added some enhancements

app.py

@@ -1,5 +1,6 @@
 # app.py
-# CLI-like Spelling Bee Tutor as a simple Gradio UI (no LLMs, all offline)
+# NeMo Guardrails Demo (starter UI) — CLI-style spelling quiz from words.csv
+# All offline; no LLMs. Ready for Hugging Face Spaces.
 
 from pathlib import Path
 import os
@@ -15,7 +16,7 @@ EXPECTED_COLS = ["word", "difficulty", "definition", "origin", "sentence"]
 def _load_words(path: Path = WORDS_PATH) -> pd.DataFrame:
     """Load a possibly-messy CSV:
        - tolerate commas inside sentence/origin
-       - accept exact 5 columns or more (extras glued into sentence)
+       - accept 5+ columns (extras glued into sentence)
        - create missing optional columns
     """
     if not path.exists():
@@ -48,20 +49,18 @@ def _load_words(path: Path = WORDS_PATH) -> pd.DataFrame:
             for row in reader:
                 if not row or all((x is None or str(x).strip() == "") for x in row):
                     continue
-                # expect at least 5 fields; if more, glue extras into sentence
                 if len(row) < 5:
-                    # pad empty fields to avoid crash
                     row = row + [""] * (5 - len(row))
                 base = row[:4]
                 sentence = ",".join(row[4:])  # glue any extras back
                 rows.append([*(str(x).strip() for x in base), sentence.strip()])
             df = pd.DataFrame(rows, columns=EXPECTED_COLS)
 
-    # --- Normalize columns ---
+    # normalize header names
     mapper = {c: c.strip().lower() for c in df.columns}
     df = df.rename(columns=mapper)
 
-    # If there are more than 5 columns, rebuild to our schema
+    # rebuild to our schema when there are >5 columns
     if df.shape[1] >= 5:
         def pick(colname, default=""):
             return df[colname] if colname in df.columns else default
@@ -87,25 +86,24 @@ def _load_words(path: Path = WORDS_PATH) -> pd.DataFrame:
         base_df["sentence"] = sentence_series
         df = base_df
     else:
-        # ensure missing optional cols exist
         for c in ["definition", "origin", "sentence", "difficulty"]:
             if c not in df.columns:
                 df[c] = ""
 
-    # Clean and type
+    # clean + types
     df["word"] = df["word"].astype(str).fillna("").str.strip()
     df["definition"] = df["definition"].astype(str).fillna("").str.strip()
     df["origin"] = df["origin"].astype(str).fillna("").str.strip()
     df["sentence"] = df["sentence"].astype(str).fillna("").str.strip()
 
-    # difficulty score
+    # difficulty score (numeric difficulty if present; else word length)
     if "difficulty" in df.columns:
         ds = pd.to_numeric(df["difficulty"], errors="coerce").fillna(0)
         df["difficulty_score"] = ds
     else:
         df["difficulty_score"] = df["word"].astype(str).str.len()
 
-    # sort hardest → easiest
+    # hardest → easiest
     df = df.sort_values("difficulty_score", ascending=False).reset_index(drop=True)
     return df
 
@@ -126,7 +124,7 @@ def start_quiz(n_words, state):
         return (state,
                 "words.csv not found or empty.",
                 gr.update(value="", interactive=False),
-                "Please add a words.csv with at least one 'word' column.",
+                "Please add a words.csv with header: word,difficulty,definition,origin,sentence",
                 "Score: 0/0",
                 gr.update(value="", interactive=False))
 
@@ -134,9 +132,11 @@ def start_quiz(n_words, state):
         n = int(n_words or 5)
     except Exception:
         n = 5
-    n = max(1, min(n, len(df)))  # clip to available size
+    n = max(1, min(n, len(df)))  # clip
+
+    # take top-n hardest words, then randomize order for the round
+    block = df.head(n).sample(frac=1, random_state=None).reset_index(drop=True)
 
