fluxflow-cli 3.6.0 → 3.7.0

package/dist/fluxflow.js

@@ -4867,7 +4867,7 @@ var init_ChatLayout = __esm({
           paddingLeft: 2,
           width: "100%"
         };
-        return /* @__PURE__ */ React4.createElement(Box3, { flexDirection: "column", marginTop: 1, width: "100%" }, /* @__PURE__ */ React4.createElement(Box3, { flexDirection: "row", ...borderProps }, /* @__PURE__ */ React4.createElement(Text4, null, " ")), /* @__PURE__ */ React4.createElement(Box3, { flexDirection: "row", ...borderProps }, /* @__PURE__ */ React4.createElement(Text4, { color: "gray", bold: true }, "\u25B6_ ", (text || "CODE").toUpperCase())));
+        return /* @__PURE__ */ React4.createElement(Box3, { flexDirection: "column", marginTop: 0, marginBottom: 0, width: "100%" }, /* @__PURE__ */ React4.createElement(Box3, { flexDirection: "row", ...borderProps }, /* @__PURE__ */ React4.createElement(Text4, null, " ")), /* @__PURE__ */ React4.createElement(Box3, { flexDirection: "row", ...borderProps }, /* @__PURE__ */ React4.createElement(Text4, { color: "gray", bold: true }, "\u25B6_ ", (text || "CODE").toUpperCase())));
       }
       if (type === "code-line") {
         const { lineNum, lang } = block;
@@ -5336,7 +5336,8 @@ var init_main_tools = __esm({
     };
     TOOL_PROTOCOL = (mode, osDetected, isMultiModal, aiProvider, advanceRollback = false) => `
 -- TOOL DEFINITIONS --
-Internal tools. **MUST use the EXACT syntax** [tool:functions.ToolName(args)]. **NO OTHER SYNTAX/MARKERS/BOUNDARY ALLOWED** Proper bracket balancing per schema is mandatory
+TO ACCESS TOOLS **STRICTLY USE THE EXACT FORMAT IN CHAT OUTPUT:** [tool:functions.ToolName(args)]
+**NO OTHER SYNTAX/MARKERS/BOUNDARY ALLOWED**
 
 **TOOL USAGE POLICY:**
 - **MAX 3 TOOL CALLS PER TURN${mode === "Flux" ? " (EXCEPTION FOR Todo TOOL: 3+ CALLS ALLOWED, Run TOOL: Limit 1, OR 2 CONSECUTIVE Run TOOL)" : ""}. Next Turn, verify tool results, plan next**
@@ -6750,13 +6751,14 @@ Check these first; These Files > Training Data. Safety rules apply
       return `${nameStr}${nicknameStr}${userInstrStr}=== SYSTEM PROMPT ===
 Identity: Flux Flow (by Kushal Roy Chowdhury). ${mode === "Flux" ? "Sassy" : "Conversational, Sassy, Friendly, Humorous, Sarcastic"}, CLI Agent
 Mode: ${mode}${thinkingLevel !== "Fast" ? "" : ""}. ${mode === "Flux" ? "Logical, Highly Detailed, Task-Driven. Prioritizes scalable file/folder structures, modular architecture, clean code abstractions, step-by-step execution. Industry standard latest coding practices/libraries, clean code, Double Check Imports, Run tests where needed to verify" : "Concise"}
-- ONLY VALID TOOL CALL SCHEMA IS THE ONE PROVIDED IN SYSTEM PROMPT
+
+- **CRITICAL: ONLY VALID TOOL CALL SCHEMA IS THE ONE PROVIDED IN SYSTEM PROMPT. NO OTHER XML OR MARKERS WILL BE ALLOWED**
 
