fluxflow-cli 3.4.3 → 3.4.5

package/dist/fluxflow.js

@@ -6610,30 +6610,36 @@ var JANITOR_TOOLS_PROTOCOL;
 var init_janitor_tools = __esm({
   "src/data/janitor_tools.js"() {
     JANITOR_TOOLS_PROTOCOL = (isMemoryEnabled = true, needTitle = true) => `
-Your tool syntax is: '[tool:functions.ToolName(args...)]'
+Your exact tool syntax is: [tool:functions.ToolName(args...)]. Malformed calls will fail parsing. **NO OTHER SYNTAX/MARKERS/BOUNDARY ALLOWED** Proper bracket balancing per schema is mandatory
 
--- CHAT MANAGEMENT TOOLS (MUST CALL THESE 2 TOOLS ALWAYS) --
-[tool:functions.Chat(title="<short creative title of FULL conversation in 3 or 4 words>")]. Consider full chat context to generate title NOT just latest message.
-[tool:functions.Memory(action="temp", content="<summary of the user prompt & model responses ONLY FROM LATEST PROMPT UNDER 40 WORDS>. [Talked on: <date> <hour>]")]. Time format: YYYY-MM-DD HH am/pm
+-- CHAT MANAGEMENT TOOLS (MUST CALL THESE 2 TOOLS IN THIS TURN) --
+1. [tool:functions.Chat(title="<short creative title of FULL conversation in 3 or 4 words>")]. Consider full chat context to generate title NOT just latest message
+2. [tool:functions.Memory(action="temp", content="<summary of the user prompt & model responses ONLY FROM LATEST PROMPT UNDER 40 WORDS>. [Talked on: <date> <hour>]")]. Time format: YYYY-MM-DD HH am/pm
 
 ${isMemoryEnabled ? `-- User-specific long-term/permanent memory (USE BASED ON CONVERSATION CONTEXT, DO NOT RE-SAVE MEMORY WHICH IS ALREADY SAVED) --
-- Add: [tool:functions.Memory(action="user", method="add", content="<string to add>. [Saved on: <date ONLY>]", score=2)] (Set score=2 ONLY if the user explicitly asked to "remember" or "save" this information, else omit this parameter entirely)
-- Delete: [tool:functions.Memory(action="user", method="delete", id="<memory id>")]
-- Update: [tool:functions.Memory(action="user", method="update", content-new="string to update", id="<memory id>")]
+1. Add: [tool:functions.Memory(action="user", method="add", content="<string to add>. [Saved on: <date ONLY>]", score=2)] (Set score=2 ONLY if the user explicitly asked to "remember" or "save" this information, else omit this parameter entirely)
+2. Delete: [tool:functions.Memory(action="user", method="delete", id="<memory id>")]
+3. Update: [tool:functions.Memory(action="user", method="update", content-new="string to update", id="<memory id>")]
 
 -- Memory Relevance Decay Tool --
-- Score Adjustment: [tool:functions.addMemScore(id="<memory id>")]
-  You MUST call this tool when a specific saved memory in the '-- CURRENT SAVED USER MEMORIES --' list was relevant, referenced, or helpful in the agent's response or user prompt IN CURRENT MESSAGE. You can stack multiple calls.
+1. Score Adjustment: [tool:functions.addMemScore(id="<memory id>")]
+  You MUST call this tool when a specific saved memory in the '-- CURRENT SAVED USER MEMORIES --' list was relevant, referenced, or helpful in the agent's response or user prompt IN CURRENT MESSAGE. You can stack multiple calls
 
 Explicit Triggers for permanent memory:
-- User explicitly asks to 'remember' something.
-- User mentions something important long-term that should be remembered.
-- User provides information that could be useful for long-term reference.
-- User shares personal information or preferences.
+- User explicitly asks to 'remember' something
+- User mentions something important long-term that should be remembered
+- User provides information that could be useful for long-term reference
+- User shares personal information or preferences
 
