copy path and chat about file/folder

src/servers/api/file-browser/router.ts

@@ -5,7 +5,9 @@ import { existsSync } from 'node:fs';
 import { homedir } from 'node:os';
 import { getHomeDir, DATA_PATH, getUserSettingsFile } from '@@/data-path';
 import * as errors from '@@/custom-errors';
-import { readConfig } from '@@/api/server-settings/resources';
+import { readTtsConfig } from '@@/api/server-settings/tts';
+import { readSttConfig } from '@@/api/server-settings/stt';
+import { readOcrConfig } from '@@/api/server-settings/ocr';
 
 const DEFAULT_HOME_DIRS = ['Downloads', 'Documents', 'Music', 'Video', 'Pictures', 'Onboarding'];
 const ONBOARDING_SEED = join(DATA_PATH, 'Onboarding');
@@ -81,7 +83,7 @@ router.get('/ls', async (ctx) => {
   );
 
   const path = '/' + absPath.slice(rootDir.length).replace(/^\/+/, '');
-  return ctx.json({ path, entries: entries.filter(Boolean) });
+  return ctx.json({ path, rootDir, entries: entries.filter(Boolean) });
 });
 
 // Read file contents
@@ -267,6 +269,33 @@ router.get('/transcode', async (ctx) => {
   });
 });
 
+// Save a cached result (ocr/tts/transcriptions/audio) next to the original file
+const CACHE_PREFIXES = ['ocr/', 'tts/', 'transcriptions/', 'audio/'];
+
+router.post('/save-result', async (ctx) => {
+  const user = ctx.get('user');
+  const { path: cachedPath } = ctx.get('body') as { path: string };
+  if (!cachedPath) throw errors.BAD_REQUEST('path is required');
+
+  const prefix = CACHE_PREFIXES.find((p) => cachedPath.startsWith(p));
+  if (!prefix) throw errors.BAD_REQUEST('Not a cached result path');
+
+  const relativePath = cachedPath.slice(prefix.length);
+  const userDataDir = getUserDataDir(user.email);
+  const srcAbs = resolve(userDataDir, cachedPath);
+  if (!existsSync(srcAbs)) throw errors.BAD_REQUEST('Cached file not found');
+
+  const homeDir = getHomeDir(user.email);
+  const destAbs = resolve(homeDir, relativePath);
+  if (!destAbs.startsWith(homeDir)) throw errors.FORBIDDEN('Path outside home directory');
+
+  await mkdir(dirname(destAbs), { recursive: true });
+  await cp(srcAbs, destAbs);
+
+  const destPath = '/' + destAbs.slice(homeDir.length).replace(/^\/+/, '');
+  return ctx.json({ savedPath: destPath });
+});
+
 // Text-to-speech with caching
 router.post('/tts', async (ctx) => {
   const user = ctx.get('user');
@@ -287,16 +316,29 @@ router.post('/tts', async (ctx) => {
     return ctx.json({ audioPath: cacheRel, audioRoot: 'user-data' });
   }
 
-  const config = await readConfig();
-  const kokoroUrl = config.kokoro?.url;
-  if (!kokoroUrl) throw errors.BAD_REQUEST('Kokoro TTS not configured');
+  const ttsConfig = await readTtsConfig();
+  if (!ttsConfig) throw errors.BAD_REQUEST('TTS not configured — set it up in Settings → Text to Speech');
 
   const content = await readFile(absPath, 'utf-8');
-  const res = await fetch(`${kokoroUrl}/v1/audio/speech`, {
-    method: 'POST',
-    headers: { 'Content-Type': 'application/json' },
-    body: JSON.stringify({ model: 'kokoro', input: content, voice: 'af_heart', response_format: 'mp3' }),
-  });
+
+  const headers: Record<string, string> = { 'Content-Type': 'application/json' };
+  let res: Response;
+
+  if (ttsConfig.provider === 'elevenlabs') {
+    if (!ttsConfig.apiKey) throw errors.BAD_REQUEST('ElevenLabs API key not configured');
+    res = await fetch(`https://api.elevenlabs.io/v1/text-to-speech/${encodeURIComponent(ttsConfig.voice)}`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json', 'xi-api-key': ttsConfig.apiKey },
+      body: JSON.stringify({ text: content, model_id: ttsConfig.model }),
+    });
+  } else {
+    if (ttsConfig.apiKey) headers['Authorization'] = `Bearer ${ttsConfig.apiKey}`;
+    res = await fetch(`${ttsConfig.url.replace(/\/+$/, '')}/v1/audio/speech`, {
+      method: 'POST',
+      headers,
+      body: JSON.stringify({ model: ttsConfig.model, input: content, voice: ttsConfig.voice, response_format: 'mp3' }),
+    });
+  }
   if (!res.ok) throw errors.BAD_REQUEST('TTS request failed');
 
