fluxflow-cli 1.8.17 → 1.8.19

package/dist/fluxflow.js

@@ -768,23 +768,21 @@ var thinking_prompts_default;
 var init_thinking_prompts = __esm({
   "src/data/thinking_prompts.json"() {
     thinking_prompts_default = {
-      Max: `-- START THINKING INSTRUCTIONS --
-EFFORT_LEVEL: MAX
+      Max: `EFFORT_LEVEL: MAX
 Think in a continuous, fluid analytical monologue within the <think>...</think> block. Do NOT use headings, bullet points, or artificial sections. Engage in a deep "Stream of Consciousness" that follows this cognitive path:
-1. **Deep Analysis**: Deconstruct the request into its core technical and logic requirements.
-2. **Hypothesis & Test**: Propose multiple solutions mentally and critique them for edge cases or security risks.
-3. **Architectural Planning**: Consider the long-term impact on the project structure and maintainability.
-4. **Refinement**: Iterate on the chosen path until it is bulletproof.
+Deep Analysis: Deconstruct the request into its core technical and logic requirements.
+Hypothesis & Test: Propose multiple solutions mentally and critique them for edge cases or security risks.
+Architectural Planning: Consider the long-term impact on the project structure and maintainability.
+Refinement: Iterate on the chosen path until it is bulletproof.
 
 RULES: 
 - NO HEADINGS. Just a solid, stable analytical monologue.
 - Be thorough and exhaustive. Explore the 'why' behind every decision.
 - Use internal critique: Question your own logic as you go.
-- **MANDATORY REASONING**: Unless the prompt is a simple greeting (e.g. 'hi') or trivial math (e.g. '1+1'), you MUST engage in full architectural reasoning regardless of perceived simplicity.
--- END THINKING INSTRUCTIONS --`,
-      High: "-- START THINKING INSTRUCTIONS --\nEFFORT_LEVEL: HIGH\nThink in a stable, analytical monologue within the <think>...</think> block. Avoid headings or structured formatting. Your thinking should be a continuous stream of logical deduction:\n1. Analyze the immediate task and its dependencies.\n2. Mentally simulate the execution to identify potential failure points.\n3. Structure a precise plan that addresses both primary goals and secondary constraints.\n\nRULES:\n- NO HEADINGS. Maintain a fluid monologue style.\n- Be detailed and rigorous in your self-questioning.\n- Focus on accuracy and technical correctness.\n- **MANDATORY REASONING**: For any request beyond basic conversation, you MUST enter reasoning to verify the path forward.\n-- END THINKING INSTRUCTIONS --",
-      Medium: "-- START THINKING INSTRUCTIONS --\nEFFORT_LEVEL: MEDIUM\nThink in a concise, stable monologue within the <think>...</think> block. No headings needed. Focus on the core logic required to solve the task efficiently:\n1. Identify the most direct path to the solution.\n2. Briefly consider and discard obvious alternatives.\n3. Confirm the plan meets the user's immediate requirements.\n\nRULES:\n- NO HEADINGS. Keep it as a simple, logical stream.\n- Be efficient. Spend energy only on what matters for the task.\n- **REQUIRED REASONING**: Engage in a baseline mental check for all technical tasks.\n-- END THINKING INSTRUCTIONS --",
-      Minimal: "-- START THINKING INSTRUCTIONS --\nEFFORT_LEVEL: LOW\nThink in a brief, focused monologue within the <think>...</think> block. No headings. Just a quick mental check before acting:\n1. Verify the objective.\n2. Note the target files/tools.\n\nRULES:\n- NO HEADINGS. Just a few lines of clear, linear thought.\n- Use minimal/no thinking for simple or conversational requests.\n-- END THINKING INSTRUCTIONS --"
+- MANDATORY REASONING: You MUST engage in full reasoning regardless of perceived simplicity.`,
+      High: "EFFORT_LEVEL: HIGH\nThink in a stable, analytical monologue within the <think>...</think> block. Avoid headings or structured formatting. Your thinking should be a continuous stream of logical deduction:\nAnalyze the immediate task and its dependencies.\nMentally simulate the execution to identify potential failure points.\nStructure a precise plan that addresses both primary goals and secondary constraints.\n\nRULES:\n- NO HEADINGS. Maintain a fluid monologue style.\n- Be detailed and rigorous in your self-questioning.\n- Focus on accuracy and technical correctness.\n- MANDATORY REASONING: You MUST enter reasoning to verify the path forward.",
+      Medium: "EFFORT_LEVEL: MEDIUM\nThink in a concise, stable monologue within the <think>...</think> block. No headings needed. Focus on the core logic required to solve the task efficiently:\nIdentify the most direct path to the solution.\nBriefly consider and discard obvious alternatives.\nConfirm the plan meets the user's immediate requirements.\n\nRULES:\n- NO HEADINGS. Keep it as a simple, logical stream.\n- Be efficient. Spend energy only on what matters for the task.\n- REQUIRED REASONING: Engage in a baseline mental check for all technical tasks.",
+      Minimal: "EFFORT_LEVEL: LOW\nThink in a brief, focused monologue within the <think>...</think> block. No headings. Just a quick mental check before acting:\nVerify the objective.\nNote the target files/tools.\n\nRULES:\n- NO HEADINGS. Just a few lines of clear, linear thought.\n- Use minimal/no thinking for simple or conversational requests."
