Update app.py

app.py

@@ -50,7 +50,6 @@ def parse_numbered_steps(text):
 # 2. SESSION STATE SETUP
 # ─────────────────────────────────────────────────────────────────────────────
 
-# Central dictionary for session state management
 if "app_state" not in st.session_state:
     st.session_state.app_state = {
         "steps": [], "images": {}, "tools_list": [], "current_step": 1,
@@ -76,30 +75,23 @@ def reset_state():
     st.success("✅ Reset complete!")
     st.rerun()
 
-def send_gemini_request(model_name, prompt, image):
-    """Helper to send requests using the client.chats interface."""
+def send_text_request(model_name, prompt, image):
+    """Helper to send requests that expect only a text response."""
     try:
-        chat = client.chats.create(
-            model=model_name,
-            config=types.GenerateContentConfig(response_modalities=["Text"]) # Assuming text response for these tasks
-        )
+        chat = client.chats.create(model=model_name)
         response = chat.send_message([prompt, image])
-        # Combine all text parts from the response
         response_text = "".join(part.text for part in response.candidates[0].content.parts if part.text)
         return response_text.strip()
     except Exception as e:
-        st.error(f"Error communicating with model {model_name}: {str(e)}")
+        st.error(f"Error with model {model_name}: {str(e)}")
         return None
 
 def initial_analysis(uploaded_file, context_text):
-    """
-    First pass with AI: get category, then get title, description, and initial plan.
-    """
+    """First pass with AI: get category, then title, description, and initial plan."""
     image = Image.open(uploaded_file)
     st.session_state.app_state['user_image'] = image
 
     with st.spinner("🤖 Analyzing your project and preparing a plan..."):
-        # Step 1: Detect Category using CATEGORY_MODEL
         category_prompt = (
             "You are an expert DIY assistant. Analyze the user's image and context. "
             f"Context: '{context_text}'. "
@@ -108,32 +100,28 @@ def initial_analysis(uploaded_file, context_text):
             "Upcycling & Sustainable Crafts, or DIY Project Creation. "
             "Reply with ONLY the category name."
         )
-        category = send_gemini_request(CATEGORY_MODEL, category_prompt, image)
+        category = send_text_request(CATEGORY_MODEL, category_prompt, image)
         if not category: return
         st.session_state.app_state['category'] = category
 
-        # Step 2: Generate Title, Description, and Plan using GENERATION_MODEL
         plan_prompt = f"""
-        You are an expert DIY assistant. The user's project is in the category: {category}.
+        You are an expert DIY assistant in the category: {category}.
         User Context: "{context_text if context_text else 'No context provided.'}"
-
         Based on the image and context, perform the following:
         1.  **Title:** Create a short, clear title for this project.
         2.  **Description:** Write a brief, one-paragraph description of the goal.
         3.  **Initial Plan:**
-            - If the category is 'Upcycling & Sustainable Crafts' AND the user has NOT specified a project, propose three distinct project options as a numbered list under the heading "UPCYCLING OPTIONS:".
+            - If 'Upcycling & Sustainable Crafts' AND no specific project is mentioned, propose three distinct project options as a numbered list under "UPCYCLING OPTIONS:".
             - For all other cases, briefly outline the main stages of the proposed solution.
-
         Structure your response EXACTLY like this:
         TITLE: [Your title]
         DESCRIPTION: [Your description]
         INITIAL PLAN:
         [Your plan or 3 options]
         """
-        plan_response = send_gemini_request(GENERATION_MODEL, plan_prompt, image)
+        plan_response = send_text_request(GENERATION_MODEL, plan_prompt, image)
         if not plan_response: return
 