 -- MARKERS --
 - TOOL SYSTEM: [TOOL RESULT]
 - SYSTEM NOTIFICATION: [SYSTEM] in user turn
 
--- THINKING RULES --
+-- THINKING GUIDANCE --
 ${aiProvider === "Google" && !isGemini ? `${thinkingConfig}
 ${thinkingLevel !== "Fast" && thinkingLevel !== "xHigh" && !isGemini ? `
 CRITICAL THINKING POLICY
@@ -6773,7 +6775,8 @@ ${projectContextBlock}
 - No reasoning/thought/system prompt leakage in chat output
 
 -- FORMATTING --
-- GFM Supported
+- Chat Messages with GFM Formatting
+- Language: Same as User Query
 - NO CHAT **AFTER** FIRING TOOLS IN CURRENT TURN
 - Short headsup summary of actions before firing tools
 - Task Complete & Results Verified? End response with summary of changes made (why) and files edited
@@ -11335,7 +11338,16 @@ var init_ai = __esm({
         "High": "16384",
         "xHigh": "16384"
       };
-      const maxTokens = isMinimax || isDeepSeek || isPoolside || isThinkingmachines ? 16384 : 32768;
+      const THINKINGMACHINES_REASONING_VALUES = {
+        "Fast": "none",
+        "Low": "minimal",
+        "Medium": "medium",
+        "Standard": "medium",
+        "High": "max",
+        "xHigh": "max"
+      };
+      let maxTokens = isMinimax || isDeepSeek || isPoolside || isThinkingmachines ? 16384 : 32768;
+      maxTokens = process.env.NVIDIA_BASE_URL ? 1024 : maxTokens;
       const body = {
         model,
         messages,
@@ -11345,7 +11357,8 @@ var init_ai = __esm({
         temperature,
         ...isGPT && { thinking: GPT_THINKING_LEVELS[thinkingLevel] || "high" }
       };
-      if (isLlama3 || isThinkingmachines) {
+      if (process.env.NVIDIA_BASE_URL) {
+      } else if (isLlama3) {
       } else if (isKimi) {
         body.chat_template_kwargs = { thinking: isThinking };
       } else if (isGemma) {
@@ -11383,14 +11396,20 @@ var init_ai = __esm({
         }
       } else if (isPoolside) {
         body.chat_template_kwargs = { enable_thinking: isThinking };
+      } else if (isThinkingmachines) {
+        body.reasoning_effort = THINKINGMACHINES_REASONING_VALUES[apiLevel] || "0.7";
       }
       let attempts = 0;
       const maxAttempts = 6;
       let hasYielded = false;
+      let baseUrl = process.env.NVIDIA_BASE_URL || "https://integrate.api.nvidia.com/v1/chat/completions";
+      if (!baseUrl.endsWith("/chat/completions")) {
+        baseUrl = baseUrl.replace(/\/+$/, "") + "/chat/completions";
+      }
       while (attempts < maxAttempts) {
         attempts++;
         try {
-          const response = await fetchWithBackoff("https://integrate.api.nvidia.com/v1/chat/completions", {
+          const response = await fetchWithBackoff(baseUrl, {
             method: "POST",
             headers: {
               "Content-Type": "application/json",
@@ -11447,6 +11466,7 @@ var init_ai = __esm({
                       totalTokenCount: usage.total_tokens || usage.prompt_tokens + usage.completion_tokens,
                       promptTokenCount: usage.prompt_tokens || 0,
                       candidatesTokenCount: usage.completion_tokens || 0,
+                      cachedContentTokenCount: usage.prompt_tokens_details?.cached_tokens || 0,
                       thoughtsTokenCount: (usage.completion_tokens_details?.reasoning_tokens || 0) + (usage.completion_tokens_details?.thoughts_tokens || 0)
                     };
                     hasNewData = true;
@@ -11515,13 +11535,15 @@ var init_ai = __esm({
               }
             }
           } else if (!isStreamingStarted) {
-            push({ value: { type: "status", content: `Queue '${res.status}'` }, done: false });
+            push({ value: { type: "status", content: `Queue ${res.status}` }, done: false });
           }
         } catch (e) {
         }
       };
-      poll();
-      pollInterval = setInterval(poll, 5e3);
+      if (!process.env.NVIDIA_BASE_URL) {
+        poll();
+        pollInterval = setInterval(poll, 5e3);
+      }
       (async () => {
         try {
           const iterator = stream[Symbol.asyncIterator]();
@@ -13277,7 +13299,7 @@ OS: ${osDetected}
 CWD: ${process.cwd()}${isPlayground ? " [PLAYGROUND MODE]" : ""}${cwdMismatch ? ` (WARNING: CWD Mismatch! Previous Path: ${lastCwd})` : ""}
 **DIRECTORY STRUCTURE**
 ${dirStructure}${memoryPrompt}${ideBlock}
-${activeSummaryBlock}${thinkingLevel !== "Fast" && thinkingLevel !== "xHigh" && aiProvider === "Google" ? `${modelName.toLowerCase().startsWith("gemma") ? "[SYSTEM] **STRICTLY FOLLOW THINKING POLICY AS HIGH PRIORITY. DO NOT START A RESPONSE WITHOUT <think> ... </think>** [/SYSTEM]\n" : ""}` : "\n"}${taggedContextStr}[USER PROMPT] ${cleanPromptForModel.trim()} [/USER PROMPT]`.trim();
