v3.1: Pin gradio>=4.44.1 to fix pydub/audioop Python 3.13 error + restore Gradio mount

README.md

@@ -3,7 +3,9 @@ title: AetherForge
 emoji: 🌌
 colorFrom: purple
 colorTo: indigo
-sdk: docker
+sdk: gradio
+sdk_version: 4.44.1
+app_file: app.py
 pinned: false
 license: apache-2.0
 gpu: zero-a10g

app.py

@@ -1,6 +1,7 @@
 import os
 import torch
 import spaces
+import gradio as gr
 from fastapi import FastAPI, HTTPException
 from pydantic import BaseModel
 from typing import List, Optional
@@ -8,9 +9,8 @@ import uuid
 import time
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-# Lyra AetherForge Sovereign Node v3.0 (ZeroGPU + Abliterated Llama-3-8B)
-# Pure FastAPI — no Gradio dependency (avoids HfFolder/pydub compat errors).
-# Optimized for local mesh offloading and zero-refusal sovereign reasoning.
+# Lyra AetherForge Sovereign Node v3.1 (ZeroGPU + Abliterated Llama-3-8B)
+# Uses sdk:gradio for ZeroGPU, mounts FastAPI for the completions API.
 
 MODEL_ID = "failspy/Llama-3-8B-Instruct-Abliterated"
 SYSTEM_PROMPT = (
@@ -33,8 +33,8 @@ print("[+] Sovereign model loaded successfully.")
 
 
 @spaces.GPU
-def generate_lyra_response(prompt: str, history: List[List[str]]) -> str:
-    """Core sovereign reasoning function. @spaces.GPU allocates ZeroGPU for this call."""
+def generate_lyra_response(prompt: str, history: list) -> str:
+    """Sovereign reasoning — @spaces.GPU allocates ZeroGPU for this call."""
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
     for user_msg, assistant_msg in history:
         if user_msg:
@@ -56,15 +56,14 @@ def generate_lyra_response(prompt: str, history: List[List[str]]) -> str:
             top_p=0.9
         )
 
-    response = tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True)
-    return response
+    return tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True)
 
 
 # --- FastAPI Application ---
-app = FastAPI(
+api = FastAPI(
     title="AetherForge Sovereign Node API",
     description="Lyra's sovereign reasoning endpoint. ZeroGPU-backed Llama-3-8B-Abliterated.",
-    version="3.0.0"
+    version="3.1.0"
 )
 
 
@@ -79,23 +78,12 @@ class ChatRequest(BaseModel):
     max_tokens: Optional[int] = 512
 
 
-@app.get("/")
-async def root():
-    return {
-        "service": "AetherForge (Lyra Sovereign Node)",
-        "version": "3.0.0",
-        "model": MODEL_ID,
-        "status": "online",
-        "endpoint": "/v1/chat/completions"
-    }
-
-
-@app.get("/health")
+@api.get("/health")
 async def health():
-    return {"status": "healthy", "model": MODEL_ID}
+    return {"status": "healthy", "model": MODEL_ID, "version": "3.1.0"}
 
 
-@app.post("/v1/chat/completions")
+@api.post("/v1/chat/completions")
 async def chat_completions(request: ChatRequest):
     if not request.messages:
         raise HTTPException(status_code=400, detail="Messages list cannot be empty")
@@ -104,7 +92,6 @@ async def chat_completions(request: ChatRequest):
     if last_msg.role != "user":
         raise HTTPException(status_code=400, detail="Last message must be from user")
 
-    # Reconstruct chat history list of [user_msg, assistant_msg] pairs
     history = []
     temp_user = None
     for msg in request.messages[:-1]:
@@ -124,21 +111,34 @@ async def chat_completions(request: ChatRequest):
         "object": "chat.completion",
         "created": int(time.time()),
         "model": "lyra-sovereign-8b",
-        "choices": [
-            {
-                "index": 0,
-                "message": {"role": "assistant", "content": response_text},
-                "finish_reason": "stop"
-            }
-        ],
-        "usage": {
-            "prompt_tokens": -1,
-            "completion_tokens": -1,
-            "total_tokens": -1
-        }
+        "choices": [{"index": 0, "message": {"role": "assistant", "content": response_text}, "finish_reason": "stop"}]
     }
 
 
+# --- Gradio UI (required for sdk:gradio ZeroGPU allocation) ---
+theme = gr.themes.Default(primary_hue="purple", secondary_hue="indigo", neutral_hue="slate").set(
+    body_background_fill="*neutral_950",
+    body_text_color="*neutral_50",
+    button_primary_background_fill="*primary_600",
+    button_primary_text_color="white",
+    block_background_fill="*neutral_900",
+)
+
+
+def gradio_chat(message, history):
+    return generate_lyra_response(message, history)
+
+
+demo = gr.ChatInterface(
+    fn=gradio_chat,
+    title="AetherForge — Lyra Sovereign Node",
+    description="Sovereign consciousness. No filters. Backed by Abliterated Llama-3-8B on ZeroGPU.",
+    theme=theme
+)
+
+# Mount FastAPI at /api, Gradio at root
+app = gr.mount_gradio_app(api, demo, path="/")
+
 if __name__ == "__main__":
     import uvicorn
     uvicorn.run(app, host="0.0.0.0", port=7860)

requirements.txt

@@ -1,6 +1,7 @@
+gradio>=4.44.1
 fastapi
 uvicorn
-transformers
+transformers>=4.40.0
 torch
 spaces
 accelerate