nyxora 26.7.5 → 26.7.9

package/CHANGELOG.md

@@ -4,6 +4,65 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Keep a Changelog](https://keepashangelog.com/en/1.0.0/),
 
+## [26.7.9]
+### Features & Architecture
+
+- **Deep Market Analysis Upgrade**: Significantly enhanced the Python ML Engine (`market.py`) to compute a comprehensive suite of advanced technical indicators including MACD, Bollinger Bands, EMA-20, ATR-14, and OBV using live OHLCV data.
+- **Smart Trend Classification**: The ML Engine now runs deterministic trend scoring (`STRONG_BULLISH` to `STRONG_BEARISH`) and generates an AI narrative summary of the market structure to prevent LLM hallucination and provide highly accurate context.
+- **Anti-Blocker Data Pipeline**: Bypassed Indonesian ISP blocks (e.g., internetsehat/ICON+) that previously blocked major crypto exchange APIs (Binance, Bybit, KuCoin, OKX) by routing historical candle data through `data-api.binance.vision`.
+
+### Refactor — Dependency & NPM Warning Cleanup
+
+- **Removed Unused & Problematic Dependencies**: Completely uninstalled `@google/genai`, `baileys`, and `@matrix-org/matrix-sdk-crypto-nodejs` from the workspace. This fully eliminates the annoying `npm warn allow-scripts` warnings that users previously encountered when running `npm install -g nyxora` or standard `npm install`.
+- **Migrated Image Analysis SDK**: Refactored `analyzeImage.ts` (Dynamic Require) to use `@google/generative-ai` (which has no install scripts) instead of the newer `@google/genai` (which triggers NPM security warnings), maintaining full Gemini 2.5 Flash compatibility.
+
+### Bug Fix — Playbooks Missing for Global npm Users
+
+- **Playbooks Not Distributed to Global Users (Critical Fix)**: Resolved a packaging omission where `packages/core/playbooks/` (containing 70 bundled SKILL.md playbooks) was missing from the `files[]` array in `package.json`. Global users running `npm install -g nyxora` never received any playbooks, causing `search_playbook` to always return "No playbooks available." and the system prompt to never list available skills. Fixed by adding `"packages/core/playbooks"` to the `files[]` field.
+
+### Bug Fix — "No response generated." Crash on Complex Multi-Step Prompts
+
+- **ContextSummarizer Redundant 4x LLM Calls (Critical Fix)**: Fixed a severe loop bug where `compressHistory()` was called once per nudge iteration (up to 4x per request) instead of once per session turn. The compressed result was never persisted back to the logger, so every iteration re-fetched the full uncompressed history and triggered another full LLM summarization call. Added a per-session in-memory cache keyed by text message count: if no new user/assistant messages have arrived, the cached compressed result is reused immediately — eliminating the redundant LLM calls entirely.
+- **TaskPlanner Context Bloat**: Refactored the TaskPlanner to inject its execution plan as a `role: 'system'` logger entry instead of prepending it to the user input string. Previously, the plan was concatenated into the user message body (`planInjection + "\n\nUSER REQUEST: " + input`), which inflated the user message size stored in logger and compounded context bloat on every subsequent turn. Also capped generated plans at 120 words to prevent runaway context consumption.
+- **Nudge Messages Now Actionable**: Replaced the generic `"[SYSTEM NUDGE] You did not output any tool calls..."` message with a structured, actionable prompt that includes: the original user task (truncated to 200 chars), the list of relevant available tools, and explicit instruction to either call a tool or output a final text answer. Applies to both OS Agent and Web3 Agent.
+- **Dead-End Fallback Replaced**: Replaced `"No response generated."` (which leaves the user with no actionable path) with `"⚠️ I encountered an issue processing your request. This can happen with very complex multi-step tasks. Please try rephrasing or breaking the request into smaller steps."` in all 4 abort paths across `osAgent.ts` and `web3Agent.ts`.