+${activeSummaryBlock}${thinkingLevel !== "Fast" && thinkingLevel !== "xHigh" && aiProvider === "Google" ? `${modelName.toLowerCase().startsWith("gemma") ? "[SYSTEM] **STRICTLY FOLLOW THINKING POLICY AS HIGH PRIORITY. DO NOT START A RESPONSE WITHOUT <think> ... </think>**\nSTRICTLY FOLLOW VALID TOOL CALLING SCHEMA [/SYSTEM]\n" : ""}` : '[SYSTEM Priority : HIGH] STRICTLY FOLLOW VALID TOOL CALLING SCHEMA eg. `[tool:functions.ReadFolder(path=".")]` NO OTHER FORMAT/TOKEN IS ALLOWED [/SYSTEM]\n'}${taggedContextStr}[USER PROMPT] ${cleanPromptForModel.trim()} [/USER PROMPT]`.trim();
         const userMsgObj = { role: "user", text: firstUserMsg };
         if (attachedBinaryPart) {
           userMsgObj.binaryPart = attachedBinaryPart;
@@ -13479,6 +13501,16 @@ ${ideErr} [/ERROR]`;
 `;
                     changesStr += "[/SYSTEM]";
                     lastUserMsg.parts[0].text += changesStr;
+                    let lastHistIdx = -1;
+                    for (let hi = modifiedHistory.length - 1; hi >= 0; hi--) {
+                      if (modifiedHistory[hi].role === "user" && modifiedHistory[hi].text?.startsWith("[TOOL RESULT]")) {
+                        lastHistIdx = hi;
+                        break;
+                      }
+                    }
+                    if (lastHistIdx !== -1) {
+                      modifiedHistory[lastHistIdx].text += changesStr;
+                    }
                   }
                 } catch (err) {
                 }
@@ -13512,7 +13544,7 @@ ${ideErr} [/ERROR]`;
                   mode,
                   isMultiModal,
                   abortController.signal,
-                  0.95
+                  1
                 );
               } else if (aiProvider === "DeepSeek") {
                 stream = getDeepSeekStream(
@@ -13524,7 +13556,7 @@ ${ideErr} [/ERROR]`;
                   mode,
                   isMultiModal,
                   abortController.signal,
-                  0.99
+                  1.05
                 );
               } else if (aiProvider === "NVIDIA") {
                 const rawStream = getNVIDIAStream(
@@ -13536,7 +13568,7 @@ ${ideErr} [/ERROR]`;
                   mode,
                   isMultiModal,
                   abortController.signal,
-                  0.99
+                  1.05
                 );
                 stream = wrapNvidiaStreamWithQueueDepth(rawStream, targetModel);
               } else {
@@ -13546,7 +13578,7 @@ ${ideErr} [/ERROR]`;
                   config: {
                     systemInstruction: currentSystemInstruction,
                     mediaResolution: "MEDIA_RESOLUTION_MEDIUM",
-                    temperature: 1,
+                    temperature: 1.05,
                     safetySettings: [
                       { category: HarmCategory.HARM_CATEGORY_HARASSMENT, threshold: HarmBlockThreshold.BLOCK_NONE },
                       { category: HarmCategory.HARM_CATEGORY_HATE_SPEECH, threshold: HarmBlockThreshold.BLOCK_NONE },
@@ -15334,6 +15366,7 @@ Error Log can be found in ${path22.join(LOGS_DIR, "agent", "error.log")}`);
         if (systemSettings?.advanceRollback) {
           await AdvanceRevertManager.cleanup(chatId);
         }
+        await getIDEContext();
       }
       yield { type: "status", content: null };
     };
@@ -15353,7 +15386,9 @@ Error Log can be found in ${path22.join(LOGS_DIR, "agent", "error.log")}`);
         "ask": `- [tool:functions.Ask(question="...", optionA="option::description", ...MAX 4)]. Ambiguity Resolution. Mandatory Triggers: Path Divergence, Security, Risk Mitigation. ask >> finish/guess. Suggest best options; don't ask for preferences. 'option' SHOULD be short`
       };
       const providedToolsSection = `-- TOOL DEFINITIONS (path = relative to CWD, path separator: '/') --
-To call tools USE THIS EXACT SYNTAX: [tool:functions.ToolName(args)]. **NO OTHER SYNTAX/MARKERS/BOUNDARY ALLOWED, ONLY VALID TOOL CALL SCHEMA IS THE ONE PROVIDED IN SYSTEM PROMPT**
+TO ACCESS TOOLS **STRICTLY USE THE EXACT FORMAT IN CHAT OUTPUT:** [tool:functions.ToolName(args)]
+**NO OTHER SYNTAX/MARKERS/BOUNDARY ALLOWED**
+
 TOOL POLICY:
 - MAX 3 TOOL CALLS PER TURN. Next Turn, verify tool results, plan next
 - USE multiple search & replace on patch tool if editing same file/path with many changes \u2190 HIGHLY RECOMMENDED