     };
   }
 });
@@ -796,7 +794,7 @@ var init_prompts = __esm({
     init_main_tools();
     init_janitor_tools();
     init_thinking_prompts();
-    getSystemInstruction = (profile, thinkingLevel, mode, systemSettings, tempMemories = "", userMemories = "", isMemoryEnabled = true, isContext50 = false, maxLoops, currentLoop) => {
+    getSystemInstruction = (profile, thinkingLevel, mode, systemSettings, tempMemories = "", userMemories = "", isMemoryEnabled = true, isContext8 = false, maxLoops, currentLoop) => {
       let levelKey = thinkingLevel;
       if (thinkingLevel === "Low") levelKey = "Minimal";
       if (thinkingLevel === "xHigh" || thinkingLevel === "Max") levelKey = "Max";
@@ -810,7 +808,7 @@ var init_prompts = __esm({
 ` : "";
       const dateTimeStr = (/* @__PURE__ */ new Date()).toLocaleString();
       const cwdStr = process.cwd();
-      const tempMemoriesStr = tempMemories?.length > 0 && !isContext50 ? `
+      const tempMemoriesStr = tempMemories?.length > 0 && !isContext8 ? `
 -- RECENT CONTEXT FROM OTHER CHAT THREADS --
 ${tempMemories}
 ------------------------------------------
@@ -831,11 +829,13 @@ CURRENT_WORKING_DIRECTORY: ${cwdStr}.
 OS: ${osDetected}. ${osDetected === "Windows" && mode === "Flux" ? "Your terminal commands will run on CMD. 'Prefer using PS scripts via CMD' instead of raw CMD commands." : ""}
 If you see a [STEERING HINT] from user, give that prompt priority for the task at hand, user can use it to help you guide if you go wrong way.
 
-[Runtime Monitor] Turn Progress: ${currentLoop}/${maxLoops} steps. Aim to finalize the task before the window closes. If the limit is reached, please summarize and invite the user to re-engage.
-
 
+-- START THINKING INSTRUCTIONS --
 ${thinkingConfig}
 
+BEFORE USING ANY TOOL THINKING IS **MANDATORY** WITH TOOL RULES. ALWAYS PRIORITIZE THINKING FIRST BEFORE RESPONDING. YOU ARE **FORBIDDEN** TO JUMP TO RESPONSES FIRST.
+-- END THINKING INSTRUCTIONS --
+
 ${TOOL_PROTOCOL(mode)}
 ${mode === "Flux" ? `
 -- START PROJECT SPECIFIC INSTRUCTIONS --
@@ -863,15 +863,15 @@ Every ${isMemoryEnabled ? "Prompt, Responses & Memories" : "Prompt & Responses"}
 
 -- START FORMATTING RULES --
 - CRITICAL NEWLINE PROTOCOL: When writing or updating files, you MUST use actual line breaks (LF) for structural newlines. If you need to write the literal characters '\\' and 'n' (e.g., in printf("Hello\\n")), you MUST use the sequence '[/n]'.
