Update app.py

app.py

@@ -94,45 +94,61 @@ def analyze_and_enhance(image, style, language, paper_file):
     if image is None:
         return None, "请先上传图表。", "", None, "", ""
 
-    # Step 0: 论文上下文
+    # ===== 0. 论文上下文 =====
     full_text = extract_full_text_from_pdf(paper_file)
-    # 先随便给一个默认 caption（用来抽关键字）
-    raw_caption_tmp = vision_pipe(image)[0]["generated_text"]
+    raw_caption_tmp = vision_pipe(image)[0]["generated_text"]  # 只当内部信息
     snippet = select_relevant_snippets(full_text, raw_caption_tmp) if full_text else ""
     paper_context = summarise_context_for_figure(snippet) if snippet else ""
 
-    # Step 2: Vision → 粗略描述（我们主要用来识别“这是柱状图”）
-    raw_caption = raw_caption_tmp
-    rc = raw_caption.lower()
-    if not any(k in rc for k in ["bar", "chart", "graph", "plot"]):
-        raw_caption = "a bar chart showing the response of different experimental groups under several treatments"
+    # ===== 1. 用论文上下文 + 常识来写 Step 2 的描述 =====
+    if paper_context:
+        figdesc_prompt = (
+            "You are given the context of a scientific study.\n\n"
+            f"{paper_context}\n\n"
+            "In ONE sentence starting with 'A bar chart showing', describe what a "
+            "typical results figure in this study would plot on the horizontal and "
+            "vertical axes and which treatments or groups are compared.\n\n"
+            "Sentence:"
+        )
+        display_caption = text_pipe(figdesc_prompt)[0]["generated_text"]
+        # 防止它没按要求开头
+        if not display_caption.lower().startswith("a bar chart"):
+            display_caption = "A bar chart showing " + display_caption.lstrip()
+    else:
+        # 没有论文就给一个通用的安全句子
+        display_caption = (
+            "A bar chart showing the response of several experimental groups under "
+            "increasing levels of environmental stress."
+        )
 
-    # Step 3: 学术化解释（重点“读柱状图的趋势”）
-    context_part = f"Paper context: {paper_context}\n\n" if paper_context else ""
+    # 下面我们不再依赖 vision 的“钟楼 caption”，
+    # 而是用刚才生成的 display_caption 作为 figure description
+    raw_caption = display_caption
 
+    # ===== 2. 学术化解释 =====
+    context_part = f"Paper context: {paper_context}\n\n" if paper_context else ""
     prompt_academic = (
         "You are writing the Results section of a scientific paper.\n\n"
         f"{context_part}"
         f"Figure description: {raw_caption}\n\n"
-        "Assume this is a bar chart with several groups and treatments. "
-        "In 3–4 sentences, describe what the figure shows using formal academic English. "
-        "Explain:\n"
-        "- what kind of response or variable is on the vertical axis in generic terms;\n"
-        "- how the bars change when moving from control to more intense treatments;\n"
-        "- which group appears more sensitive or shows a stronger change, if any.\n"
-        "Do not invent exact numbers. Answer directly with the explanation:"
+        "In 3–4 sentences, explain what this bar chart shows using formal academic "
+        "English. Describe in generic terms what is on the vertical axis, how the "
+        "bars change from control to stronger treatments, and which group appears "
+        "more sensitive if any. Do not invent exact numbers.\n\n"
+        "Explanation:"
     )
     academic_desc = text_pipe(prompt_academic)[0]["generated_text"]
 
-    # Step 5a: Caption
+    # ===== 3. Caption =====
     caption_prompt = (
-        "Based on the explanation below, write a 1–2 sentence figure caption for a scientific paper. "
-        "Mention that this is a bar chart and highlight the main comparison or trend.\n\n"
+        "Based on the explanation below, write a 1–2 sentence figure caption for a "
+        "scientific paper. Mention that this is a bar chart and highlight the main "
+        "comparison or trend.\n\n"
         f"{academic_desc}\n\nCaption:"
     )
     caption_text = text_pipe(caption_prompt)[0]["generated_text"]
 
-    # Step 5b: Summary / 讨论
+    # ===== 4. Summary / 讨论 =====
     summary_prompt = (
         "Rewrite the following explanation as a short paragraph (3–4 sentences) "
         "for the Results section, highlighting the main pattern and conclusion. "
@@ -141,7 +157,7 @@ def analyze_and_enhance(image, style, language, paper_file):
     )
     summary_text = text_pipe(summary_prompt)[0]["generated_text"]
 
-    # 语言切换
+    # ===== 5. 语言切换 =====
     if language == "中文":
         academic_desc_out = text_pipe(
             f"Translate the following academic explanation into Chinese:\n{academic_desc}"
@@ -155,11 +171,12 @@ def analyze_and_enhance(image, style, language, paper_file):
     else:
         academic_desc_out = academic_desc
 
-    # Step 4: 图像增强
+    # ===== 6. 图像增强 =====
     img_rgb = image.convert("RGB")
     img_c = ImageEnhance.Contrast(img_rgb).enhance(1.4)
     enhanced_image = ImageEnhance.Brightness(img_c).enhance(1.1)
 
+    # 注意这里输出的是 display_caption，而不是 vision 的原始结果
     return image, raw_caption, academic_desc_out, enhanced_image, caption_text, summary_text