sygal-cli 0.3.1 → 0.3.5

package/CHANGELOG.md

@@ -9,6 +9,21 @@ partagent toujours le même numéro de version, publiés ensemble, même si un
 seul a réellement changé de contenu — même principe que l'app desktop.
 Détails du process : `CLI_RELEASE_GUIDE.md` à la racine du repo.
 
+## 0.3.1
+
+- **Corrige la 0.3.0**, publiée avec un bundle `sygal-cli` (dist/index.js)
+  périmé — construit avant les changements ci-dessous, donc `.rtf` restait
+  "Extension non supportée" et aucun des correctifs OCR n'était réellement
+  présent malgré le numéro de version. Ajout d'un hook `prepublishOnly`
+  (rebuild automatique avant toute publication) pour que ça ne puisse plus
+  se reproduire. **Si vous avez installé la 0.3.0, mettez à jour vers la
+  0.3.1.**
+- Nouveau : support des formats **AVIF** et **RTF** (texte brut, tableaux
+  RTF rendus lisibles ligne par ligne).
+- Fix : messages d'erreur clairs pour HEIC/HEIF (HEVC non pris en charge),
+  AVIF "séquence"/rafale, et WebP envoyé au fournisseur NVIDIA — au lieu
+  d'erreurs génériques ou d'un faux "erreur réseau".
+
 ## 0.2.1
 
 - Notice de mise à jour après chaque conversion (terminal interactif

package/dist/index.js

@@ -71,6 +71,10 @@ var SUPPORTED_EXTENSIONS = {
   heif: "image/heif",
   avif: "image/avif"
 };
+var PACKAGE_DIR_EXTENSIONS = /* @__PURE__ */ new Set(["epub"]);
+function isPackageDirExtension(extension) {
+  return PACKAGE_DIR_EXTENSIONS.has(extension.toLowerCase().replace(/^\./, ""));
+}
 function mimeForExtension(extension) {
   return SUPPORTED_EXTENSIONS[extension.toLowerCase()] ?? null;
 }
@@ -121,10 +125,16 @@ var OCR_MODELS = {
   openai: "gpt-4o",
   mistral: "mistral-ocr-latest",
   claude: "claude-sonnet-4-6",
-  // Bascule 18/07/2026 (test comparatif, validation Cyrill) : nemotron nano VL
-  // remplace meta/llama-3.2-11b-vision-instruct — 0/5 confabulation sur portrait
-  // sans texte (contre 5/5), français fidèle, plus rapide et déterministe.
-  nvidia: "nvidia/llama-3.1-nemotron-nano-vl-8b-v1",
+  // Bascule 27/07/2026 (banc comparatif sur images réelles, validation Cyrill) :
+  // nemotron-nano-12b-v2-vl remplace llama-3.1-nemotron-nano-vl-8b-v1. Sur une
+  // capture de dossier de 22 lignes, le 8b s'arrêtait au premier en-tête (« Nom »,
+  // 1 token) 3 fois sur 3 ; le 12b la transcrit intégralement 3 fois sur 3, accents
+  // préservés. Il respecte aussi la consigne de langue sur une image sans texte
+  // (description en français, là où le 8b répondait en anglais), sans confabuler.
+  // Entraîné sur DocLayNet / PubTables-1M / PDF CommonCrawl — OCRBench 85,6.
+  // Précédemment : llama-3.2-11b-vision-instruct, écarté le 18/07/2026 (5/5
+  // confabulations sur un portrait sans texte).
+  nvidia: "nvidia/nemotron-nano-12b-v2-vl",
   gemini: "gemini-2.0-flash"
 };
 var MISTRAL_TRANSLATE_MODEL = "mistral-large-latest";
@@ -410,7 +420,7 @@ var LOCALE_LANGUAGE = {
   es: "Spanish"
 };
 function buildOcrPrompt(labelLanguage = "French") {
