Upload 9 files

app.py

@@ -0,0 +1,41 @@
+import streamlit as st
+from config import Config
+from helpers.image_helper import create_temp_file
+from helpers.llm_helper import analyze_image_file, stream_parser
+
+page_title = Config.PAGE_TITLE
+
+# configures page settings
+st.set_page_config(
+    page_title=page_title,
+    initial_sidebar_state="expanded",
+)
+
+# page title
+st.title(page_title)
+
+st.markdown("Select an image file to analyze.")
+
+# displays file upload widget
+uploaded_file = st.file_uploader("Choose image file", type=['png', 'jpg', 'jpeg'] )
+
+# sets up sidebar nav widgets
+#with st.sidebar:   
+    # creates selectbox to pick the model we would like to use
+image_model = 'llava:7b'
+
+if chat_input := st.chat_input("What would you like to ask?"):
+    if uploaded_file is None:
+        st.error('You must select an image file to analyze!')
+        st.stop()
+
+    # Color formatting example https://docs.streamlit.io/library/api-reference/text/st.markdown
+    with st.status(":red[Processing image file. DON'T LEAVE THIS PAGE WHILE IMAGE FILE IS BEING ANALYZED...]", expanded=True) as status:
+        st.write(":orange[Analyzing Image File...]")
+
+        # creates the audio file
+        stream = analyze_image_file(uploaded_file, model=image_model, user_prompt=chat_input)
+
+        stream_output = st.write_stream(stream_parser(stream))
+
+        st.write(":green[Done analyzing image file]")

helpers/image_helper.py

@@ -0,0 +1,24 @@
+
+import io
+import tempfile
+from PIL import Image
+    
+def create_temp_file(text_file):
+    # create a local tempfile of file that was selected to be uploaded
+    with tempfile.NamedTemporaryFile(delete=False, suffix='.png') as tmp:
+        tmp.write(text_file.getvalue())
+        tmp_path = tmp.name  # Save the path where the tempfile has been written
+
+        return tmp_path
+
+def get_image_bytes(image_file):
+    # Open the image file
+    image_path = image_file  # Replace with the path to your image file
+    image = Image.open(image_path)
+
+    # Convert the image to bytes
+    with io.BytesIO() as output:
+        image.save(output, format="png")  # Change the format as needed (e.g., JPEG, PNG)
+        image_bytes = output.getvalue()
+    
+    return image_bytes

helpers/llm_helper.py

@@ -0,0 +1,22 @@
+from ollama import generate
+from config import Config
+from helpers.image_helper import get_image_bytes
+
+system_prompt = Config.SYSTEM_PROMPT
+
+def analyze_image_file(image_file, model, user_prompt):
+    # gets image bytes using helper function
+    image_bytes = get_image_bytes(image_file)
+
+    # calls the llava model using Ollama SDK
+    stream = generate(model=model, 
+            prompt=user_prompt, 
+            images=[image_bytes], 
+            stream=True)
+
+    return stream
+
+# handles stream response back from LLM
+def stream_parser(stream):
+    for chunk in stream:
+        yield chunk['response']

requirements.txt

@@ -0,0 +1,3 @@
+streamlit
+python-dotenv
+ollama