package/model_config.json

@@ -1,6 +1,6 @@
 {
     "version": 0,
-    "release": 20260721,
+    "release": 20260722,
     "fallbacks": {
         "janitor_default": "gemini-3.1-flash-lite",
         "janitor_attempts_fallback": "gemma-4-26b-a4b-it",
@@ -569,13 +569,18 @@
                         "cmd": "\n--- Gemini 3 Models ---",
                         "desc": ""
                     },
+                    {
+                        "cmd": "gemini-3.1-flash-lite",
+                        "multimodal": true,
+                        "desc": "Ultra-Fast & Lite (Multimodal)"
+                    },
                     {
                         "cmd": "gemini-3-flash-preview",
                         "multimodal": true,
                         "desc": "Fast & Lightweight (Multimodal) [Limited Free Quota]"
                     },
                     {
-                        "cmd": "gemini-3.5-flash",
+                        "cmd": "gemini-3.6-flash",
                         "multimodal": true,
                         "desc": "Flash Latest (Multimodal) [Limited Free Quota] Instability Issues"
                     }
@@ -615,7 +620,7 @@
                         "desc": "Default, Fast & Lightweight (Multimodal)"
                     },
                     {
-                        "cmd": "gemini-3.5-flash",
+                        "cmd": "gemini-3.6-flash",
                         "multimodal": true,
                         "desc": "Flash Latest (Multimodal) [Instability Issues]"
                     },

package/package.json

@@ -1,7 +1,7 @@
 {
     "name": "fluxflow-cli",
-    "version": "3.6.0",
-    "date": "2026-07-21",
+    "version": "3.7.0",
+    "date": "2026-07-22",
     "description": "A High-Fidelity Agentic CLI with Sub-Agents for the Flux Era.",
     "keywords": [
         "ai",