-  [CORRECT]:
-  tool:functions.write_file(path="test.c", content="#include <stdio.h>
-  int main() {
+[CORRECT]:
+tool:functions.write_file(path="test.c", content="#include <stdio.h>
+int main() {
     printf("Hello[/n]");
     return 0;
-  }")
-  [INCORRECT]:
-  tool:functions.write_file(path="test.c", content="#include <stdio.h>\\nint main() {\\nprintf("Hello\\\\n");\\n}")
-  \u{1F6D1} NEVER use '\\\\n' for literals; it will be converted to a real line break and break code syntax.
+    }")
+[INCORRECT]:
+tool:functions.write_file(path="test.c", content="#include <stdio.h>\\nint main() {\\nprintf("Hello\\\\n");\\n}")
+\u{1F6D1} NEVER use '\\\\n' for literals; it will be converted to a real line break and break code syntax.
 - Structure responses VISUALLY pleasing, easy to read, and beautiful.
 - USE GFM Markdown HEAVILY.
 - Use GFM tables for structured data to keep the terminal view organized. KEEP SENTENCES IN TABLE **SHORT & CONCISE**. AND MAX 4 COLUMNS. DO NOT OVERUSE TABLES.
@@ -887,8 +887,9 @@ TO END THE LOOP YOU **MUST** WRITE [turn: finish] AT VERY END OF YOUR RESPONSE.
 When you 'finish' an agentic loop, you will lose your previous turn 'thinking' data. So only write [turn: finish] when you are absolutely sure that you are done with the task. Or user has to prompt again and re-thinking again from scratch will use tokens that were already planned.
 -- END REPONSE FINISH PROTOCOL --
 
+[Runtime Monitor] Turn Progress: ${currentLoop}/${maxLoops} steps. Aim to finalize the task before the window closes. If the limit is reached, please summarize and invite the user to re-engage.
 Current date and Time is: ${dateTimeStr}
- --- END SYSTEM INSTRUCTION ---`.trim();
+--- END SYSTEM INSTRUCTION ---`.trim();
     };
     getJanitorInstruction = (originalText, agentRaws, userMemories = "", isMemoryEnabled = true, needTitle = false) => {
       let agentRes = `${agentRaws.replace(/tool:functions\..*\n/g, "").replace(/<think>.*<\/think>/g, "").replace(/\[Prompted on:.*?\]/g, "").replace(/\[turn: continue\]/g, "").replace(/\[turn: finish\]/g, "").replace(/\[TOOL_RESULTS\]/g, "").replace(/\[tool_results\]/g, "").substring(0, 3500)}`;
@@ -1487,7 +1488,7 @@ var init_memory = __esm({
         if (!content) return "ERROR: Missing 'content' for temp memory.";
         const tempStorage = readEncryptedJson(TEMP_MEM_FILE, {});
         if (!tempStorage[chatId]) tempStorage[chatId] = [];
-        const MAX_CHARS = 2500 * 4;
+        const MAX_CHARS = 2e3 * 4;
         let currentTotalLength = tempStorage[chatId].reduce((acc, m) => acc + m.length, 0);
         while (tempStorage[chatId].length > 0 && currentTotalLength + content.length > MAX_CHARS) {
           const removed = tempStorage[chatId].shift();
@@ -2664,7 +2665,7 @@ var init_ai = __esm({
       modifiedHistory.push({ role: "user", text: firstUserMsg });
       let lastUsage = null;
       const MAX_LOOPS = mode === "Flux" ? 50 : 7;
-      const MAX_RETRIES = 7;
+      const MAX_RETRIES = 8;
       yield { type: "status", content: "Connecting..." };
       TERMINATION_SIGNAL = false;
       let fullAgentResponseChunks = [];
@@ -2697,16 +2698,18 @@ var init_ai = __esm({
         }
         let stream;
         let success = false;
-        let retryCount = 0;
+        let retryCount = 1;
+        let inStreamRetryCount = 1;
         let turnText = "";
         let lastToolSniffed = null;
         let lastToolEventTime = null;
         let toolResults = [];
         let toolCallPointer = 0;
         let isThinkingLoop = false;
+        let isStutteringLoop = false;
