Create handler.py

handler.py

@@ -0,0 +1,29 @@
+# handler.py
+from transformers import pipeline
+
+# this loads the files in the repo root
+generator = pipeline(
+    "text2text-generation",
+    model=".", 
+    tokenizer="."
+)
+
+def init():
+    # optional: run once at container startup
+    pass
+
+def inference(payload: dict):
+    """
+    payload is the body of the POST, e.g.
+      { "inputs": "VIBE: Agnostic\nQUESTION: …\nAnswer:" }
+    """
+    text = payload.get("inputs", "")
+    out = generator(
+        text, 
+        max_new_tokens=128, 
+        do_sample=True,
+        top_p=0.9,
+        temperature=0.7
+    )
+    # return whatever JSON you want clients to see
+    return {"generated_text": out[0]["generated_text"]}