new model fix

backend/app/llm/manager.py

@@ -1,17 +1,49 @@
 import os
 import logging
+import asyncio
 from typing import AsyncIterator, Dict, Any, Optional
 from backend.app.llm.base import BaseLLMProvider
 from backend.app.llm.local import LocalLLMProvider
+from backend.app.config import settings
 
 logger = logging.getLogger(__name__)
 
+MODEL_REPO = "bartowski/SmolLM2-360M-Instruct-GGUF"
+MODEL_FILE = "SmolLM2-360M-Instruct-Q4_K_M.gguf"
+
+
 class LLMManager:
     def __init__(self):
         self.provider: Optional[BaseLLMProvider] = None
 
+    async def ensure_model_downloaded(self) -> bool:
+        model_path = os.path.abspath(settings.LOCAL_MODEL_PATH)
+        if os.path.exists(model_path):
+            logger.info(f"Model found at {model_path}")
+            return True
+
+        os.makedirs(os.path.dirname(model_path), exist_ok=True)
+        logger.info(f"Downloading {MODEL_FILE} (~180MB) from {MODEL_REPO}...")
+        try:
+            from huggingface_hub import hf_hub_download
+
+            def download():
+                return hf_hub_download(
+                    repo_id=MODEL_REPO, filename=MODEL_FILE,
+                    local_dir=os.path.dirname(model_path),
+                    local_dir_use_symlinks=False,
+                )
+
+            await asyncio.to_thread(download)
+            logger.info("Model download complete!")
+            return True
+        except Exception as e:
+            logger.error(f"Model download failed: {e}")
+            return False
+
     async def setup_provider(self):
         logger.info("Initializing local provider...")
+        await self.ensure_model_downloaded()
         local = LocalLLMProvider()
         await local.initialize()
         self.provider = local
@@ -34,4 +66,5 @@ class LLMManager:
         p = await self.get_active_provider()
         return p.get_info()
 
+
 llm_manager = LLMManager()

backend/app/main.py

@@ -6,7 +6,7 @@ import asyncio
 from contextlib import asynccontextmanager
 from typing import AsyncGenerator, Dict, Any
 
-from fastapi import FastAPI, HTTPException, status
+from fastapi import FastAPI, HTTPException, status, APIRouter
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import StreamingResponse
 from fastapi.staticfiles import StaticFiles
@@ -37,6 +37,7 @@ async def lifespan(app: FastAPI):
 
 
 app = FastAPI(title="Levi AI Coder", version="1.0.0", lifespan=lifespan)
+api = APIRouter(prefix="/api")
 
 app.add_middleware(LoggingMiddleware)
 app.add_middleware(
@@ -109,7 +110,7 @@ async def run_standard(
         raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=str(e))
 
 
-@app.post("/chat")
+@api.post("/chat")
 async def chat(request: ChatRequest):
     msgs = [m.model_dump() for m in request.messages]
     system_prompt = None
@@ -138,7 +139,7 @@ async def chat(request: ChatRequest):
     )
 
 
-@app.post("/complete")
+@api.post("/complete")
 async def complete(request: CompletionRequest):
     prompt = get_completion_prompt(request.prefix, request.suffix, request.language or "python")
     if request.stream:
@@ -151,7 +152,7 @@ async def complete(request: CompletionRequest):
                               max_tokens=request.max_tokens or 128)
 
 
-@app.get("/health")
+@api.get("/health")
 async def health():
     info = await llm_manager.get_status_info()
     return {
@@ -160,6 +161,8 @@ async def health():
     }
 
 
+app.include_router(api)
+
 frontend_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..", "frontend", "dist"))
 if os.path.exists(frontend_dir):
     app.mount("/", StaticFiles(directory=frontend_dir, html=True), name="frontend")

railway.json

@@ -7,7 +7,7 @@
   "deploy": {
     "numReplicas": 1,
     "restartPolicyType": "ON_FAILURE",
-    "healthcheckPath": "/health",
+    "healthcheckPath": "/api/health",
     "healthcheckTimeout": 120
   }
 }

render.yaml

@@ -1,31 +1,19 @@
 services:
   - type: web
-    name: antigravity-ai-coder
+    name: levi-ai-coder
     env: docker
     dockerfilePath: Dockerfile
-    plan: free # Can be upgraded for RAM/volumes
+    plan: free
     envVars:
       - key: PORT
         value: 8000
-      - key: HOST
-        value: 0.0.0.0
-      - key: INFERENCE_MODE
-        value: auto
       - key: LOCAL_MODEL_PATH
-        value: models/Qwen2.5-Coder-0.5B-Instruct-Q4_K_M.gguf
-      - key: HF_MODEL_ID
-        value: Qwen/Qwen2.5-Coder-0.5B-Instruct
-      - key: DEFAULT_TEMPERATURE
-        value: 0.7
+        value: models/SmolLM2-360M-Instruct-Q4_K_M.gguf
+      - key: DEFAULT_CONTEXT_LENGTH
+        value: 512
       - key: DEFAULT_MAX_TOKENS
-        value: 1024
-      - key: RATE_LIMIT_PER_MINUTE
-        value: 60
-      - key: SECRET_KEY
-        generateValue: true
-      - key: HF_API_TOKEN
-        sync: false # Set in Render UI
+        value: 512
     disk:
       name: models-storage
       mountPath: /app/models
-      sizeGB: 10
+      sizeGB: 2