Create app.py

app.py

@@ -0,0 +1,38 @@
+from fastapi import FastAPI
+from pydantic import BaseModel
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+app = FastAPI()
+
+model_name = "deepseek-ai/deepseek-coder-1.3b-instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float32)
+
+class CodeRequest(BaseModel):
+    code: str
+    language: str = "python"
+    max_tokens: int = 128
+
+@app.get("/")
+def root():
+    return {"status": "DevOS AI is running"}
+
+@app.post("/complete")
+def complete_code(request: CodeRequest):
+    prompt = f"Continue the following {request.language} code:\n{request.code}"
+    inputs = tokenizer(prompt, return_tensors="pt")
+    
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=request.max_tokens,
+            temperature=0.2,
+            do_sample=True,
+            pad_token_id=tokenizer.eos_token_id
+        )
+    
+    generated = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    suggestion = generated[len(prompt):]
+    
+    return {"suggestion": suggestion.strip()}