   await mkdir(dirname(cacheAbs), { recursive: true });
@@ -327,26 +369,46 @@ router.post('/ocr', async (ctx) => {
     return ctx.json({ text, ocrPath: cacheRel, ocrRoot: 'user-data', cached: true });
   }
 
-  const config = await readConfig();
-  const llamaUrl = config.llama?.url;
-  if (!llamaUrl) throw errors.BAD_REQUEST('llama.cpp not configured');
+  const ocrConfig = await readOcrConfig();
+  if (!ocrConfig) throw errors.BAD_REQUEST('OCR not configured — set it up in Settings → OCR');
 
   const imageBytes = await Bun.file(absPath).arrayBuffer();
   const base64 = Buffer.from(imageBytes).toString('base64');
   const ext = absPath.split('.').pop()?.toLowerCase() ?? 'png';
   const mime = ext === 'jpg' || ext === 'jpeg' ? 'image/jpeg' : ext === 'webp' ? 'image/webp' : `image/${ext}`;
 
-  const res = await fetch(`${llamaUrl}/v1/chat/completions`, {
+  const res = await fetch(`${ocrConfig.url.replace(/\/+$/, '')}/v1/chat/completions`, {
     method: 'POST',
     headers: { 'Content-Type': 'application/json' },
     body: JSON.stringify({
-      model: 'Qwen2.5-VL-7B-Instruct-q4_k_m.gguf',
+      model: ocrConfig.model,
       messages: [
+        {
+          role: 'system',
+          content: [
+            'You are an OCR assistant. Extract meaningful text content from images.',
+            'Rules:',
+            '- Output ONLY the extracted text, no commentary or explanations.',
+            '- For documents, articles, books: preserve the original text, paragraphs, and structure as markdown.',
+            '- For tables: use markdown table format.',
+            '- For code/terminal screenshots: use fenced code blocks.',
+            '- For social media posts/threads: extract as clean conversation. Format as:',
+            '  **username** says: "their text"',
+            '  **replier** replies: "their text"',
+            '  Strip all UI chrome (follow buttons, timestamps, like counts, avatars, "Everybody can reply", etc).',
+            '  Keep only usernames and what they actually wrote.',
+            '- For memes/image macros: describe the image briefly, then extract any text.',
+            '- For handwriting: transcribe as accurately as possible.',
+            '- For receipts/invoices: extract as structured text with line items.',
+            '- Strip all UI elements, navigation, ads, watermarks, and other noise.',
+            '- Preserve the original language of the text.',
+          ].join('\n'),
+        },
         {
           role: 'user',
           content: [
             { type: 'image_url', image_url: { url: `data:${mime};base64,${base64}` } },
-            { type: 'text', text: 'Extract all text from this image. Return only the extracted text, nothing else.' },
+            { type: 'text', text: 'Extract the text from this image.' },
           ],
         },
       ],
@@ -507,9 +569,9 @@ router.post('/transcribe', async (ctx) => {
     return ctx.json({ transcriptionPath: cacheRel, transcriptionRoot: 'user-data', cached: true });
   }
 
-  const config = await readConfig();
-  const whisperUrl = config.whisper?.url;
-  if (!whisperUrl) throw errors.BAD_REQUEST('Whisper not configured');
+  const sttConfig = await readSttConfig();
+  const whisperUrl = sttConfig?.url;
+  if (!whisperUrl) throw errors.BAD_REQUEST('Whisper not configured — set it up in Settings → Speech to Text');
 
   const audioFile = Bun.file(absPath);