(ui): add multi-modality routing skeleton to dashboard.

- Restructured dashboard to support multi-modality routing.
- Built a robust interface foundation that enables long-term support for Raman, Image, FTIR pathways.
- Per-modality page routing: Raman(Dashboard, Model MGMT, Inferece); Image & FTIR(placeholder shells).
- All modlaity and page selections preserved via 'st.session_state'.

dashboard/app.py

@@ -0,0 +1,138 @@
+"""
+Streamlit UI for ml-polymer-recycling — Step 3a: Multi-Modality Interface Skeleton
+- Introduces top-level modality selector (Raman | Image | FTIR)
+- Modular routing for each modality with scoped navigation
+- Only Raman pages are active (Model Management, Inference)
+- Image/FTIR pages show 'Coming Soon' placeholders
+- All routing and session state preserved cleanly
+"""
+
+import streamlit as st
+from pathlib import Path
+
+# --- PAGE CONFIGURATION ---
+st.set_page_config(
+    page_title="ML Polymer Recycling",
+    page_icon="🧪",
+    layout="wide"
+)
+
+# --- SESSION STATE INITIALIZATION ---
+def init_session_state():
+    if "status_message" not in st.session_state:
+        st.session_state.status_message = "Ready."
+    if "status_type" not in st.session_state:
+        st.session_state.status_type = "ok"
+    if "modality" not in st.session_state:
+        st.session_state.modality = "Raman"
+
+# --- STATUS BANNER ---
+def display_status():
+    style_map = {
+        "ok": ("#e8f5e9", "#2e7d32"),
+        "warn": ("#fff8e1", "#f9a825"),
+        "err": ("#ffebee", "#c62828")
+    }
+    bg_color, text_color = style_map.get(st.session_state.status_type, ("#f0f0f0", "#333"))
+    st.markdown(f"""
+        <div style='background-color:{bg_color}; padding:0.75em 1em; border-radius:8px; color:{text_color};'>
+            <b>Status:</b> {st.session_state.status_message}
+        </div>
+    """, unsafe_allow_html=True)
+
+# --- SIDEBAR NAVIGATION ---
+def display_sidebar():
+    with st.sidebar:
+        st.header("🧪 ML Polymer Dashboard")
+
+        modality = st.radio("Modality", ["Raman", "Image", "FTIR"])
+        st.session_state.modality = modality
+
+        if modality == "Raman":
+            page = st.radio("Raman Pages", ["Dashboard", "Model Management", "Inference"])
+        elif modality == "Image":
+            page = st.radio("Image Pages", ["Model Management", "Inference"])
+        elif modality == "FTIR":
+            page = st.radio("FTIR Pages", ["Model Management", "Inference"])
+
+        return modality, page
+
+# --- HEADER BAR ---
+def display_header(title: str):
+    st.title(title)
+    display_status()
+    st.markdown("---")
+
+# --- MODEL DISCOVERY (Raman only for now) ---
+def discover_models(outputs_dir="outputs"):
+    out = []
+    root = Path(outputs_dir)
+    if not root.exists():
+        return []
+    for p in sorted(root.rglob("*.pth")):
+        out.append(p)
+    return out
+
+# --- RAMAN PAGES ---
+def raman_dashboard():
+    display_header("Raman Dashboard")
+    st.write("This will house future metrics, model count, and version history.")
+
+def raman_model_management():
+    display_header("Raman Model Management")
+    models = discover_models()
+    if not models:
+        st.info("No model weights found in outputs/. Place .pth files there to make them discoverable.")
+    else:
+        st.markdown(f"**Discovered {len(models)} model weight file(s):**")
+        for m in models:
+            st.code(str(m), language="text")
+
+def raman_inference():
+    display_header("Raman Inference")
+    st.write("This will house the real-time spectrum upload and model prediction interface.")
+
+# --- IMAGE + FTIR PLACEHOLDERS ---
+def image_model_management():
+    display_header("Image Model Management")
+    st.info("Image-based model integration is coming soon.")
+
+def image_inference():
+    display_header("Image Inference")
+    st.info("This page will allow batch image upload and multi-model prediction.")
+
+def ftir_model_management():
+    display_header("FTIR Model Management")
+    st.info("FTIR model support is planned and will be developed after clarification with Dr. K.")
+
+def ftir_inference():
+    display_header("FTIR Inference")
+    st.info("FTIR input and prediction support will be added in a future phase.")
+
+# --- MAIN ENTRY POINT ---
+def main():
+    init_session_state()
+    modality, page = display_sidebar()
+
+    if modality == "Raman":
+        if page == "Dashboard":
+            raman_dashboard()
+        elif page == "Model Management":
+            raman_model_management()
+        elif page == "Inference":
+            raman_inference()
+
+    elif modality == "Image":
+        if page == "Model Management":
+            image_model_management()
+        elif page == "Inference":
+            image_inference()
+
+    elif modality == "FTIR":
+        if page == "Model Management":
+            ftir_model_management()
+        elif page == "Inference":
+            ftir_inference()
+
+if __name__ == "__main__":
+    main()