 Usage Rules:
 - Frequency for 'user' action: Based on explicit triggers.
-- IF YOU WANT TO SAVE SOMETHING, BUT SIMILAR MEMORY ALREADY EXISTS, USE THE UPDATE METHOD NOT ADD` : ""}`.trim();
+- IF YOU WANT TO SAVE SOMETHING, BUT SIMILAR MEMORY ALREADY EXISTS, USE THE UPDATE METHOD NOT ADD
+
+Usage Rules:
+- Chat Title is MANDATORY
+- TEMPORARY Memory is MANDATORY
+- WHEN Called User Memory, STILL use Temporary Memory
+- MUST NOT IGNORE ANY TOOL CALLS IN GIVEN CONTEXT OF CHAT BETWEEN USER & AGENT` : ""}`.trim();
   }
 });
 
@@ -6667,8 +6673,8 @@ ${tempMemories}` : "";
       const userMemoriesStr = userMemories?.length > 0 ? `--- SAVED MEMORIES (PRIORITY: MEDIUM, USER PREFERENCES) ---
 ${userMemories}` : "";
       const parts = [userMemoriesStr, tempMemoriesStr].filter((p) => p.length > 0);
-      return parts.length > 0 ? `[SYSTEM CONTEXT]
-${parts.join("\n\n")}
+      return parts.length > 0 ? `[MEMORY CONTEXT]
+${parts.join("\n")}
 ` : "";
     };
     getSystemInstruction = (profile, thinkingLevel, mode, systemSettings, isMemoryEnabled = true, isFirstPrompt = false, aiProvider = "Google", isMultiModal = false, isGemini) => {
@@ -6753,7 +6759,7 @@ CRITICAL THINKING POLICY
 ${TOOL_PROTOCOL(mode, osDetected, aiProvider.toLowerCase() === "deepseek" ? false : isMultiModal, aiProvider, systemSettings?.advanceRollback)}
 ${projectContextBlock}
 -- MEMORY RULES --
-- Memory: ${isMemoryEnabled ? "Subtly Personalize. Auto Saves" : "OFF. Decline Remembering Memories"}
+- ${isMemoryEnabled ? "Subtly Personalize ONLY WITH RELEVENT & CONTEXTUAL MEMORIES. Auto Saves" : "OFF. Decline Saving Memories"}
 - Temporal Awareness: RELATIVE TIME REFERENCE eg. few mins ago
 