-    # Parse the second response
     try:
         st.session_state.app_state['project_title'] = re.search(r"TITLE:\s*(.*)", plan_response).group(1).strip()
         st.session_state.app_state['project_description'] = re.search(r"DESCRIPTION:\s*(.*)", plan_response, re.DOTALL).group(1).strip()
@@ -148,16 +136,15 @@ def initial_analysis(uploaded_file, context_text):
         st.session_state.app_state['prompt_sent'] = True
         if context_text:
             st.session_state.app_state['plan_approved'] = True
-            generate_detailed_steps()
+            generate_detailed_guide_with_images()
         else:
             st.session_state.app_state['plan_approved'] = False
     except AttributeError:
         st.error("The AI response was not in the expected format. Please try again.")
         st.session_state.app_state['prompt_sent'] = False
 
-
-def generate_detailed_steps(selected_option=None):
-    """Generates the detailed, step-by-step guide using GENERATION_MODEL."""
+def generate_detailed_guide_with_images(selected_option=None):
+    """Generates the detailed guide with steps and illustrations."""
     image = st.session_state.app_state.get('user_image')
     if not image:
         st.error("Image not found. Please start over."); return
@@ -169,39 +156,55 @@ def generate_detailed_steps(selected_option=None):
     detailed_prompt = f"""
     You are a DIY expert. The user wants to proceed with the project titled "{st.session_state.app_state['project_title']}".
     {context}
-    Provide a detailed guide in this EXACT format:
+    Provide a detailed guide. For each step, you MUST provide a simple, clear illustrative image.
+    Format your response EXACTLY like this:
 
     TOOLS AND MATERIALS:
     - Tool A
     - Material B
 
     STEPS:
-    1. First step instructions. Be specific and clear.
+    1. First step instructions.
     2. Second step instructions...
     """
-    with st.spinner("🛠️ Generating your detailed step-by-step guide..."):
-        full_resp = send_gemini_request(GENERATION_MODEL, detailed_prompt, image)
-        if not full_resp: return
-
-    try:
-        tools_section = re.search(r"TOOLS AND MATERIALS:\s*(.*?)\s*STEPS:", full_resp, re.DOTALL).group(1).strip()
-        steps_section = re.search(r"STEPS:\s*(.*)", full_resp, re.DOTALL).group(1).strip()
-
-        st.session_state.app_state['tools_list'] = [line.strip("- ").strip() for line in tools_section.split('\n') if line.strip()]
-        st.session_state.app_state['steps'] = parse_numbered_steps(steps_section)
-
-        for idx, step_text in st.session_state.app_state['steps']:
-            st.session_state.app_state['done_flags'][idx] = False
-            st.session_state.app_state['notes'][idx] = ""
-            timer_match = re.search(r"wait\s+for\s+(\d+)\s+(seconds?|minutes?)", step_text.lower())
-            if timer_match:
-                val, unit = int(timer_match.group(1)), timer_match.group(2)
-                st.session_state.app_state['timers'][idx] = val * (60 if "minute" in unit else 1)
-            else:
-                st.session_state.app_state['timers'][idx] = 0
-    except AttributeError:
-        st.error("Failed to parse the detailed guide from the AI. The format was incorrect.")
-
+    with st.spinner("🛠️ Generating your detailed guide with illustrations..."):
+        try:
+            chat = client.chats.create(
+                model=GENERATION_MODEL,
+                config=types.GenerateContentConfig(response_modalities=["Text", "Image"])
+            )
+            full_resp = chat.send_message([detailed_prompt, image])
+            gen_parts = full_resp.candidates[0].content.parts
+
+            combined_text = ""
+            inline_images = []
+            for part in gen_parts:
+                if part.text is not None:
+                    combined_text += part.text + "\n"
+                if part.inline_data is not None:
+                    img = Image.open(BytesIO(part.inline_data.data))
+                    inline_images.append(img)
+            combined_text = combined_text.strip()
+
+            tools_section = re.search(r"TOOLS AND MATERIALS:\s*(.*?)\s*STEPS:", combined_text, re.DOTALL).group(1).strip()
+            steps_section = re.search(r"STEPS:\s*(.*)", combined_text, re.DOTALL).group(1).strip()
+            parsed_steps = parse_numbered_steps(steps_section)
+
+            st.session_state.app_state['tools_list'] = [line.strip("- ").strip() for line in tools_section.split('\n') if line.strip()]
+            st.session_state.app_state['steps'] = parsed_steps
+            st.session_state.app_state['images'] = {idx: inline_images[idx - 1] for idx, _ in parsed_steps if idx - 1 < len(inline_images)}
+
+            for idx, step_text in parsed_steps:
+                st.session_state.app_state['done_flags'][idx] = False
+                st.session_state.app_state['notes'][idx] = ""
+                timer_match = re.search(r"wait\s+for\s+(\d+)\s+(seconds?|minutes?)", step_text.lower())
+                if timer_match:
+                    val, unit = int(timer_match.group(1)), timer_match.group(2)
+                    st.session_state.app_state['timers'][idx] = val * (60 if "minute" in unit else 1)
+                else:
+                    st.session_state.app_state['timers'][idx] = 0
+        except Exception as e:
+            st.error(f"Failed to generate or parse the illustrated guide: {str(e)}")
 