-  return `Transcribe ALL visible text in this image verbatim as clean Markdown, preserving the document layout: headings, lists and tables exactly as shown. Transcribe EVERY line, row and table cell faithfully \u2014 never summarize, skip, merge, reorder, or invent rows, values, or text. If a cell is empty, leave it empty. Keep the document text in its ORIGINAL language exactly as written \u2014 do not translate it. Do NOT invent a table of contents, section titles, or any heading/structure that is not visually present in the document. If you must add a label of your own that is not in the document, write it in ${labelLanguage}. While transcribing, do NOT describe the scene, the people, or any visual elements that are not text. ONLY IF the image contains no readable text at all: instead of an empty document, describe the image in ${labelLanguage} as clean Markdown \u2014 one short paragraph covering the subject, the setting and the dominant colors, factual and concise, without inventing any text that is not visible. When describing, NEVER invent or guess names, ages, occupations, employers, biographies, locations, dates, or any information about people or organizations. Describe only what is physically visible in the image; if something is not visible, do not mention it. Output only the Markdown.`;
+  return `Transcribe ALL visible text in this image verbatim as clean Markdown, preserving the document layout: headings, lists and tables exactly as shown. Render every table as a GitHub-flavored Markdown table using pipes (| \u2026 | \u2026 |) with a header separator row \u2014 NEVER output LaTeX, never use \\begin{tabular}, \\multicolumn or any TeX command. Transcribe EVERY line, row and table cell faithfully \u2014 never summarize, skip, merge, reorder, or invent rows, values, or text. If a cell is empty, leave it empty. Keep the document text in its ORIGINAL language exactly as written \u2014 do not translate it. Do NOT invent a table of contents, section titles, or any heading/structure that is not visually present in the document. If you must add a label of your own that is not in the document, write it in ${labelLanguage}. While transcribing, do NOT describe the scene, the people, or any visual elements that are not text. ONLY IF the image contains no readable text at all: instead of an empty document, describe the image in ${labelLanguage} as clean Markdown \u2014 one short paragraph covering the subject, the setting and the dominant colors, factual and concise, without inventing any text that is not visible. When describing, NEVER invent or guess names, ages, occupations, employers, biographies, locations, dates, or any information about people or organizations. Describe only what is physically visible in the image; if something is not visible, do not mention it. Output only the Markdown.`;
 }
 function buildTranslatePrompt(targetLocale) {
   const language = LOCALE_LANGUAGE[targetLocale] ?? "English";
@@ -668,6 +678,30 @@ var mistralProvider = {
   }
 };
 