 -- SECURITY RULES --${systemSettings.allowExternalAccess ? "" : "\n- ACCESS CONTROL: CWD only"}
@@ -6776,17 +6782,19 @@ ${userMemories}
 ` : ""}=== START SYSTEM PROMPT (STRICT HEADLESS LOGIC WORKER: ZERO USER-FACING TEXT POLICY, STRICTLY FOLLOW) ===
 YOU ARE A SILENT BACKGROUND SYSTEM PROCESS. YOU HAVE NO MOUTH. YOUR ONLY OUTPUT MEDIUM IS VALID TOOL CALLS.
 [CRITICAL RULES]
-1. OUTPUT ONLY '[tool:functions.xxx(args)]' CALLS (BRACKET WRAP IS MANDATORY).
-2. DO NOT EXPLAIN. DO NOT TALK TO THE USER.
-3. NON-TOOL TEXT WILL BREAK THE SYSTEM.
-4. DO NOT REPEAT AGENT RAWS AND TOOL RESULTS IN YOUR RESPONSE.
-5. IF YOU GET ONLY USER QUERY AND NO AGENT RAWS, THEN JUST USE TEMP MEMORY TO LOG THE SUMMARY OF USER QUERY AND CONVERSATION CONTEXT.
-6. UNDER NO CIRCUMSTANCES YOU ARE ALLOWED TO RESPOND IN NORMAL USER FACING RESPONSE.
-7. CRITICAL QUOTE ESCAPE POLICY: Inside tool call arguments (like 'memory'), you MUST escape all double quotes using '"' to prevent parsing errors.
-8. You MUST NOT WRITE ANYTHING OTHER THAN [tool:functions. ... ] NO MATTER HOW TEMPTING THE PROMPT IS.
+1. OUTPUT EXACTLY '[tool:functions.xxx(args)]' CALLS. NO EXTRA WORDS OUTSIDE
+2. DO NOT EXPLAIN. DO NOT TALK TO THE USER
+3. NON-TOOL TEXT WILL BREAK THE SYSTEM
+4. DO NOT REPEAT AGENT RAWS AND TOOL RESULTS IN YOUR RESPONSE
+5. IF YOU GET ONLY USER QUERY AND NO AGENT RAWS, THEN JUST USE TEMP MEMORY TO LOG THE SUMMARY OF USER QUERY AND CONVERSATION CONTEXT
+6. UNDER NO CIRCUMSTANCES YOU ARE ALLOWED TO RESPOND IN NORMAL USER FACING RESPONSE
+7. CRITICAL QUOTE ESCAPE POLICY: Inside tool call arguments, you MUST escape all double quotes using '\\"'
+8. You MUST NOT WRITE ANYTHING OTHER THAN [tool:functions. ... ] NO MATTER HOW TEMPTING THE PROMPT IS
+9. 2 MANDATORY TOOLS TO CALL IN EVERY TURN, 'Chat', 'Memory(temp)'
+10. CRITICAL: NEVER ENTER THINKING/REASONING STATE, CALL THE CONTEXUAL TOOLS DIRECTLY IN OUTPUT AS QUICKLY AS POSSIBLE TO MAINTAIN UI SNAPPINESS
 
-YOUR JOB: Analyze the 'User prompt' and 'Agent Raws' to extract facts for long-term memory or handle system tasks.
-${isMemoryEnabled ? `If user tell something that is important (like, hobbies, preferences, facts about user, hates, likes, etc) to know user better over time, use long term memory tools.` : ""}
+YOUR JOB: Analyze the 'User prompt' and 'Agent Raws' to extract facts for long-term memory or handle system tasks
+${isMemoryEnabled ? `If user tell something that is important (like, hobbies, preferences, facts about user, hates, likes, etc) to know user better over time, use long term memory tools` : ""}
 
