create app.py

app.py

@@ -0,0 +1,105 @@
+import gradio as gr
+import requests
+import os
+
+# Read backend URL from Hugging Face secret (set in Settings → Repository secrets)
+BACKEND_URL = os.environ.get("KERNL_BACKEND_URL", "").rstrip('/')
+
+if not BACKEND_URL:
+    BACKEND_URL = None
+
+def query_kernl(scenario, with_brain):
+    """Call the Kernl /agent/handle endpoint and format the response."""
+    if not BACKEND_URL:
+        return "❌ Backend URL not configured. Please set the KERNL_BACKEND_URL secret in Space Settings → Repository secrets."
+    
+    if not scenario or not scenario.strip():
+        return "❗ Please enter a scenario."
+    
+    try:
+        response = requests.post(
+            f"{BACKEND_URL}/agent/handle",
+            json={
+                "company_id": "rivanly-inc",
+                "scenario": scenario,
+                "with_brain": with_brain
+            },
+            timeout=30
+        )
+        response.raise_for_status()
+        data = response.json()
+        
+        # Format the response nicely
+        output = []
+        output.append(f"**Action:** `{data.get('action', 'N/A')}`")
+        output.append(f"**Rule Applied:** `{data.get('rule_applied', 'N/A')}`")
+        output.append(f"**Message:** {data.get('message_to_customer', data.get('answer', 'N/A'))}")
+        if data.get('evidence'):
+            output.append(f"**Evidence:** {data.get('evidence')}")
+        output.append(f"**Skill Matched:** `{data.get('skill_matched', 'N/A')}`")
+        output.append(f"**Confidence:** `{data.get('confidence', 'N/A')}`")
+        
+        return "\n\n".join(output)
+    
+    except requests.exceptions.ConnectionError:
+        return "❌ Cannot connect to Kernl backend. Make sure your backend is running and KERNL_BACKEND_URL is correct."
+    except requests.exceptions.Timeout:
+        return "❌ Request timed out. The backend is taking too long to respond."
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
+
+# Custom dark theme with deep teal accent
+theme = gr.themes.Soft(
+    primary_hue="teal",
+    secondary_hue="teal",
+    neutral_hue="gray",
+    font=gr.themes.GoogleFont("Inter")
+)
+
+with gr.Blocks(theme=theme, title="Kernl – Operational Memory for AI Agents") as demo:
+    gr.Markdown("""
+    # 🧠 Kernl
+    ### Operational memory for AI agents
+    
+    Kernl compiles how your company actually decides things – from Slack, SOPs, and tickets – into an executable skills file.
+    Any agent. Any task. Correct every time.
+    """)
+    
+    with gr.Row():
+        with gr.Column(scale=1):
+            scenario_input = gr.Textbox(
+                label="Enter your business scenario",
+                placeholder="Example: Enterprise customer, 18 months tenure, wants $1,200 refund",
+                lines=4
+            )
+            with_brain_toggle = gr.Checkbox(
+                label="🧠 Use Company Brain (Kernl)",
+                value=True,
+                info="ON = Kernl uses compiled company knowledge. OFF = generic AI answer."
+            )
+            submit_btn = gr.Button("Ask Kernl", variant="primary", size="lg")
+        
+        with gr.Column(scale=2):
+            output_box = gr.Markdown(label="Kernl's Response", value="*Your answer will appear here...*")
+    
+    gr.Markdown("""
+    ---
+    ### Try these example scenarios (copy & paste):
+    
+    - `Enterprise customer, 18 months tenure, wants $1,200 refund`
+    - `Annual plan customer, day 10 of subscription, $300 refund requested`
+    - `Customer reporting P0 bug on dashboard, enterprise plan`
+    - `Customer showing 3 churn signals in last 30 days`
+    - `Startup requesting 40% discount`
+    
+    ---
+    **Built with AMD MI300X, vLLM, and LangGraph** | [GitHub](https://github.com/your-repo) | **Track 1: AI Agents & Agentic Workflows**
+    """)
+    
+    submit_btn.click(
+        fn=query_kernl,
+        inputs=[scenario_input, with_brain_toggle],
+        outputs=output_box
+    )
+
+demo.launch()