || error}`);
     }
     return false;
@@ -1003,15 +1045,15 @@ async function executeSelfCorrectionLoop(chat, initialResult, workspaceRoot, inp
     let intentVerified = false;
     let isFixingCodeError = false;
     while (correctionAttempts <= MAX_CORRECTIONS) {
-        if (finalText === '[Generation stopped by user]')
+        if (signal.aborted || finalText === '[Generation stopped by user]')
             break;
         const historyLengthBefore = chat.getRawHistory().length;
         const currentText = await processResponse(chat, result, workspaceRoot, inputHandler, agentState, signal);
         const historyLengthAfter = chat.getRawHistory().length;
         const usedTools = historyLengthAfter > historyLengthBefore;
         debugLog(`processResponse returned text (length ${currentText.length}): "${currentText.substring(0, 10)}..."`);
-        if (currentText === '[Generation stopped by user]') {
-            finalText = currentText;
+        if (currentText === '[Generation stopped by user]' || signal.aborted) {
+            finalText = '[Generation stopped by user]';
             break;
         }
         if (currentText.startsWith('[The AI repeatedly returned empty responses')) {
@@ -1037,7 +1079,18 @@ async function executeSelfCorrectionLoop(chat, initialResult, workspaceRoot, inp
             debugLog('AI failed to execute tools or modify files during the correction attempt. Retrying...');
             const forcePrompt = `AUTOMATED SYSTEM CHECK: You did not execute any tools or modify files. You MUST use your tools (such as run_command or modify_file) to investigate and apply a fix for the previously mentioned errors. Do not just explain the issue.`;
             spinner.start('Thinking (Correction)...');
-            result = await chat.sendMessage(forcePrompt);
+            try {
+                result = await chat.sendMessage(forcePrompt, undefined, signal);
+            }
+            catch (e) {
+                spinner.stop();
+                process.stdout.write('\x1b[2K\r');
+                if (e.name === 'AbortError' || e.message?.includes('abort') || signal.aborted) {
+                    p.log.warn(pc.yellow('Generation stopped by user during correction.'));
+                    break;
+                }
+                throw e;
+            }
             spinner.stop();
             process.stdout.write('\x1b[2K\r');
             correctionAttempts++;
@@ -1062,7 +1115,7 @@ If you are not done, please continue working using your other tools.`;
                 catch (e) {
                     spinner.stop();
                     process.stdout.write('\x1b[2K\r');
-                    if (e.name === 'AbortError' || e.message?.includes('abort')) {
+                    if (e.name === 'AbortError' || e.message?.includes('abort') || signal.aborted) {
                         p.log.warn(pc.yellow('Generation stopped by user during verification.'));
                         break;
                     }
@@ -1135,7 +1188,18 @@ If you are not done, please continue working using your other tools.`;
         // Compile compilation and syntax diagnostic warnings into an auto-correction prompt
         const correctionPrompt = `AUTOMATED SYSTEM CHECK: Your previous changes resulted in the following issues:\n\n${combinedIssuesForAI}\n\nPlease analyze these issues and use your file modification tools to fix them.`;
         spinner.start('Thinking (Correction)...');
-        result = await chat.sendMessage(correctionPrompt);
+        try {
+            result = await chat.sendMessage(correctionPrompt, undefined, signal);
+        }
+        catch (e) {
+            spinner.stop();
+            process.stdout.write('\x1b[2K\r');
+            if (e.name === 'AbortError' || e.message?.includes('abort') || signal.aborted) {
+                p.log.warn(pc.yellow('Generation stopped by user during correction.'));
+                break;
+            }
+            throw e;
+        }
         spinner.stop();
         process.stdout.write('\x1b[2K\r');
     }

package/dist/services/ai.d.ts

@@ -80,7 +80,7 @@ export declare function getPlanModeConfig(): {
  * Compresses a large string of text using gemini-3.5-flash-lite.
  * Used for shrinking context payloads to prevent OOM/choking.
  */
-export declare function compressTextUsingFlashLite(text: string, instruction?: string, inlineData?: any, force?: boolean): Promise<string>;
+export declare function compressTextUsingFlashLite(text: string, instruction?: string, inlineData?: any, force?: boolean, abortSignal?: AbortSignal): Promise<string>;
 /**
  * Returns the tool declarations for the read-only Context Agent.
  * Extracted as a reusable function so investigation sub-agents can import
@@ -96,8 +96,8 @@ export declare function createHistorySummarizerSession(): any;
 /**
  * Summarizes an array of Content history entries using Gemini Flash Lite.
  */
-export declare function summarizeChatHistory(history: Content[]): Promise<string>;
+export declare function summarizeChatHistory(history: Content[], abortSignal?: AbortSignal): Promise<string>;
 /**
  * Generates a concise title for a chat session based on the user's first message.
  */
-export declare function generateChatTitle(firstMessage: string): Promise<string>;
+export declare function generateChatTitle(firstMessage: string, abortSignal?: AbortSignal): Promise<string>;