+- **Thinking-Prefill Continuation (Critical Fix)**: Ported the core silent-stop recovery technique used by the Hermes agent framework. When a reasoning model (e.g. Gemini 2.5 Pro with `reasoning_effort: high`) produces internal thinking content but no visible text or tool calls, Nyxora now appends the assistant's reasoning turn as-is and continues the loop — allowing the model to see its own reasoning on the next turn and naturally produce a tool call or text response. This is a fundamentally different approach from system nudges (which restart the reasoning chain from scratch). Prefill is attempted up to 2 times before falling back to the nudge path. Applies to both OS Agent and Web3 Agent.
+- **Anti-Silent-Stop System Prompt Rule**: Added `CRITICAL RULE 6` to the OS agent system prompt explicitly forbidding silent stops: after reasoning is complete, the model MUST produce either tool calls or a visible text answer. Ending a turn with only internal thinking is explicitly stated as not acceptable.
+- **MAX_TURNS Raised (OS Agent)**: Increased the stream loop iteration cap from 10 to 15 to give complex multi-step tasks (e.g. write file → send to Telegram) sufficient room to complete without hitting the ceiling.
+
+## [26.7.8]
+### Bug Fixes — Global Compatibility & Dashboard Crash
+
+- **Dashboard Not Accessible via `nyxora dashboard` (Critical Fix)**: Resolved a crash-on-startup bug where the entire Core gateway process would exit with code `1` before binding to port 3000, making `http://localhost:3000` unreachable. Root cause: `channels/index.ts` performed static top-level imports of `baileys`, `@slack/bolt`, and `@line/bot-sdk` — packages that are **not listed in `package.json` dependencies** and therefore not installed on global npm users' machines. Resolved by refactoring `channels/index.ts` into a `registerAllAdapters()` async function that uses dynamic `import()` with `MODULE_NOT_FOUND` error handling. Missing adapters now log a clear warning and are skipped instead of crashing the process.
+- **WhatsApp Adapter Lazy Load**: Migrated `whatsappAdapter.ts` from a static `import makeWASocket from 'baileys'` to a lazy `await import('baileys')` inside `start()`, with a graceful error and install instructions if `baileys` is absent.
+- **Slack Adapter Lazy Load**: Migrated `slackAdapter.ts` `@slack/bolt` import to lazy initialization inside `start()`.
+- **LINE Adapter Lazy Load**: Migrated `lineAdapter.ts` `@line/bot-sdk` import to lazy initialization inside `start()`.
+
+### Bug Fixes — Windows & Cross-Platform Compatibility
+
+- **Unix Socket Crash on Windows** (`ENOENT`/`ENOTSUP`): Replaced hardcoded Unix Domain Socket (`/tmp/nyxora-signer.sock`) in `signer/server.ts` with a cross-platform IPC strategy. On Windows, the signer now listens on TCP `127.0.0.1:3002`; Unix/Mac keeps the existing UDS path for security.
+- **Policy Engine Cross-Platform IPC**: Refactored all Signer proxy calls in `policy/server.ts` into a unified `signerRequestOptions()` helper that automatically selects Unix socket (Linux/Mac) or TCP (Windows). The optional UDS IPC server is now skipped entirely on Windows.
+- **vaultClient TCP Fallback**: Refactored `vaultClient.ts` with a `getPolicyOptions()` helper — on Windows, all Policy Engine requests use TCP; on Linux/Mac, Unix socket is preferred if available with automatic TCP fallback.
+- **Python Path on Windows**: Launcher now resolves the ML Engine Python executable to `venv/Scripts/python.exe` on Windows instead of `venv/bin/python`, preventing silent "Python not found" at startup.
+- **`pkill` Skipped on Windows**: Graceful shutdown in `launcher.ts` now guards `pkill -f ts-node` and `pkill -f uvicorn` behind a `process.platform !== 'win32'` check.