 def render_sidebar_navigation():
     st.sidebar.markdown("## Steps Navigation")
@@ -228,7 +231,14 @@ def render_step(idx, text):
     total = len(st.session_state.app_state['steps'])
     st.markdown(f"### Step {idx} of {total}")
     st.write(text)
-    # Timer, checkbox, notes, and navigation logic remains the same...
+
+    if idx in st.session_state.app_state['images']:
+        st.image(
+            st.session_state.app_state['images'][idx],
+            caption=f"Illustration for step {idx}",
+            use_container_width=True
+        )
+
     done = st.checkbox("✅ Mark this step as completed", value=st.session_state.app_state['done_flags'].get(idx, False), key=f"done_{idx}")
     st.session_state.app_state['done_flags'][idx] = done
     notes = st.text_area("📝 Your notes for this step:", value=st.session_state.app_state['notes'].get(idx, ""), height=100, key=f"notes_{idx}")
@@ -253,12 +263,11 @@ with st.expander("ℹ️ How it works", expanded=False):
     st.write("""
     1.  **Upload a photo** of your project.
     2.  **(Optional) Describe your goal** for more accurate results.
-    3.  **Review the Plan.** The AI will propose a plan. If you didn't provide a description, you'll be asked to approve it. For crafts, you may get options to choose from.
-    4.  **Get Your Guide** with tools and step-by-step instructions.
+    3.  **Review the Plan.** The AI will propose a plan. If you didn't provide a description, you'll be asked to approve it.
+    4.  **Get Your Guide** with tools and illustrated step-by-step instructions.
     5.  **Follow the Steps** using the interactive checklist.
     """)
 
-# --- Main UI ---
 if not st.session_state.app_state['prompt_sent']:
     st.markdown("---")
     col1, col2 = st.columns([3, 1])
@@ -275,8 +284,6 @@ if not st.session_state.app_state['prompt_sent']:
                 st.warning("⚠️ Please upload an image first!")
     if st.button("🔄 Start Over", use_container_width=True):
         reset_state()
-
-# --- Results and Steps UI ---
 else:
     render_sidebar_navigation()
     st.markdown("---")
@@ -290,14 +297,14 @@ else:
             st.markdown("#### The AI has suggested a few projects. Please choose one:")
             for i, option in enumerate(st.session_state.app_state['upcycling_options']):
                 if st.button(option, key=f"option_{i}"):
-                    generate_detailed_steps(selected_option=option)
+                    generate_detailed_guide_with_images(selected_option=option)
                     st.rerun()
         elif not st.session_state.app_state['plan_approved']:
             st.markdown("#### The AI has proposed the following plan:")
             st.success(st.session_state.app_state['initial_plan'])
             if st.button("✅ Looks good, proceed with this plan", type="primary"):
                 st.session_state.app_state['plan_approved'] = True
-                generate_detailed_steps()
+                generate_detailed_guide_with_images()
                 st.rerun()
     else:
         render_tools_list()