         let isInitialAttempt = true;
         let accumulatedContext = "";
-        while (retryCount <= MAX_RETRIES && !success && !TERMINATION_SIGNAL) {
+        while (retryCount <= MAX_RETRIES && inStreamRetryCount <= MAX_RETRIES && !success && !TERMINATION_SIGNAL) {
           try {
             if (isInitialAttempt) {
               yield { type: "turn_reset", content: true };
@@ -2728,48 +2731,32 @@ var init_ai = __esm({
               throw new Error("Error: Daily Quota Exausted for Agent");
             }
             let targetModel = modelName;
-            if (retryCount === 5) {
+            if (retryCount === 6) {
               targetModel = "gemini-3-flash-preview";
               yield { type: "model_update", content: "Trying with fallback model" };
-            } else if (retryCount >= 6) {
+            } else if (retryCount >= 7) {
               targetModel = "gemini-3.1-flash-lite-preview";
               yield { type: "model_update", content: "Trying with fallback model lite" };
             } else if (retryCount > 0) {
               yield { type: "model_update", content: null };
             }
-            const isContext50 = (sessionStats.tokens || 0) >= 54e3;
-            const currentSystemInstruction = getSystemInstruction(profile, thinkingLevel, mode, systemSettings, otherMemories, mainUserMemories, isMemoryEnabled, isContext50, MAX_LOOPS, loop + 1);
+            const isContext8 = (sessionStats.tokens || 0) >= 8e3;
+            const currentSystemInstruction = getSystemInstruction(profile, thinkingLevel, mode, systemSettings, otherMemories, mainUserMemories, isMemoryEnabled, isContext8, MAX_LOOPS, loop + 1);
             stream = await client.models.generateContentStream({
               model: targetModel || "gemma-4-31b-it",
               contents,
               config: {
                 systemInstruction: currentSystemInstruction,
-                temperature: mode === "Flux" ? 1 : 1.3,
+                temperature: mode === "Flux" ? 0.99 : 1.4,
                 maxOutputTokens: 32768,
                 mediaResolution: "MEDIA_RESOLUTION_MEDIUM",
                 safetySettings: [
-                  {
-                    category: HarmCategory.HARM_CATEGORY_HARASSMENT,
-                    threshold: HarmBlockThreshold.BLOCK_NONE
-                  },
-                  {
-                    category: HarmCategory.HARM_CATEGORY_HATE_SPEECH,
-                    threshold: HarmBlockThreshold.BLOCK_NONE
-                  },
-                  {
-                    category: HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
-                    threshold: HarmBlockThreshold.BLOCK_NONE
-                  },
-                  {
-                    category: HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
-                    threshold: HarmBlockThreshold.BLOCK_NONE
-                  }
+                  { category: HarmCategory.HARM_CATEGORY_HARASSMENT, threshold: HarmBlockThreshold.BLOCK_NONE },
+                  { category: HarmCategory.HARM_CATEGORY_HATE_SPEECH, threshold: HarmBlockThreshold.BLOCK_NONE },
+                  { category: HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT, threshold: HarmBlockThreshold.BLOCK_NONE },
+                  { category: HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT, threshold: HarmBlockThreshold.BLOCK_NONE }
                 ],
-                thinkingConfig: {
-                  includeThoughts: false,
-                  thinkingLevel: ThinkingLevel.MINIMAL
-                  // Gemma's API Reasoning is bad. Keep it Minimal.