+- **Unix Socket Doctor Checks Skipped on Windows**: `nyxora doctor` now conditionally skips UDS socket health checks on Windows where they are not applicable.
+
+### Improvements — Configuration & Telemetry
+
+- **Cloudflare Tunnel Configurable**: Cloudflare auto-tunnel can now be disabled by setting `cloudflare_tunnel: false` in `~/.nyxora/config/config.yaml`. Previously it always launched unconditionally.
+- **ML Engine Health Check in `nyxora doctor`**: Added a dedicated check (#7) that pings `http://127.0.0.1:8000/health` to detect if the Python ML Engine sidecar is running. When daemon is stopped, it verifies the venv is installed and guides with `nyxora setup`.
+- **Node.js Version Enforcement**: Added `"engines": { "node": ">=22.0.0" }` to `package.json`. npm will now warn users attempting to install on incompatible Node.js versions instead of silently installing and crashing at runtime.
+- **Locale-Neutral System Prompt**: Replaced Indonesian-specific examples (`"cek saldo gue dirupiahin"`, `"fiat/rupiah"`) in `promptBuilder.ts` with multilingual examples (`"0.5 BTC in EUR"`, `"convert 100 SOL to JPY"`) to align with global user base.
+- **English-Only Source Code**: Translated the remaining Indonesian comment in `nyxDaemon.ts` (`// Kirim riwayat percakapan...`) to English for open-source contributor clarity.
+
+## [26.7.6]
+### Features & Architecture
+- **ML Engine Custom Provider Support**: Fixed a `500 Internal Server Error` in the Python ML Engine by explicitly supporting the `custom` and `openrouter` providers and gracefully handling empty API keys with a local fallback, ensuring robust execution for cognitive reasoning and RAG memory tasks.
+
+### Bug Fixes & Dashboard
+- **Dashboard Overview Key Detection**: Resolved a UI bug in the Dashboard Overview where custom and proxy providers (`custom_provider_key`, `9router_key`) incorrectly rendered as "Missing Key". The status engine now dynamically queries the exact credential mapping for off-standard providers.
+- **LLM Configuration Persistence Bug**: Fixed a configuration merging flaw in `Settings.tsx` where switching away from a `custom_provider` left the old `base_url` intact in the persistent payload. This previously caused subsequent standard LLM providers (e.g., OpenAI) to silently inherit the custom URL and crash. The dashboard now cleanly purges the `base_url` state upon provider transitions.
+
 ## [26.7.5]
 ### Reasoning & Agent Engine Architecture
 - **Event-Driven Tool Execution**: Fully refactored the core agent loop (`osAgent.ts`) from a linear synchronous executor to a highly scalable, event-driven architecture using Lifecycle Hooks (`beforeToolCall`, `afterToolCall`). This completely decouples security guardrails and domain-specific logic from the core loop.

package/README.md

@@ -123,43 +123,39 @@ It operates under a **Zero-Trust, Defense-in-Depth Cryptographically Bound Human
 
 The following diagram illustrates Nyxora's **6-Tier Hybrid Architecture**, showing the isolated communication channels across the ecosystem.