+// ../core/src/ocr/latex-tables.ts
+var TABULAR_RE = /\\begin\{tabular\}(?:\{[^}]*\})?([\s\S]*?)\\end\{tabular\}/g;
+function cleanCell(cell) {
+  return cell.replace(/\\multicolumn\{\d+\}\{[^}]*\}\{([\s\S]*?)\}/g, "$1").replace(/\\multirow\{\d+\}\{[^}]*\}\{([\s\S]*?)\}/g, "$1").replace(/\\textbf\{([\s\S]*?)\}/g, "**$1**").replace(/\\textit\{([\s\S]*?)\}/g, "*$1*").replace(/\\(?:hline|toprule|midrule|bottomrule)\b/g, "").replace(/\\\\/g, "").replace(/\|/g, "\\|").replace(/\s+/g, " ").trim();
+}
+function tabularToMarkdown(body) {
+  const rows = body.split(/\\\\/).map((row) => row.trim()).filter((row) => row.replace(/\\(?:hline|toprule|midrule|bottomrule)\b/g, "").trim().length > 0).map((row) => row.split("&").map(cleanCell)).filter((cells) => cells.some((cell) => cell.length > 0));
+  if (rows.length === 0) return null;
+  const columns = Math.max(...rows.map((row) => row.length));
+  if (columns < 2) return rows.map((row) => row[0]).join("\n\n");
+  const pad = (cells) => `| ${[...cells, ...Array(columns - cells.length).fill("")].join(" | ")} |`;
+  const [header, ...body_] = rows;
+  return [pad(header), `| ${Array(columns).fill("---").join(" | ")} |`, ...body_.map(pad)].join(
+    "\n"
+  );
+}
+function convertLatexTables(markdown) {
+  if (!markdown.includes("\\begin{tabular}")) return markdown;
+  return markdown.replace(TABULAR_RE, (whole, body) => {
+    const converted = tabularToMarkdown(body);
+    return converted ?? whole;
+  });
+}
+
 // ../core/src/ocr/nvidia-provider.ts
 var NVIDIA_OCR_MODEL = OCR_MODELS.nvidia;
 var ENDPOINT4 = "https://integrate.api.nvidia.com/v1/chat/completions";
@@ -812,6 +846,28 @@ function collapseRepeatedLines(markdown) {
   }
   return kept.join("\n");
 }
+var TRUNCATED_COMPLETION_TOKENS = 3;
+function completionTokensOf(raw) {
+  if (typeof raw !== "object" || raw === null) return null;
+  const usage = raw.usage;
+  if (typeof usage !== "object" || usage === null) return null;
+  const value = usage.completion_tokens ?? usage.output_tokens;
+  return typeof value === "number" ? value : null;
+}
+function looksTruncated(result) {
+  const tokens = completionTokensOf(result.raw);
+  return tokens !== null && tokens <= TRUNCATED_COMPLETION_TOKENS;
+}
+function stripEnclosingCodeFence(markdown) {
+  const trimmed = markdown.trim();
+  const opening = /^```([a-zA-Z]*)\r?\n?/.exec(trimmed);
+  if (!opening || !trimmed.endsWith("```") || trimmed.length < 6) return markdown;
+  const language = opening[1].toLowerCase();
+  if (language && language !== "markdown" && language !== "md") return markdown;
+  const inner = trimmed.slice(opening[0].length, -3);
+  if (inner.includes("```")) return markdown;
+  return inner.trim();
+}
 var REFUSAL_RE = /^(i['’]?m sorry|i cannot|i can['’]?t|i am (unable|not able)|sorry,|je suis désolé|désolé,|je ne peux pas|lo siento|mi dispiace|es tut mir leid|ich kann)/i;
 function isProviderRefusal(markdown) {
   return REFUSAL_RE.test(markdown.trim().slice(0, 80));
@@ -885,13 +941,23 @@ var ApiOcrRunner = class {
       if (!apiKey2) {
         throw new ConversionError("API_KEY_MISSING", `Aucune cl\xE9 API pour ${custom.name}`);
       }
-      const result2 = await this.callCustomWithRetry(custom, img, sendMime, apiKey2, prompt);
+      let result2 = await this.callCustomWithRetry(custom, img, sendMime, apiKey2, prompt);
+      if (looksTruncated(result2) && !isProviderRefusal(result2.markdown)) {
+        const second = await this.callCustomWithRetry(custom, img, sendMime, apiKey2, prompt);
+        if (second.markdown.trim().length > result2.markdown.trim().length) result2 = second;
+      }
       if (isProviderRefusal(result2.markdown)) {
         throw new ConversionError("OCR_REFUSED", `${custom.name} : refus du mod\xE8le`);
       }
       return {
-        markdown: collapseRepeatedLines(result2.markdown),
-        meta: { provider: custom.name, model: custom.model }
+        markdown: convertLatexTables(
+          stripEnclosingCodeFence(collapseRepeatedLines(result2.markdown))
+        ),
+        meta: {
+          provider: custom.name,
+          model: custom.model,
+          ...looksTruncated(result2) ? { truncated: true } : {}
+        }
       };
     }
