gdfgdf

vision_runner.ts

@@ -47,7 +47,7 @@ interface Args {
   api: string;
   limit: number;
   only: string | null;
-  dpi: number;
+  dpi: number | "auto";
   maxPages: number;
   force: boolean;
   outDir: string;
@@ -73,7 +73,10 @@ function parseArgs(): Args {
     api,
     limit: parseInt(get("--limit", "0")!, 10) || 0,
     only: get("--only"),
-    dpi: parseInt(get("--dpi", "150")!, 10) || 150,
+    dpi: ((): number | "auto" => {
+      const v = get("--dpi", "auto")!;
+      return v === "auto" ? "auto" : parseInt(v, 10) || 150;
+    })(),
     maxPages: parseInt(get("--max-pages", "8")!, 10) || 8,
     force: a.includes("--force"),
     outDir: get("--out-dir", path.dirname(input))!,
@@ -234,6 +237,19 @@ async function renderPdf(pdfPath: string, dpi: number, maxPages: number, tmpDir:
   return files;
 }
 
+/** Hat das PDF eine nennenswerte Textebene? (Hybrid: Werte getippt → 150 DPI
+ *  reicht. Reiner Scan: keine Textebene, Handschrift möglich → 200 DPI.) */
+async function hasTextLayer(pdfPath: string, maxPages: number): Promise<boolean> {
+  try {
+    const { stdout } = await execFileP("pdftotext", ["-l", String(maxPages), pdfPath, "-"], {
+      timeout: 30_000, maxBuffer: 10 * 1024 * 1024,
+    });
+    return stdout.replace(/\s+/g, "").length > 100;
+  } catch {
+    return false; // im Zweifel als Scan behandeln → hohe Auflösung
+  }
+}
+
 // ---------------------------------------------------------------------------
 // VLM-Aufruf (llama-server, OpenAI-kompatibel, JSON-Schema)
 // ---------------------------------------------------------------------------
@@ -319,6 +335,22 @@ async function fetchModelId(api: string): Promise<string> {
   }
 }
 
+/** Wartet nach einem Server-Neustart (503/Verbindungsfehler), bis /health
+ *  wieder OK meldet — Modell-Reload dauert 1-2 Minuten. */
+async function waitForHealthy(api: string, timeoutMs: number): Promise<void> {
+  const healthUrl = api.replace(/\/v1\/?$/, "") + "/health";
+  const deadline = Date.now() + timeoutMs;
+  while (Date.now() < deadline) {
+    try {
+      const r = await fetch(healthUrl, { signal: AbortSignal.timeout(5000) });
+      if (r.ok) return;
+    } catch {
+      /* Server noch weg */
+    }
+    await new Promise((res) => setTimeout(res, 5000));
+  }
+}
+
 async function callVision(api: string, images: string[], timeoutMs: number): Promise<VisionRaw> {
   const content: Array<Record<string, unknown>> = [{ type: "text", text: USER_PROMPT }];
   for (const img of images) {
@@ -481,8 +513,11 @@ async function main(): Promise<void> {
 
     const tmpDir = await fsp.mkdtemp(path.join(os.tmpdir(), "bvs-"));
     try {
-      progress(`${tag} … rendere`);
-      const images = await renderPdf(pdfPath, args.dpi, args.maxPages, tmpDir);
+      const dpi = args.dpi === "auto"
+        ? ((await hasTextLayer(pdfPath, args.maxPages)) ? 150 : 200)
+        : args.dpi;
+      progress(`${tag} … rendere (${dpi} dpi)`);
+      const images = await renderPdf(pdfPath, dpi, args.maxPages, tmpDir);
 
       let vis: VisionRaw | null = null;
       let lastErr = "";
@@ -493,7 +528,15 @@ async function main(): Promise<void> {
           break;
         } catch (e) {
           lastErr = e instanceof Error ? e.message : String(e);
-          if (attempt < RETRIES) await new Promise((res) => setTimeout(res, 5000 * attempt));
+          if (attempt < RETRIES) {
+            if (/HTTP 50[23]|fetch failed|aborted|ECONNREFUSED|ECONNRESET/i.test(lastErr)) {
+              // Server crasht/lädt neu → auf /health warten (Modell-Reload dauert)
+              progress(`${tag} … Server neu am Laden, warte auf /health`);
+              await waitForHealthy(args.api, 300_000);
+            } else {
+              await new Promise((res) => setTimeout(res, 5000 * attempt));
+            }
+          }
         }
       }