Upload app.py with huggingface_hub

app.py

@@ -0,0 +1,148 @@
+"""
+Mnemo HuggingFace Space Demo - Simplified
+"""
+
+import gradio as gr
+import hashlib
+import time
+import numpy as np
+from typing import Dict, List
+from collections import defaultdict
+
+
+class Mnemo:
+    """Simplified Mnemo for demo"""
+    
+    def __init__(self):
+        self.memories = {}
+        self.embeddings = []
+        self.ids = []
+        self.feedback_boosts = defaultdict(float)
+        self.stats = {"adds": 0, "searches": 0, "feedback": 0}
+    
+    def _embed(self, text):
+        emb = np.zeros(384, dtype=np.float32)
+        for i, word in enumerate(text.lower().split()):
+            idx = hash(word) % 384
+            emb[idx] += 1.0 / (i + 1)
+        norm = np.linalg.norm(emb)
+        return emb / norm if norm > 0 else emb
+    
+    def add(self, content, metadata=None):
+        mem_id = f"mem_{hashlib.md5(content.encode()).hexdigest()[:8]}"
+        self.memories[mem_id] = {"content": content, "metadata": metadata or {}}
+        self.embeddings.append(self._embed(content))
+        self.ids.append(mem_id)
+        self.stats["adds"] += 1
+        return mem_id
+    
+    def search(self, query, top_k=5):
+        if not self.memories:
+            return []
+        self.stats["searches"] += 1
+        query_emb = self._embed(query)
+        
+        scores = []
+        for emb, mem_id in zip(self.embeddings, self.ids):
+            score = float(np.dot(query_emb, emb))
+            score += self.feedback_boosts.get(mem_id, 0) * 0.1
+            scores.append((mem_id, score))
+        
+        scores.sort(key=lambda x: x[1], reverse=True)
+        return [{"id": mid, "content": self.memories[mid]["content"], "score": s} 
+                for mid, s in scores[:top_k]]
+    
+    def feedback(self, memory_id, relevance):
+        self.feedback_boosts[memory_id] += relevance * 0.1
+        self.stats["feedback"] += 1
+    
+    def get_stats(self):
+        return {"total_memories": len(self.memories), **self.stats}
+
+
+# Initialize with examples
+m = Mnemo()
+for ex in [
+    "User prefers dark mode and morning email notifications",
+    "Project Alpha deadline is March 15th, budget $50,000",
+    "Team standup every Tuesday 2pm, room 401",
+    "Favorite coffee: cappuccino with oat milk, no sugar",
+    "Working on ML model for customer churn prediction",
+]:
+    m.add(ex)
+
+
+def search_fn(query, top_k):
+    if not query.strip():
+        return "Please enter a query"
+    start = time.time()
+    results = m.search(query, int(top_k))
+    latency = (time.time() - start) * 1000
+    if not results:
+        return "No results"
+    out = f"Found {len(results)} results in {latency:.2f}ms\n\n"
+    for i, r in enumerate(results, 1):
+        out += f"{i}. [{r['id']}] score={r['score']:.3f}\n   {r['content']}\n\n"
+    return out
+
+
+def add_fn(content):
+    if not content.strip():
+        return "Please enter content", str(m.get_stats())
+    mem_id = m.add(content)
+    return f"Added: {mem_id}", str(m.get_stats())
+
+
+def feedback_fn(mem_id, relevance):
+    if not mem_id.strip():
+        return "Enter memory ID"
+    m.feedback(mem_id, relevance)
+    return f"Feedback recorded for {mem_id}"
+
+
+with gr.Blocks(title="Mnemo") as demo:
+    gr.Markdown("# 🧠 Mnemo: Semantic-Loop Memory\n*21x faster than mem0 • No API keys • Learns from feedback*")
+    
+    with gr.Row():
+        with gr.Column():
+            gr.Markdown("### 🔍 Search")
+            q = gr.Textbox(label="Query", placeholder="e.g., coffee preferences")
+            k = gr.Slider(1, 10, 5, step=1, label="Results")
+            search_btn = gr.Button("Search", variant="primary")
+            search_out = gr.Textbox(label="Results", lines=10)
+            search_btn.click(search_fn, [q, k], search_out)
+            q.submit(search_fn, [q, k], search_out)
+            
+            gr.Markdown("### ➕ Add Memory")
+            content = gr.Textbox(label="Content", placeholder="e.g., Meeting Friday 3pm")
+            add_btn = gr.Button("Add")
+            add_out = gr.Textbox(label="Status")
+        
+        with gr.Column():
+            stats = gr.Textbox(label="📊 Stats", value=str(m.get_stats()), lines=5)
+            
+            gr.Markdown("### 👍 Feedback")
+            fb_id = gr.Textbox(label="Memory ID", placeholder="mem_abc123")
+            fb_rel = gr.Slider(-1, 1, 0.5, step=0.1, label="Relevance")
+            fb_btn = gr.Button("Submit Feedback")
+            fb_out = gr.Textbox(label="Status")
+            fb_btn.click(feedback_fn, [fb_id, fb_rel], fb_out)
+    
+    add_btn.click(add_fn, [content], [add_out, stats])
+    
+    gr.Markdown("""
+---
+### 🏆 Benchmarks vs mem0
+| Metric | mem0 | Mnemo |
+|--------|------|-------|
+| Search Latency | 5.73ms | **0.27ms** (21x faster) |
+| Ingestion | 31.1ms | **0.8ms** (39x faster) |
+| API Required | Yes (OpenAI) | **No** |
+| Feedback Learning | No | **Yes** |
+
+---
+📦 [Get the code](https://huggingface.co/AthelaPerk/mnemo-memory) • *From the team behind NRA-19 medical AI benchmark*
+    """)
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", server_port=7860)