Input and results

check_connection.py

@@ -0,0 +1,15 @@
+import requests
+
+url = "http://127.0.0.1:11434/api/generate"
+
+payload = {
+    "model": "llama3.2-vision",
+    "prompt": "Csak ezt írd vissza: ok",
+    "stream": False
+}
+
+r = requests.post(url, json=payload, timeout=60)
+r.raise_for_status()
+
+data = r.json()
+print(data["response"])

main.py

@@ -2,7 +2,7 @@ from pathlib import Path
 import base64
 import requests
 
-OLLAMA_URL = "http://localhost:11434/api/chat"
+OLLAMA_URL = "http://127.0.0.1:11434/api/chat"
 MODEL = "llama3.2-vision"
 IMAGE_EXTS = {".png", ".jpg", ".jpeg", ".webp", ".bmp", ".gif"}
 
@@ -28,18 +28,22 @@ def ocr_image(path: Path) -> str:
             }
         ]
     }
-    r = requests.post(OLLAMA_URL, json=payload, timeout=300)
+    
+    r = requests.post(OLLAMA_URL, json=payload, timeout=900000)
     r.raise_for_status()
     data = r.json()
+    print(data)
     return data.get("message", {}).get("content", "").strip()
 
-def process_folder(folder: str):
+def process_folder(folder: str, result : str):
     folder_path = Path(folder)
+    result_path = Path(result)
     for img in folder_path.iterdir():
         if img.is_file() and img.suffix.lower() in IMAGE_EXTS:
             print(f"Processing: {img.name}")
             text = ocr_image(img)
-            img.with_suffix(".txt").write_text(text, encoding="utf-8")
+            out_file = result_path / f"{img.stem}.txt"
+            out_file.write_text(text, encoding="utf-8")
 
 if __name__ == "__main__":
-    process_folder("kepek")
+    process_folder("images", "output")

output/a.txt

@@ -0,0 +1 @@
+The text in the image is: "Because be doing".

output/b.txt

@@ -0,0 +1 @@
+The text in the image is "Congratulations".