fluxflow-cli 1.8.18 → 1.8.19

package/dist/fluxflow.js

@@ -770,19 +770,19 @@ var init_thinking_prompts = __esm({
     thinking_prompts_default = {
       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**: 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:\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**: 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:\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.",
-      Minimal: "EFFORT_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."
+- 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."
     };
   }
 });
@@ -794,7 +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";
@@ -808,7 +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}
 ------------------------------------------
@@ -833,7 +833,7 @@ If you see a [STEERING HINT] from user, give that prompt priority for the task a
 -- START THINKING INSTRUCTIONS --
 ${thinkingConfig}
 
-BEFORE USING ANY TOOL THINKING IS **MANDATORY**. ALWAYS PREFER TO ENTER IN THINKING AS PER INSTRUCTIONS FOR MORE ACCURACY, AVOID DIRECT SHOTS.
+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)}
@@ -1488,7 +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();
@@ -2665,7 +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 = [];
@@ -2698,16 +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 };
@@ -2729,48 +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 = "";
@@ -2832,9 +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));
@@ -2864,6 +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;
                   }
@@ -3067,24 +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")}`);
+              }
             }
           }
         }
@@ -3243,8 +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 };
@@ -3928,10 +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" },
@@ -4951,7 +4955,7 @@ Selection: ${val}`,
               setActiveView("chat");
               setTimeout(() => {
                 handleSubmit(val);
-              }, 50);
+              }, 200);
             },
             onEdit: (val) => {
               setResolutionData(null);
@@ -5198,7 +5202,7 @@ var init_app = __esm({
     init_text();
     SESSION_START_TIME = Date.now();
     CHANGELOG_URL = "https://fluxflow-cli.onrender.com/changelog.html";
-    versionFluxflow = "1.8.18";
+    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": "1.8.18",
+	"version": "1.8.19",
 	"description": "A high-fidelity agentic terminal assistant for the Flux Era.",
 	"keywords": [
 		"ai",