     const providerId = settings.ocrProvider;
@@ -903,13 +969,24 @@ var ApiOcrRunner = class {
         `Aucune cl\xE9 API enregistr\xE9e pour le provider ${providerId}`
       );
     }
-    const result = await this.callWithRetry(provider, img, sendMime, apiKey, prompt);
+    let result = await this.callWithRetry(provider, img, sendMime, apiKey, prompt);
+    if (looksTruncated(result) && !isProviderRefusal(result.markdown)) {
+      const second = await this.callWithRetry(provider, img, sendMime, apiKey, prompt);
+      if (second.markdown.trim().length > result.markdown.trim().length) result = second;
+    }
     if (isProviderRefusal(result.markdown)) {
       throw new ConversionError("OCR_REFUSED", `${provider.id} : refus du mod\xE8le`);
     }
     return {
-      markdown: collapseRepeatedLines(result.markdown),
-      meta: { provider: provider.id, model: provider.model }
+      markdown: convertLatexTables(stripEnclosingCodeFence(collapseRepeatedLines(result.markdown))),
+      meta: {
+        provider: provider.id,
+        model: provider.model,
+        // Rejeu compris, le modèle s'est arrêté au premier libellé : le
+        // ConversionService en fait un avertissement dans le journal plutôt
+        // que d'enregistrer silencieusement une transcription amputée.
+        ...looksTruncated(result) ? { truncated: true } : {}
+      }
     };
   }
   /** Tentative initiale + retries backoff sur erreurs retryable uniquement */
@@ -1212,7 +1289,7 @@ var TranslationRunner = class {
       const timer = setTimeout(() => controller.abort(), this.timeoutMs);
       try {
         const result = await attempt(controller.signal);
-        return collapseRepeatedLines(result.markdown);
+        return convertLatexTables(stripEnclosingCodeFence(collapseRepeatedLines(result.markdown)));
       } catch (err) {
         if (controller.signal.aborted) {
           throw new ConversionError("TIMEOUT", `${providerLabel} : d\xE9lai de traduction d\xE9pass\xE9`);
@@ -1667,7 +1744,7 @@ async function expandInputs(inputs) {
       });
       continue;
     }
-    if (stat.isDirectory()) {
+    if (stat.isDirectory() && !isPackageDirExtension((0, import_node_path5.extname)(path).slice(1))) {
       await collectFromDir(path, jobs);
     } else {
       jobs.push({ input: path, kind: "file" });
@@ -1687,7 +1764,11 @@ async function collectFromDir(dir, jobs) {
     if (entry.name.startsWith(".")) continue;
     const full = (0, import_node_path5.join)(dir, entry.name);
     if (entry.isDirectory()) {
-      await collectFromDir(full, jobs);
+      if (isPackageDirExtension((0, import_node_path5.extname)(entry.name).slice(1))) {
+        jobs.push({ input: full, kind: "file" });
+      } else {
+        await collectFromDir(full, jobs);
+      }
     } else if (entry.isFile() && mimeForExtension((0, import_node_path5.extname)(entry.name).slice(1))) {
       jobs.push({ input: full, kind: "file" });
     }
@@ -2700,7 +2781,7 @@ Exemples :
     let singleFile = false;
     if (paths.length === 1 && !/^https?:\/\//i.test(paths[0])) {
       try {
-        singleFile = (0, import_node_fs8.statSync)(paths[0]).isFile();
+        singleFile = (0, import_node_fs8.statSync)(paths[0]).isFile() || isPackageDirExtension((0, import_node_path10.extname)(paths[0]).slice(1));
       } catch {
         singleFile = true;
       }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "sygal-cli",
-  "version": "0.3.1",
+  "version": "0.3.5",
   "description": "Sygal CLI - convert files and web pages to Markdown/JSON (local MarkItDown pipeline, OCR for images)",
   "author": "Cyrill Semah",
   "license": "SEE LICENSE IN LICENSE",
@@ -37,9 +37,9 @@
     "@sygal/core": "0.1.0"
   },
   "optionalDependencies": {
-    "@sygal/sidecar-darwin-arm64": "0.3.1",
-    "@sygal/sidecar-win32-x64": "0.3.1",
-    "@sygal/sidecar-linux-x64": "0.3.1"
+    "@sygal/sidecar-win32-x64": "0.3.5",
+    "@sygal/sidecar-darwin-arm64": "0.3.5",
+    "@sygal/sidecar-linux-x64": "0.3.5"
   },
   "scripts": {
     "build": "esbuild src/index.ts --bundle --platform=node --target=node20 --format=cjs --outfile=dist/index.js --external:sharp --external:gpt-tokenizer --external:commander --external:picocolors",