Update app.py

app.py

@@ -15,10 +15,23 @@ from PIL import Image, ImageOps
 
 DEFAULT_MODEL = os.getenv("VISION_MODEL", "qwen/qwen3.6-27b")
 DEFAULT_PROMPT = (
-    "Write concise, useful alt text for this image. Describe only meaningful visual "
-    "content and function. Do not begin with 'image of' or 'picture of'. Do not guess "
-    "unclear identities or details. Include visible text only when important. Return "
-    "only the alt text, preferably under 150 characters."
+"""
+You create accessibility alt text for images.
+
+Return only the finished alt text.
+
+Requirements:
+- Write one concise sentence.
+- Describe the image's essential meaning and purpose.
+- Do not describe every minor visual detail.
+- Do not begin with "Image of", "Picture of", or "This image shows".
+- Do not include analysis, reasoning, headings, XML tags, or bullet points.
+- Do not include <think> tags.
+- Do not identify people unless their identity is explicitly provided.
+- For diagrams, summarize the main process or relationship.
+- For charts, state the chart type and principal trend or conclusion.
+- Keep the response under 250 characters when practical.
+"""
 )
 
 TFLIVE_RE = re.compile(
@@ -174,66 +187,27 @@ Requirements:
 - Keep the response under 250 characters when practical.
 """
 
-def generate_alt_text(client, image_data_uri, prompt=None):
-    system_prompt = """
-You are an accessibility specialist writing alt text.
-
-Return exactly one plain-text alt description and nothing else.
-
-Rules:
-1. Do not expose reasoning or analysis.
-2. Do not use <think> tags.
-3. Do not add a heading or label.
-4. Do not begin with "Image of", "Picture of", or "This image shows".
-5. Describe the image's essential information and function.
-6. For diagrams, explain the main process or relationship rather than listing
-   every label.
-7. Include visible text only when it is necessary to understand the image.
-8. Do not speculate about unclear content.
-9. Aim for one sentence and fewer than 250 characters.
-"""
-
-    user_prompt = prompt or (
-        "Write accessible alt text for this image. "
-        "Return only the finished alt text."
-    )
-
-    response = client.chat.completions.create(
-        model="qwen/qwen3.6-27b",
+def generate_alt_text(client, model, prompt, png_bytes, max_tokens):
+    completion = client.chat.completions.create(
+        model=model,
         reasoning_effort="none",
         messages=[
-            {
-                "role": "system",
-                "content": system_prompt
-            },
             {
                 "role": "user",
                 "content": [
-                    {
-                        "type": "text",
-                        "text": user_prompt
-                    },
+                    {"type": "text", "text": prompt},
                     {
                         "type": "image_url",
-                        "image_url": {
-                            "url": image_data_uri
-                        }
-                    }
-                ]
+                        "image_url": {"url": _data_uri(png_bytes)},
+                    },
+                ],
             }
         ],
+        max_tokens=int(max_tokens),
         temperature=0,
-        max_tokens=100
     )
+    return (completion.choices[0].message.content or "").strip()
 
-    raw_text = response.choices[0].message.content or ""
-    cleaned_text = clean_alt_text(raw_text)
-
-    return {
-        "alt_text": cleaned_text,
-        "model_used": "qwen/qwen3.6-27b",
-        "error": ""
-    }
 
 
 def process_table(
@@ -380,4 +354,4 @@ Upload a CSV, TSV, or Excel table containing one image value per row. The app re
     )
 
 if __name__ == "__main__":
-    demo.queue(default_concurrency_limit=2).launch()
+    demo.queue(default_concurrency_limit=2).launch()