 
-```text
-+-------------------------------------------------------------+
-|                     Nyxora 6-Tier Architecture              |
-+-------------------------------------------------------------+
-
-    [ User / External Client ]
-               |
-               v
-+-----------------------------+       +-------------------------+
-|     Dashboard (UI)          |       |      MCP Server         |
-|        Port 5173            |       |       Port 3001         |
-+-----------------------------+       +-------------------------+
-               |                                  |
-               +---------------+------------------+
-                               |
-                               v
-                    +--------------------+
-                    |   Core LLM Runtime | <--- (NLP Parsing, Routing,
-                    |      Port 3000     |       Agent Logic)
-                    +--------------------+
-                      ^                |
-       (RAG & Math)   |                |  (Draft Transaction)
-                      v                v
-+-------------------------+   +-------------------------------+
-|       ML Engine         |   |    Policy Engine (Guard)      |
-|       Port 8000         |   |  Unix Socket (IPC) / Loopback |
-+-------------------------+   +-------------------------------+
-                                               |
-                                               | (Approved Payload)
-                                               v
-                              +-------------------------------+
-                              |    Signer Vault (Safe)        |
-                              |       Unix Socket (IPC)       |
-                              +-------------------------------+
-                                               |
-                                               v
-                                      [ Blockchain RPC ]
+```mermaid
+graph TD
+    classDef default fill:#1E1E2E,stroke:#4C4F69,stroke-width:2px,color:#CDD6F4;
+    classDef external fill:#11111B,stroke:#F38BA8,stroke-width:2px,color:#F38BA8,stroke-dasharray: 5 5;
+    classDef ui fill:#89B4FA,stroke:#1E1E2E,stroke-width:2px,color:#11111B;
+    classDef core fill:#A6E3A1,stroke:#1E1E2E,stroke-width:2px,color:#11111B;
+    classDef policy fill:#F9E2AF,stroke:#1E1E2E,stroke-width:2px,color:#11111B;
+    classDef signer fill:#F38BA8,stroke:#1E1E2E,stroke-width:2px,color:#11111B;
+    classDef blockchain fill:#CBA6F7,stroke:#1E1E2E,stroke-width:2px,color:#11111B;
+
+    User["User / External Client"]:::external
+
+    Dashboard["Dashboard (UI)<br/>Port 5173"]:::ui
+    MCP["MCP Server<br/>Port 3001"]:::ui
+
+    Core["Core LLM Runtime<br/>Port 3000<br/>(NLP Parsing, Routing, Agent Logic)"]:::core
+
+    ML["ML Engine<br/>Port 8000"]:::core
+    Policy["Policy Engine (Guard)<br/>Unix Socket (IPC) / Loopback"]:::policy
+    Signer["Signer Vault (Safe)<br/>Unix Socket (IPC)"]:::signer
+
+    RPC["Blockchain RPC"]:::blockchain
+
+    User --> Dashboard
+    User --> MCP
+
+    Dashboard --> Core
+    MCP --> Core
+
+    Core <-->|" RAG & Math "| ML
+    Core -->|" Draft Transaction "| Policy
+    Policy -->|" Approved Payload "| Signer
+    Signer --> RPC
 ```
 
 *Nyxora separates its duties into 6 independent layers for absolute security and cognitive depth:*

package/dist/launcher.js

@@ -71,7 +71,6 @@ const spawnService = (name, command, args, env, inheritStdio = false, cwd) => {
                 try {
                     const yaml = require('yaml');
                     const configPath = path_1.default.join(process.env.HOME || process.env.USERPROFILE || '', '.nyxora', 'config', 'config.yaml');
-                    // Fallback check just in case it hasn't migrated yet
                     const actualConfigPath = fs_1.default.existsSync(configPath) ? configPath : path_1.default.join(process.env.HOME || process.env.USERPROFILE || '', '.nyxora', 'config.yaml');
                     if (fs_1.default.existsSync(actualConfigPath)) {
                         const configStr = fs_1.default.readFileSync(actualConfigPath, 'utf8');
@@ -89,7 +88,12 @@ const spawnService = (name, command, args, env, inheritStdio = false, cwd) => {
                     }
                 }
                 catch (e) { }
-                process.exit(1);
+                console.log('\n[Launcher] Shutting down all services due to emergency shutdown...');