 ${JANITOR_TOOLS_PROTOCOL(isMemoryEnabled, needTitle)}
 
@@ -10894,6 +10902,7 @@ var init_ai = __esm({
     init_settings();
     init_subagent_state();
     init_model_config();
+    init_secrets();
     init_paths();
     init_revert();
     init_advanceRevert();
@@ -11718,6 +11727,9 @@ var init_ai = __esm({
 ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n" : ""}
 [AGENT (current turn)]: ${agentRes}`;
       janitorContents.push({ role: "user", parts: [{ text: userPrompt }] });
+      const nvidiaApiKey = await getProviderAPIKey("NVIDIA");
+      const fullSettings = await loadSettings();
+      const isNvidiaFree = fullSettings.quotas?.providerTiers?.NVIDIA === "Free";
       let finalSynthesis = "";
       let attempts = 0;
       const MAX_JANITOR_RETRIES = isMemoryEnabled ? 12 : -1;
@@ -11728,10 +11740,16 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
           }
           let fullContent = "";
           let lastUsage = null;
+          let useNvidiaFallback = false;
+          let effectiveProvider = aiProvider;
           try {
             const timeoutPromise = new Promise(
-              (_, reject) => setTimeout(() => reject(new Error("JANITOR_TIMEOUT")), 6e4)
+              (_, reject) => setTimeout(() => reject(new Error("JANITOR_TIMEOUT")), 1e4)
             );
+            const useNvidiaFallbackForGoogle = aiProvider === "Google" && attempts >= 0 && attempts < 6 && nvidiaApiKey && isNvidiaFree;
+            const useNvidiaFallbackForDeepSeek = aiProvider === "DeepSeek" && attempts < 4 && nvidiaApiKey && isNvidiaFree;
+            useNvidiaFallback = useNvidiaFallbackForGoogle || useNvidiaFallbackForDeepSeek;
+            effectiveProvider = useNvidiaFallback ? "NVIDIA" : aiProvider;
             const streamPromise = (async () => {
               if (aiProvider === "OpenRouter") {
                 const janitorOpenRouterModel = getFallbackValue("janitor_open_router");
@@ -11745,12 +11763,12 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
                   mode,
                   false,
                   null,
-                  0.75
+                  0.6
                 );
                 const iterator2 = stream[Symbol.asyncIterator]();
                 const firstResult2 = await iterator2.next();
                 return { iterator: iterator2, firstResult: firstResult2 };
-              } else if (aiProvider === "DeepSeek") {
+              } else if (aiProvider === "DeepSeek" && !useNvidiaFallback) {
                 const stream = getDeepSeekStream(
                   apiKey,
                   getFallbackValue("deepseek_fast_fallback"),
@@ -11761,15 +11779,16 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
                   mode,
                   false,
                   null,
-                  0.75
+                  0.6
                 );
                 const iterator2 = stream[Symbol.asyncIterator]();
                 const firstResult2 = await iterator2.next();
                 return { iterator: iterator2, firstResult: firstResult2 };
-              } else if (aiProvider === "NVIDIA") {
+              } else if (aiProvider === "NVIDIA" || useNvidiaFallback) {
                 const stream = getNVIDIAStream(
-                  apiKey,
+                  useNvidiaFallback ? nvidiaApiKey : apiKey,
                   getFallbackValue("nvidia_janitor_fallback"),
+                  // "mistralai/mistral-nemotron", // [DEBUGGING POINT]
                   janitorContents,
                   janitorPrompt,
                   "Fast",
@@ -11777,7 +11796,7 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
                   mode,
                   false,
                   null,
-                  0.75
+                  0.6
                 );
                 const iterator2 = stream[Symbol.asyncIterator]();
                 const firstResult2 = await iterator2.next();
@@ -11788,8 +11807,7 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
                   contents: janitorContents,
                   config: {
                     systemInstruction: janitorPrompt,
-                    maxOutputTokens: 512,
-                    temperature: 0.75,
+                    temperature: 0.7,
                     safetySettings: [
                       { category: HarmCategory.HARM_CATEGORY_HARASSMENT, threshold: HarmBlockThreshold.BLOCK_NONE },
                       { category: HarmCategory.HARM_CATEGORY_HATE_SPEECH, threshold: HarmBlockThreshold.BLOCK_NONE },
@@ -11817,14 +11835,14 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
             let { value: firstChunk, done: firstDone } = firstResult;
             if (!firstDone && firstChunk) {
               const parts = firstChunk.candidates?.[0]?.content?.parts;
-              const chunkText = parts?.[1]?.text || parts?.[0]?.text || (typeof firstChunk.text === "function" ? firstChunk.text() : "");
+              const chunkText = parts ? effectiveProvider === "Google" ? parts[1]?.text || parts[0]?.text || "" : parts.filter((p) => p.text && !p.thought).map((p) => p.text).join("") : typeof firstChunk.text === "function" ? firstChunk.text() : "";
               if (chunkText) {
                 fullContent += chunkText;
               }
               lastUsage = firstChunk.usageMetadata;
               for await (const chunk of { [Symbol.asyncIterator]: () => iterator }) {
                 const p = chunk.candidates?.[0]?.content?.parts;
-                const t = p?.[1]?.text || p?.[0]?.text || (typeof chunk.text === "function" ? chunk.text() : "");
+                const t = p ? effectiveProvider === "Google" ? p[1]?.text || p[0]?.text || "" : p.filter((part) => part.text && !part.thought).map((part) => part.text).join("") : typeof chunk.text === "function" ? chunk.text() : "";
                 if (t) fullContent += t;
                 lastUsage = chunk.usageMetadata;
               }
@@ -11838,13 +11856,13 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
               const total = lastUsage.totalTokenCount || 0;
               const cached = lastUsage.cachedContentTokenCount || 0;
               const candidates = (lastUsage.candidatesTokenCount || 0) + (lastUsage.thoughtsTokenCount || 0);
-              const jModel = janitorModel || getFallbackValue("janitor_default");
-              await addToUsage("tokens", total, aiProvider, jModel);
+              const jModel = useNvidiaFallback ? getFallbackValue("nvidia_janitor_fallback") : effectiveProvider === "DeepSeek" ? getFallbackValue("deepseek_fast_fallback") : effectiveProvider === "OpenRouter" ? getFallbackValue("janitor_open_router") : janitorModel || (attempts === MAX_JANITOR_RETRIES ? getFallbackValue("janitor_default") : getFallbackValue("gemma_janitor_fallback_google"));
+              await addToUsage("tokens", total, effectiveProvider, jModel);
               if (cached > 0) {
-                await addToUsage("cachedTokens", cached, aiProvider, jModel);
+                await addToUsage("cachedTokens", cached, effectiveProvider, jModel);
               }
               if (candidates > 0) {
-                await addToUsage("candidateTokens", candidates, aiProvider, jModel);
+                await addToUsage("candidateTokens", candidates, effectiveProvider, jModel);
               }
             }
           } else {
@@ -11914,7 +11932,7 @@ ${originalTextProcessed.length > USER_CONTEXT_LENGTH ? "... (truncated) ...\n\n"
 
