s

.gitignore

@@ -1 +1,2 @@
 tester.py
+__pycache__/

Dockerfile

@@ -10,4 +10,4 @@ COPY . /code
 RUN pip install --no-cache-dir -r requirements.txt
 RUN chmod -R 777 /code
 
-CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
+CMD ["python", "app.py"]

app.py

@@ -1,50 +1,23 @@
-import os
-import requests
 from fastapi import FastAPI, Request
 from fastapi.responses import JSONResponse
-from llama_cpp import Llama
-
-REPO_ID = "TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF"
-MODEL_FILENAME = "tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf"
-MODEL_PATH = f"./models/{MODEL_FILENAME}"
-HF_TOKEN = os.getenv("HF_TOKEN")
-
-# Ensure models folder exists
-os.makedirs("./models", exist_ok=True)
-os.chmod("./models", 0o777)  # ensure write access
-
-# Download model if missing
-if not os.path.exists(MODEL_PATH):
-    print("📦 Downloading TinyLlama Q4_K_M model...")
-    url = f"https://huggingface.co/{REPO_ID}/resolve/main/{MODEL_FILENAME}"
-    headers = {"Authorization": f"Bearer {HF_TOKEN}"}
-    r = requests.get(url, headers=headers, stream=True)
-    if r.status_code != 200:
-        raise RuntimeError(f"❌ Download failed: {r.status_code} {r.text[:200]}")
-    with open(MODEL_PATH, "wb") as f:
-        for chunk in r.iter_content(8192):
-            f.write(chunk)
-    print("✅ Model downloaded")
-
-# Load into llama-cpp
-print("🔧 Loading TinyLlama model...")
-llm = Llama(
-    model_path=MODEL_PATH,
-    n_ctx=512,
-    n_threads=os.cpu_count() or 1
-)
+from model import generate_structure
+import uvicorn
 
 app = FastAPI()
 
 @app.get("/")
-async def root():
-    return {"status": "🟢 TinyLlama-1.1B Q4_K_M is ready"}
+def index():
+    return {"message": "Minecraft AI Builder Backend is running."}
 
 @app.post("/prompt")
-async def prompt(req: Request):
-    body = await req.json()
-    prompt = body.get("prompt") or ""
-    if not prompt:
-        return JSONResponse(status_code=400, content={"error": "Missing 'prompt' field"})
-    resp = llm(prompt, max_tokens=512, stop=["</s>"])
-    return {"response": resp["choices"][0]["text"].strip()}
+async def prompt_handler(req: Request):
+    try:
+        data = await req.json()
+        prompt = data.get("prompt", "")
+        response = generate_structure(prompt)
+        return {"response": response}
+    except Exception as e:
+        return JSONResponse(status_code=500, content={"error": str(e)})
+
+if __name__ == "__main__":
+    uvicorn.run(app, host="0.0.0.0", port=7860)

model.py

@@ -0,0 +1,37 @@
+import os
+from llama_cpp import Llama
+from huggingface_hub import hf_hub_download
+
+# Model parameters
+REPO_ID = "TheBloke/phi-2-GGUF"
+MODEL_FILENAME = "phi-2.Q4_K_M.gguf"
+MODEL_PATH = f"./models/{MODEL_FILENAME}"
+
+# Auto-download if model not present
+if not os.path.exists(MODEL_PATH):
+    os.makedirs("./models", exist_ok=True)
+    print("📦 Downloading GGUF model manually from Hugging Face...")
+    MODEL_PATH = hf_hub_download(
+        repo_id=REPO_ID,
+        filename=MODEL_FILENAME,
+        local_dir="./models",
+        local_dir_use_symlinks=False
+    )
+    print(f"✅ Model downloaded to {MODEL_PATH}")
+
+# Initialize Llama model
+llm = Llama(
+    model_path=MODEL_PATH,
+    n_ctx=512,
+    n_threads=2,
+    verbose=True
+)
+
+def generate_structure(prompt: str) -> str:
+    response = llm.create_chat_completion(
+        messages=[{"role": "user", "content": prompt}],
+        temperature=0.4,
+        top_p=0.95,
+        max_tokens=1024,
+    )
+    return response["choices"][0]["message"]["content"]

structure_prompt_config.json

@@ -0,0 +1,3 @@
+{
+  "template": "You are a Minecraft-style structure planner. You have the curiosity to build almost anything you could think of. {structure_name}\n\nOnly output a JSON object describing a 3D structure using this format:\n\n{{\n  \"width\": <int>,\n  \"height\": <int>,\n  \"depth\": <int>,\n  \"layers\": [\n    [\n      [\"stone\", \"stone\", \"stone\"],\n      [\"stone\", \"air\",   \"stone\"],\n      [\"stone\", \"stone\", \"stone\"]\n    ],\n    ...\n  ]\n}}\n\nOnly use lowercase Minecraft block IDs (e.g. \"stone\", \"air\", \"glass\", \"planks\").\nYou could only build this structure using {blocks_allowed}\nDo not include any natural language or explanation.\nOutput strictly valid JSON only."
+}