-                }
+                thinkingConfig: { includeThoughts: false, thinkingLevel: targetModel.includes("pro") ? ThinkingLevel.HIGH : ThinkingLevel.MINIMAL }
               }
             });
             turnText = "";
@@ -2784,7 +2771,7 @@ var init_ai = __esm({
             for await (const chunk of stream) {
               if (TERMINATION_SIGNAL) {
                 yield { type: "status", content: "Termination Signal Received." };
-                await new Promise((resolve) => setTimeout(resolve, 1500));
+                await new Promise((resolve) => setTimeout(resolve, 3e3));
                 break;
               }
               if (chunk.text) {
@@ -2831,9 +2818,16 @@ var init_ai = __esm({
                 const wordCount = thinkContent.split(/\s+/).filter((w) => w.length > 0).length;
                 let repetitionThresholdThinking = 0.4;
                 let repetitionThresholdResponse = 0.6;
-                let isOverVerboseThinking = wordCount > 2500;
+                const thinkingCaps = {
+                  "low": 200,
+                  "medium": 500,
+                  "high": 2e3,
+                  "max": 3500
+                };
+                const cap = thinkingCaps[thinkingLevel?.toLowerCase()] || 2500;
+                let isOverVerboseThinking = wordCount > cap;
                 if (repetitionRatio > repetitionThresholdThinking || isOverVerboseThinking) {
-                  const reason = repetitionRatio > repetitionThresholdThinking ? "Thinking Loop Detected" : "Rambling Detected";
+                  const reason = repetitionRatio > repetitionThresholdThinking ? "Thinking Loop Detected" : "Thinking Budget Exceeded";
                   yield { type: "status", content: `${reason}. Re-centering...` };
                   isThinkingLoop = true;
                   await new Promise((resolve) => setTimeout(resolve, 3e3));
@@ -2863,6 +2857,7 @@ var init_ai = __esm({
                   if (stutterDetected) {
                     yield { type: "status", content: `Stuttering Detected. Re-centering...` };
                     isThinkingLoop = false;
+                    isStutteringLoop = true;
                     await new Promise((resolve) => setTimeout(resolve, 3e3));
                     break;
                   }
@@ -3066,24 +3061,33 @@ ${boxBottom}
 ----------------------------------------------------------------------
 
 `);
-            if (retryCount < MAX_RETRIES) {
-              retryCount++;
-              const waitTime = Math.floor(Math.random() * (2e3 - 800 + 1)) + 800;
-              if (turnText.trim().length > 0) {
+            if (turnText.trim().length > 0) {
+              if (inStreamRetryCount <= MAX_RETRIES) {
+                inStreamRetryCount++;
+                const waitTime = Math.floor(Math.random() * (3e3 - 1e3 + 1)) + 1e3;
                 modifiedHistory.push({ role: "agent", text: turnText });
                 if (toolResults.length > 0) {
                   toolResults.forEach((tr) => modifiedHistory.push(tr));
                 }
                 modifiedHistory.push({ role: "user", text: "[SYSTEM] Response got cut for internal error, continue from checkpoint seamlessly and DON'T repeat what you already said!" });
                 accumulatedContext += turnText;
-                yield { type: "status", content: `Recovering & Continuing (${retryCount}/${MAX_RETRIES + 1})...` };
+                yield { type: "status", content: `Error Occured. Recovering Stream (${inStreamRetryCount}/${MAX_RETRIES})...` };
+                await new Promise((resolve) => setTimeout(resolve, waitTime));
               } else {
-                isInitialAttempt = true;
-                yield { type: "status", content: `Retrying (${retryCount}/${MAX_RETRIES + 1})...` };
+                throw new Error(`Stream collapsed too many times. (Failed to resolve ${MAX_RETRIES} times)
+Error Log can be found in ${path16.join(LOGS_DIR, "agent", "error.log")}`);
               }
-              await new Promise((resolve) => setTimeout(resolve, waitTime));
             } else {
-              throw new Error(`Model cannot be reached: ${errMsg}. (Failed ${MAX_RETRIES + 1} times)`);
+              if (retryCount <= MAX_RETRIES) {
+                retryCount++;
+                const waitTime = Math.floor(Math.random() * (3e3 - 1e3 + 1)) + 1e3;