+                children.forEach(c => c.kill());
+                // Give children time to exit cleanly
+                setTimeout(() => {
+                    process.exit(1);
+                }, 1000);
                 return;
             }
             console.log(`[Launcher] Restarting ${name} in 3 seconds... (Attempt ${crashCount}/5)`);
@@ -139,7 +143,11 @@ setTimeout(() => {
     children.push(policy);
     setTimeout(() => {
         // Spawn ML Engine (Python Sidecar)
-        const pythonPath = path_1.default.join(process.env.HOME || process.env.USERPROFILE || '', '.nyxora', 'ml-engine', 'venv', 'bin', 'python');
+        // Cross-platform: Windows uses Scripts/, Unix uses bin/
+        const IS_WINDOWS_LAUNCHER = process.platform === 'win32';
+        const pythonBinDir = IS_WINDOWS_LAUNCHER ? 'Scripts' : 'bin';
+        const pythonExe = IS_WINDOWS_LAUNCHER ? 'python.exe' : 'python';
+        const pythonPath = path_1.default.join(process.env.HOME || process.env.USERPROFILE || '', '.nyxora', 'ml-engine', 'venv', pythonBinDir, pythonExe);
         if (fs_1.default.existsSync(pythonPath)) {
             let mlDir = path_1.default.join(__dirnameResolved, 'packages', 'ml-engine');
             if (!fs_1.default.existsSync(mlDir))
@@ -149,7 +157,7 @@ setTimeout(() => {
             children.push(mlEngine);
         }
         else {
-            console.warn('[Launcher] Warning: Python virtual environment not found. Did you run setup?');
+            console.warn('[Launcher] Warning: Python virtual environment not found. ML Engine features (market analysis etc.) will be unavailable. Run: nyxora setup');
         }
         setTimeout(() => {
             const corePath = path_1.default.join(__dirnameResolved, `packages/core/src/gateway/cli${ext}`);
@@ -157,30 +165,46 @@ setTimeout(() => {
             const core = spawnService('Core', cmd, [...baseArgs, corePath, ...args], env, true);
             children.push(core);
             // --- AUTO-TUNNEL (Cloudflare) ---
+            // Respects config: set cloudflare_tunnel: false in ~/.nyxora/config/config.yaml to disable.
             setTimeout(() => {
-                console.log('[Launcher] Starting Auto-Tunnel (Cloudflare) on port 3000...');
-                const cf = (0, child_process_1.spawn)('npx', ['cloudflared', 'tunnel', '--url', 'http://localhost:3000'], { env, shell: true });
-                children.push({
-                    kill: () => { try {
-                        process.kill(cf.pid, 'SIGTERM');
-                    }
-                    catch (e) { } },
-                    forceKill: () => { try {
-                        process.kill(cf.pid, 'SIGKILL');
-                    }
-                    catch (e) { } }
-                });
-                cf.stderr.on('data', (data) => {
-                    const msg = data.toString();
-                    // Extract the Cloudflare URL (e.g., https://xyz.trycloudflare.com)
-                    const match = msg.match(/https:\/\/[a-zA-Z0-9-]+\.trycloudflare\.com/);
-                    if (match) {
-                        const url = match[0];
-                        console.log(`\n[Auto-Tunnel] Secure Public URL generated: ${url}\n`);
-                        fs_1.default.writeFileSync(path_1.default.join(nyxoraDir, 'public_url.txt'), url, 'utf8');
+                let cfEnabled = true;
+                try {
+                    const yaml = require('yaml');
+                    const configPath = path_1.default.join(nyxoraDir, 'config', 'config.yaml');
+                    const fallbackPath = path_1.default.join(nyxoraDir, 'config.yaml');
+                    const actualPath = fs_1.default.existsSync(configPath) ? configPath : fallbackPath;
+                    if (fs_1.default.existsSync(actualPath)) {
+                        const cfg = yaml.parse(fs_1.default.readFileSync(actualPath, 'utf8'));
+                        if (cfg?.cloudflare_tunnel === false) {
+                            cfEnabled = false;
+                            console.log('[Launcher] Cloudflare Auto-Tunnel is disabled by config (cloudflare_tunnel: false).');
+                        }
                     }
-                    // Optionally print errors if they are real errors, but cloudflared logs everything to stderr
-                });
+                }
+                catch (e) { }
+                if (cfEnabled) {