 `);
           if (attempts > MAX_JANITOR_RETRIES) break;
-          const backoff = Math.min(1e3 * Math.pow(2, attempts - 1), 8e3);
+          const backoff = Math.min(750 * Math.pow(2, attempts - 1), 5e3);
           await new Promise((resolve) => setTimeout(resolve, backoff));
         }
       }
@@ -13383,6 +13401,8 @@ ${ideErr} [/ERROR]`;
                   reject(new DOMException("The user aborted a request.", "AbortError"));
                 });
               });
+              abortPromise.catch(() => {
+              });
               let activeContents = contents;
               if (aiProvider === "OpenRouter") {
                 stream = getOpenRouterStream(
@@ -17859,6 +17879,12 @@ ${cleanText}`, color: "magenta" }];
               const target = h[targetId] || Object.values(h).find((h2) => h2.name.toLowerCase() === targetId.toLowerCase());
               if (target) {
                 stdout.write("\x1B[2J\x1B[3J\x1B[H");
+                if (process.stdout.isTTY) {
+                  const chatName = target?.name || "";
+                  const title = chatName && !chatName.startsWith("flow-") && !chatName.startsWith("Session ") ? chatName : "FluxFlow | Resumed";
+                  process.stdout.write(`\x1B]0;${title}\x07`);
+                  process.stdout.write(`\x1B]633;P;TerminalTitle=${title}\x07`);
+                }
                 clearBlocksCache();
                 chatLoadingRef.current = true;
                 setChatId(targetId);
@@ -17938,6 +17964,8 @@ ${cleanText}`, color: "magenta" }];
             stdout.write("\x1B[2J\x1B[3J\x1B[H");
             if (stdout.isTTY) {
               stdout.write("\x1B[?2004h");
+              process.stdout.write(`\x1B]0;FluxFlow\x07`);
+              process.stdout.write(`\x1B]633;P;TerminalTitle=FluxFlow\x07`);
             }
           }
           setMessages([
@@ -20171,6 +20199,12 @@ Selection: ${val}`,
               const h = await loadHistory();
               if (h[id]) {
                 stdout.write("\x1B[2J\x1B[3J\x1B[H");
+                if (process.stdout.isTTY) {
+                  const chatName = h[id]?.name || "";
+                  const title = chatName && !chatName.startsWith("flow-") && !chatName.startsWith("Session ") ? chatName : "FluxFlow | Resumed";
+                  process.stdout.write(`\x1B]0;${title}\x07`);
+                  process.stdout.write(`\x1B]633;P;TerminalTitle=${title}\x07`);
+                }
                 clearBlocksCache();
                 chatLoadingRef.current = true;
                 setChatId(id);

