Update app.py

app.py

@@ -1,4 +1,4 @@
-# app.py — EduPrompt API (per-task lazy load + cache-safe on Spaces, debug prints)
+# app.py — EduPrompt API (per-task lazy load, cache-safe, no cache_dir in inference)
 
 import os
 
@@ -87,6 +87,11 @@ class InputData(BaseModel):
     input: str
     params: dict | None = None
 
+def filter_model_kwargs(params):
+    # Remove keys not accepted by model.__call__()
+    forbidden = {"cache_dir"}
+    return {k: v for k, v in (params or {}).items() if k not in forbidden}
+
 @app.post("/run")
 async def run_task(data: InputData):
     start = time.time()
@@ -105,19 +110,22 @@ async def run_task(data: InputData):
         print(traceback.format_exc())
         return {"error": f"model_load_failed: {type(e).__name__}: {str(e)}"}
 
+    # Filter out forbidden kwargs
+    params = filter_model_kwargs(data.params)
+
     try:
         if task == "summarize":
             prompt = f"You are an expert explainer. Summarize clearly and concisely:\n{text}"
-            output = model(prompt, max_length=120, min_length=30, truncation=True, do_sample=False)[0]["summary_text"]
+            output = model(prompt, max_length=120, min_length=30, truncation=True, do_sample=False, **params)[0]["summary_text"]
         elif task == "rewrite":
             prompt = f"You are a writing assistant. Rewrite this text for clarity and tone:\n{text}"
-            output = model(prompt, max_new_tokens=150, truncation=True)[0]["generated_text"]
+            output = model(prompt, max_new_tokens=150, truncation=True, **params)[0]["generated_text"]
         elif task == "proofread":
             prompt = f"Correct and improve grammar and style:\n{text}"
-            output = model(prompt, max_new_tokens=150, truncation=True)[0]["generated_text"]
+            output = model(prompt, max_new_tokens=150, truncation=True, **params)[0]["generated_text"]
         else:  # explain_code
             prompt = f"Explain what this code does in simple language:\n{text}"
-            output = model(prompt, max_new_tokens=200, truncation=True)[0]["generated_text"]
+            output = model(prompt, max_new_tokens=200, truncation=True, **params)[0]["generated_text"]
     except Exception as e:
         import traceback
         print(traceback.format_exc())