+                    console.log('[Launcher] Starting Auto-Tunnel (Cloudflare) on port 3000...');
+                    const cf = (0, child_process_1.spawn)('npx', ['cloudflared', 'tunnel', '--url', 'http://localhost:3000'], { env, shell: true });
+                    children.push({
+                        kill: () => { try {
+                            process.kill(cf.pid, 'SIGTERM');
+                        }
+                        catch (e) { } },
+                        forceKill: () => { try {
+                            process.kill(cf.pid, 'SIGKILL');
+                        }
+                        catch (e) { } }
+                    });
+                    cf.stderr.on('data', (data) => {
+                        const msg = data.toString();
+                        const match = msg.match(/https:\/\/[a-zA-Z0-9-]+\.trycloudflare\.com/);
+                        if (match) {
+                            const url = match[0];
+                            console.log(`\n[Auto-Tunnel] Secure Public URL generated: ${url}\n`);
+                            fs_1.default.writeFileSync(path_1.default.join(nyxoraDir, 'public_url.txt'), url, 'utf8');
+                        }
+                    });
+                }
             }, 3000);
         }, 1000);
     }, 1000);
@@ -201,11 +225,17 @@ const cleanup = () => {
             }
             catch (e) { }
         });
-        try {
-            require('child_process').execSync('pkill -f ts-node');
-            require('child_process').execSync('pkill -f uvicorn');
+        // pkill is Unix-only — skip on Windows
+        if (process.platform !== 'win32') {
+            try {
+                require('child_process').execSync('pkill -f ts-node');
+            }
+            catch (e) { }
+            try {
+                require('child_process').execSync('pkill -f uvicorn');
+            }
+            catch (e) { }
         }
-        catch (e) { }
         process.exit(0);
     }, 1000);
 };

package/dist/packages/core/src/agent/llmProvider.js

@@ -8,10 +8,18 @@ class OpenAIAdapter {
     }
     async chat(request) {
         const response = await this.client.chat.completions.create(request);
+        let content = response.choices[0].message.content || '';
+        let reasoning = response.choices[0].message.reasoning_content || null;
+        // Extract <thinking> tags from content if present
+        const thinkingMatch = content.match(/<(think|thought|thinking|reasoning|analysis|reflection)>([\s\S]*?)<\/\1>/i);
+        if (thinkingMatch) {
+            reasoning = (reasoning || '') + thinkingMatch[2].trim();
+            content = content.replace(/<(think|thought|thinking|reasoning|analysis|reflection)>[\s\S]*?<\/\1>\n?/i, '').trim();
+        }
         return {
             message: {
-                content: response.choices[0].message.content,
-                reasoning_content: response.choices[0].message.reasoning_content || null,
+                content: content || null,
+                reasoning_content: reasoning || null,
                 tool_calls: response.choices[0].message.tool_calls
             },
             usage: response.usage ? { total_tokens: response.usage.total_tokens } : undefined
@@ -29,8 +37,8 @@ class OpenAIAdapter {
                     fullContent += delta.content;
                     onChunk(delta.content);
                 }
-                if (delta?.reasoning_content) {
-                    reasoningContent += delta.reasoning_content;
+                if (delta?.reasoning_content || delta?.reasoning) {
+                    reasoningContent += (delta.reasoning_content || delta.reasoning);
                 }
                 if (delta?.tool_calls) {
                     for (const tc of delta.tool_calls) {
@@ -49,6 +57,12 @@ class OpenAIAdapter {
                 }
             }
             const toolCalls = Object.values(toolCallsMap);
+            // Post-process to extract <thinking> tags that were streamed as part of content
+            const thinkingMatch = fullContent.match(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>([\s\S]*?)<\/\1>/i);
+            if (thinkingMatch) {
+                reasoningContent = (reasoningContent || '') + thinkingMatch[2].trim();
+                fullContent = fullContent.replace(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>[\s\S]*?<\/\1>\n?/i, '').trim();
+            }
             return {