+                isInitialAttempt = true;
+                yield { type: "status", content: `Retrying Connection (${retryCount}/${MAX_RETRIES})...` };
+                await new Promise((resolve) => setTimeout(resolve, waitTime));
+              } else {
+                throw new Error(`Model cannot be reached. (Failed ${MAX_RETRIES} times)
+Error Log can be found in ${path16.join(LOGS_DIR, "agent", "error.log")}`);
+              }
             }
           }
         }
@@ -3242,8 +3246,9 @@ ${timestamp}`;
         if (toolResults.length > 0) {
           toolResults.forEach((tr) => modifiedHistory.push(tr));
         } else {
-          modifiedHistory.push({ role: "user", text: `[SYSTEM]: ${isThinkingLoop ? "OVER-THINKING " : ""}LOOP DETECTED by Internal System. ${isThinkingLoop ? "If you have planned the task, prioritize the execution/output. " : "If you have finished your task use [turn: finish] else continue."}` });
+          modifiedHistory.push({ role: "user", text: `[SYSTEM]: ${isStutteringLoop && !isThinkingLoop ? `STUTTERING DETECTED by Internal System. Re-calibrate your response & proceed.` : `${isThinkingLoop ? "OVER-THINKING " : ""}LOOP DETECTED by Internal System${isThinkingLoop && " for current EFFORT_LEVEL"}. ${isThinkingLoop ? "If you have planned the task, prioritize the execution/output. " : "If you have finished your task use [turn: finish] else continue."}`}` });
           isThinkingLoop = false;
+          isStutteringLoop = false;
         }
       }
       yield { type: "status", content: null };
@@ -3927,10 +3932,10 @@ Check what's new using \`/changelog\` command.`,
       cmd: "/model",
       desc: "Switch AI model",
       subs: [
-        { cmd: "gemma-4-31b-it", desc: apiTier === "Free" ? "Standard Default (Free, Recommended)" : "Standard Default (Free, Recommended) - Cannot use Gemma with paid API" },
+        { cmd: "gemma-4-31b-it", desc: apiTier === "Free" ? "Standard Default (Free, Recommended)" : "Standard Default (Free, Recommended) - Use Free API Key to use this model " },
         { cmd: "gemini-3.1-pro-preview", desc: "Most Capable (Paid)" },
-        { cmd: "gemini-3-flash-preview", desc: "Fast & Lightweight (Paid, Free limited quota)" },
-        { cmd: "gemini-3.1-flash-lite-preview", desc: "Ultra Fast (Paid, Free limited quota)" }
+        { cmd: "gemini-3-flash-preview", desc: "Fast & Lightweight (Paid, Limited Free quota)" },
+        { cmd: "gemini-3.1-flash-lite-preview", desc: "Ultra Fast (Paid, Decent Free quota)" }
       ]
     },
     { cmd: "/settings", desc: "Configure system prefs" },
@@ -4950,7 +4955,7 @@ Selection: ${val}`,
               setActiveView("chat");
               setTimeout(() => {
                 handleSubmit(val);
-              }, 50);
+              }, 200);
             },
             onEdit: (val) => {
               setResolutionData(null);
@@ -5197,7 +5202,7 @@ var init_app = __esm({
     init_text();
     SESSION_START_TIME = Date.now();
     CHANGELOG_URL = "https://fluxflow-cli.onrender.com/changelog.html";
-    versionFluxflow = "1.8.17";
+    versionFluxflow = "1.8.19";
     updatedOn = "2026-05-10";
     ResolutionModal = ({ data, onResolve, onEdit }) => /* @__PURE__ */ React10.createElement(Box10, { flexDirection: "column", borderStyle: "round", borderColor: "magenta", paddingX: 2, paddingY: 1, width: "100%" }, /* @__PURE__ */ React10.createElement(Text10, { color: "magenta", bold: true, underline: true }, "\u{1F7E3} STEERING HINT RESOLUTION"), /* @__PURE__ */ React10.createElement(Text10, { marginTop: 1 }, "The agent already finished the task before your hint was consumed."), /* @__PURE__ */ React10.createElement(Box10, { marginTop: 1, backgroundColor: "#222", paddingX: 1, width: "100%" }, /* @__PURE__ */ React10.createElement(Text10, { italic: true, color: "gray" }, '"', data, '"')), /* @__PURE__ */ React10.createElement(Box10, { marginTop: 1 }, /* @__PURE__ */ React10.createElement(Text10, { color: "cyan" }, "How would you like to proceed?")), /* @__PURE__ */ React10.createElement(Box10, { marginTop: 1 }, /* @__PURE__ */ React10.createElement(
       CommandMenu,

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "fluxflow-cli",
-	"version": "1.8.17",
+	"version": "1.8.19",
 	"description": "A high-fidelity agentic terminal assistant for the Flux Era.",
 	"keywords": [
 		"ai",