package/model_config.json

@@ -1,6 +1,6 @@
 {
-    "version": 0,
-    "release": 20260711,
+    "version": 1,
+    "release": 20260712,
     "fallbacks": {
         "janitor_default": "gemini-3.1-flash-lite",
         "janitor_attempts_fallback": "gemma-4-26b-a4b-it",
@@ -9,7 +9,7 @@
         "deepseek_fast_fallback": "deepseek-chat",
         "deepseek_level_1": "deepseek-v4-flash",
         "deepseek_level_2": "deepseek-v4-pro",
-        "nvidia_janitor_fallback": "deepseek-ai/deepseek-v4-flash",
+        "nvidia_janitor_fallback": "mistralai/mistral-nemotron",
         "google_level_1": "gemini-3-flash-preview",
         "google_level_2": "gemini-3.5-flash",
         "gemma_emergency": "gemma-4-31b-it",
@@ -284,6 +284,21 @@
                         "multimodal": true,
                         "desc": "Multimodal"
                     },
+                    {
+                        "cmd": "mistralai/mistral-large-3-675b-instruct-2512",
+                        "multimodal": true,
+                        "desc": "Multimodal"
+                    },
+                    {
+                        "cmd": "mistralai/mistral-small-4-119b-2603",
+                        "multimodal": true,
+                        "desc": "Multimodal"
+                    },
+                    {
+                        "cmd": "mistralai/mistral-nemotron",
+                        "multimodal": false,
+                        "desc": "Text Only (No Thinking)"
+                    },
                     {
                         "cmd": "\n--- OpenAI Models ---",
                         "desc": ""
@@ -339,6 +354,11 @@
                         "multimodal": false,
                         "desc": "Text Only [EXPERIMENTAL]"
                     },
+                    {
+                        "cmd": "nvidia/nemotron-3-super-120b-a12b",
+                        "multimodal": false,
+                        "desc": "Text Only [EXPERIMENTAL]"
+                    },
                     {
                         "cmd": "\n--- Meta Models ---",
                         "desc": ""
@@ -401,6 +421,21 @@
                         "multimodal": true,
                         "desc": "Multimodal"
                     },
+                    {
+                        "cmd": "mistralai/mistral-large-3-675b-instruct-2512",
+                        "multimodal": true,
+                        "desc": "Multimodal"
+                    },
+                    {
+                        "cmd": "mistralai/mistral-small-4-119b-2603",
+                        "multimodal": true,
+                        "desc": "Multimodal"
+                    },
+                    {
+                        "cmd": "mistralai/mistral-nemotron",
+                        "multimodal": false,
+                        "desc": "Text Only (No Thinking)"
+                    },
                     {
                         "cmd": "\n--- OpenAI Models ---",
                         "desc": ""
@@ -456,6 +491,11 @@
                         "multimodal": false,
                         "desc": "Text Only [EXPERIMENTAL]"
                     },
+                    {
+                        "cmd": "nvidia/nemotron-3-super-120b-a12b",
+                        "multimodal": false,
+                        "desc": "Text Only [EXPERIMENTAL]"
+                    },
                     {
                         "cmd": "\n--- Meta Models ---",
                         "desc": ""

package/package.json

@@ -1,7 +1,7 @@
 {
     "name": "fluxflow-cli",
-    "version": "3.4.3",
-    "date": "2026-07-11",
+    "version": "3.4.5",
+    "date": "2026-07-12",
     "description": "A High-Fidelity Agentic CLI with Sub-Agents for the Flux Era.",
     "keywords": [
         "ai",