                 message: {
                     content: fullContent || null,
@@ -176,10 +190,17 @@ class AnthropicAdapter {
                 });
             }
         }
+        if (contentStr) {
+            const thinkingMatch = contentStr.match(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>([\s\S]*?)<\/\1>/i);
+            if (thinkingMatch) {
+                reasoningStr = (reasoningStr || '') + thinkingMatch[2].trim();
+                contentStr = contentStr.replace(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>[\s\S]*?<\/\1>\n?/i, '').trim();
+            }
+        }
         return {
             message: {
-                content: contentStr,
-                reasoning_content: reasoningStr,
+                content: contentStr || null,
+                reasoning_content: reasoningStr || null,
                 tool_calls: toolCalls.length > 0 ? toolCalls : undefined
             },
             usage: response.usage ? { total_tokens: response.usage.input_tokens + response.usage.output_tokens } : undefined
@@ -264,6 +285,13 @@ class AnthropicAdapter {
                         last.function.arguments += event.delta.partial_json;
                 }
             }
+            if (fullContent) {
+                const thinkingMatch = fullContent.match(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>([\s\S]*?)<\/\1>/i);
+                if (thinkingMatch) {
+                    reasoningContent = (reasoningContent || '') + thinkingMatch[2].trim();
+                    fullContent = fullContent.replace(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>[\s\S]*?<\/\1>\n?/i, '').trim();
+                }
+            }
             return { message: { content: fullContent || null, reasoning_content: reasoningContent || null, tool_calls: toolCalls.length > 0 ? toolCalls : undefined } };
         }
         catch {
@@ -406,9 +434,18 @@ class GeminiAdapter {
         if (data.usageMetadata && data.usageMetadata.totalTokenCount) {
             totalTokens = data.usageMetadata.totalTokenCount;
         }
+        let reasoningContent = null;
+        if (contentStr) {
+            const thinkingMatch = contentStr.match(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>([\s\S]*?)<\/\1>/i);
+            if (thinkingMatch) {
+                reasoningContent = thinkingMatch[2].trim();
+                contentStr = contentStr.replace(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>[\s\S]*?<\/\1>\n?/i, '').trim();
+            }
+        }
         return {
             message: {
-                content: contentStr,
+                content: contentStr || null,
+                reasoning_content: reasoningContent || null,
                 tool_calls: toolCalls.length > 0 ? toolCalls : undefined
             },
             usage: totalTokens > 0 ? { total_tokens: totalTokens } : undefined
@@ -525,8 +562,16 @@ class GeminiAdapter {
                     }
                 }
             }
+            let reasoningContent = null;
+            if (contentStr) {
+                const thinkingMatch = contentStr.match(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>([\s\S]*?)<\/\1>/i);
+                if (thinkingMatch) {
+                    reasoningContent = thinkingMatch[2].trim();
+                    contentStr = contentStr.replace(/<(think|thought|thinking|reasoning|analysis|reflection|ant-thinking|ant_thinking)[^>]*>[\s\S]*?<\/\1>\n?/i, '').trim();
+                }
+            }
             return {
-                message: { content: contentStr || null, tool_calls: toolCalls.length > 0 ? toolCalls : undefined },
+                message: { content: contentStr || null, reasoning_content: reasoningContent || null, tool_calls: toolCalls.length > 0 ? toolCalls : undefined },
                 usage: totalTokens > 0 ? { total_tokens: totalTokens } : undefined
             };
         }

package/dist/packages/core/src/agent/nyxDaemon.js

@@ -55,7 +55,7 @@ class NyxDaemon {
                 this.isProcessing = false;
                 return;
             }
-            // Kirim riwayat percakapan ke Python ML Engine untuk diproses oleh LangChain
+            // Send conversation history to Python ML Engine for cognitive modeling
             const res = await fetch('http://127.0.0.1:8000/cognitive/reason', {
                 method: 'POST',
                 headers: { 'Content-Type': 'application/json' },