Update app.py

app.py

@@ -16,7 +16,6 @@ import edge_tts
 from audio_recorder_streamlit import audio_recorder
 import nest_asyncio
 import re
-from streamlit_paste_button import paste_image_button
 import pytz
 import shutil
 import anthropic
@@ -128,7 +127,7 @@ def init_session_state():
         'autosend': True, 'autosearch': True, 'last_message': "", 'last_query': "",
         'mp3_files': {}, 'timer_start': time.time(), 'quote_index': 0,
         'quote_source': "famous", 'last_sent_transcript': "", 'old_val': None,
-        'last_refresh': time.time(), 'paper_metadata': {}, 'paste_trigger': False
+        'last_refresh': time.time(), 'paper_metadata': {}, 'paste_image_base64': None
     }
     for k, v in defaults.items():
         if k not in st.session_state:
@@ -564,6 +563,53 @@ def create_zip_of_files(md_files, mp3_files, png_files, mp4_files, query):
         [z.write(f) for f in all_files]
     return zip_name
 
+# Custom Paste Image Component
+def paste_image_component():
+    # JavaScript to capture clipboard image and send it to Streamlit
+    paste_script = """
+    <script>
+    function pasteImage() {
+        navigator.clipboard.read().then(clipboardItems => {
+            for (const item of clipboardItems) {
+                if (item.types.includes('image/png') || item.types.includes('image/jpeg')) {
+                    item.getType('image/png').then(blob => {
+                        const reader = new FileReader();
+                        reader.onload = function(event) {
+                            document.getElementById('paste_input').value = event.target.result.split(',')[1];
+                            document.getElementById('paste_form').submit();
+                        };
+                        reader.readAsDataURL(blob);
+                    });
+                    break;
+                }
+            }
+        }).catch(err => {
+            console.error('Failed to read clipboard: ', err);
+            document.getElementById('paste_input').value = 'ERROR: ' + err.message;
+            document.getElementById('paste_form').submit();
+        });
+    }
+    </script>
+    <form id="paste_form" action="/" method="POST" style="display: none;">
+        <input type="text" id="paste_input" name="paste_data" />
+    </form>
+    <button onclick="pasteImage()">Paste Image 📋</button>
+    """
+    paste_data = components.html(paste_script, height=50)
+    if paste_data and 'paste_data' in paste_data:
+        base64_str = paste_data['paste_data']
+        if base64_str.startswith('ERROR:'):
+            st.warning(f"Paste failed: {base64_str}")
+            return None
+        try:
+            img_bytes = base64.b64decode(base64_str)
+            img = Image.open(io.BytesIO(img_bytes))
+            return img
+        except Exception as e:
+            st.warning(f"Error decoding pasted image: {e}")
+            return None
+    return None
+
 # 🎮 Main Interface
 def main():
     init_session_state()
@@ -610,40 +656,35 @@ def main():
             st.code(numbered_content, language="python")
 
         message = st.text_input(f"Message as {st.session_state.username}", key="message_input")
-        paste_result = paste_image_button("Paste Image 📋", key=f"paste_button_msg_{int(time.time())}")  # Dynamic key to force re-render
-        if paste_result.image_data is not None and not st.session_state['paste_trigger']:
-            st.session_state['paste_trigger'] = True
+        
+        # Custom Paste Image Button
+        pasted_image = paste_image_component()
+        if pasted_image is not None and st.session_state['paste_image_base64'] != pasted_image.tobytes():
+            st.session_state['paste_image_base64'] = pasted_image.tobytes()
             voice = FUN_USERNAMES.get(st.session_state.username, "en-US-AriaNeural")
-            if isinstance(paste_result.image_data, str):
-                st.session_state.message_text = paste_result.image_data
-                message = st.text_input(f"Message as {st.session_state.username}", key="message_input_paste", value=st.session_state.message_text)
-            else:
-                st.image(paste_result.image_data, caption="Pasted Image")
-                filename = asyncio.run(save_pasted_image(paste_result.image_data, st.session_state.username))
-                if filename:
-                    st.session_state.pasted_image_data = filename
-                    # Allow user to add text to accompany the image
-                    image_prompt = st.text_input("Add a prompt for Claude (e.g., 'OCR this image')", key="image_prompt")
-                    if image_prompt:
-                        # Run Claude search with both text and image
-                        md_file_claude, audio_file_claude, claude_result = asyncio.run(
-                            perform_claude_search(image_prompt, st.session_state.username, paste_result.image_data)
-                        )
-                        if audio_file_claude:
-                            play_and_download_audio(audio_file_claude)
-                        # Feed Claude result to ArXiv search
-                        md_file_arxiv, audio_file_arxiv = asyncio.run(
-                            perform_arxiv_search(image_prompt, st.session_state.username, claude_result)
-                        )
-                        if audio_file_arxiv:
-                            play_and_download_audio(audio_file_arxiv)
-                        st.session_state.pasted_image_data = None
-                        st.session_state['paste_trigger'] = False
-                        st.session_state.timer_start = time.time()
-                        save_username(st.session_state.username)
-                        st.rerun()
+            st.image(pasted_image, caption="Pasted Image")
+            filename = asyncio.run(save_pasted_image(pasted_image, st.session_state.username))
+            if filename:
+                st.session_state.pasted_image_data = filename
+                image_prompt = st.text_input("Add a prompt for Claude (e.g., 'OCR this image')", key="image_prompt")
+                if image_prompt:
+                    md_file_claude, audio_file_claude, claude_result = asyncio.run(
+                        perform_claude_search(image_prompt, st.session_state.username, pasted_image)
+                    )
+                    if audio_file_claude:
+                        play_and_download_audio(audio_file_claude)
+                    md_file_arxiv, audio_file_arxiv = asyncio.run(
+                        perform_arxiv_search(image_prompt, st.session_state.username, claude_result)
+                    )
+                    if audio_file_arxiv:
+                        play_and_download_audio(audio_file_arxiv)
+                    st.session_state.pasted_image_data = None
+                    st.session_state['paste_image_base64'] = None
+                    st.session_state.timer_start = time.time()
+                    save_username(st.session_state.username)
+                    st.rerun()
 
-        if (message and message != st.session_state.last_message) or (st.session_state.pasted_image_data and not st.session_state['paste_trigger']):
+        if (message and message != st.session_state.last_message) or (st.session_state.pasted_image_data and not st.session_state['paste_image_base64']):
             st.session_state.last_message = message
             col_send, col_claude, col_arxiv = st.columns([1, 1, 1])