-    block = df.head(n).reset_index(drop=True)
     s = {
         "i": 0,
         "n": n,
@@ -148,7 +148,7 @@ def start_quiz(n_words, state):
     }
 
     current = s["words"][0]
-    hist = f"Okay! We'll do {n} words this round, hardest → easiest.\nSpell this word: {current}"
+    hist = f"Okay! We'll do {n} words this round, hardest → easiest (shuffled).\nSpell this word: {current}"
     status = "Type your spelling attempt, or click definition/origin/sentence."
     return s, hist, gr.update(value=current, interactive=False), status, f"Score: 0/{n}", gr.update(value="", interactive=True)
 
@@ -204,6 +204,14 @@ def show_sentence(state, history, current_word, score_md):
     history = f"{history}\nTutor (sentence with {word}): {text}"
     return state, history, current_word, f"Sentence: {text}", score_md
 
+def show_answer(state, history, current_word, score_md):
+    if not _check_state(state):
+        return state, history, current_word, "No round active. Click Start.", score_md
+    i = state["i"]
+    word = state["words"][i]
+    history = f"{history}\nTutor: The correct spelling is **{word}**."
+    return state, history, current_word, f"Answer: {word}", score_md
+
 def next_word(state, history, current_word, score_md):
     if not _check_state(state):
         return state, history, current_word, "No round active. Click Start.", score_md
@@ -224,6 +232,20 @@ def stop_round(state, history, current_word, score_md):
     history = f"{history}\n{summary}"
     return {}, history, gr.update(value="", interactive=False), "Stopped.", f"Score: {state['score']}/{state['n']}"
 
+# ---------- Upload handling ----------
+
+def load_uploaded(file):
+    """Replace words.csv with the uploaded file and refresh DF."""
+    global DF
+    try:
+        # Copy uploaded file bytes into WORDS_PATH
+        with open(file.name, "rb") as src, open(WORDS_PATH, "wb") as dst:
+            dst.write(src.read())
+        DF = _load_words()  # refresh global dataset
+        return "Loaded new words.csv!", f"Loaded {len(DF)} rows."
+    except Exception as e:
+        return "", f"Upload failed: {e}"
+
 # ---------- UI ----------
 
 with gr.Blocks() as demo:
@@ -233,17 +255,23 @@ with gr.Blocks() as demo:
         n_words = gr.Number(value=5, precision=0, label="Words this round")
         start = gr.Button("Start quiz")
 
+    # Optional: allow swapping in a new CSV live
+    upload = gr.File(label="Upload words.csv", file_types=[".csv"], interactive=True)
+
     current_word = gr.Textbox(label="Spell this word", interactive=False)
     attempt = gr.Textbox(label="Your attempt")
+
     with gr.Row():
-        check = gr.Button("Check")
+        check = gr.Button("check")
         bdef = gr.Button("definition")
         borg = gr.Button("origin")
         bsent = gr.Button("sentence")
+        bshow = gr.Button("show answer")
         bnext = gr.Button("next")
         bstop = gr.Button("stop")
 
     status = gr.Markdown("")
+    upload_status = gr.Markdown("")  # feedback for uploads
     score_md = gr.Markdown("Score: 0/0")
     history = gr.Textbox(label="History", lines=14)
 
@@ -262,11 +290,15 @@ with gr.Blocks() as demo:
                [state, history, current_word, status, score_md], queue=False)
     bsent.click(show_sentence, [state, history, current_word, score_md],
                 [state, history, current_word, status, score_md], queue=False)
+    bshow.click(show_answer, [state, history, current_word, score_md],
+                [state, history, current_word, status, score_md], queue=False)
     bnext.click(next_word, [state, history, current_word, score_md],
                 [state, history, current_word, status, score_md], queue=False)
     bstop.click(stop_round, [state, history, current_word, score_md],
                 [state, history, current_word, status, score_md], queue=False)
 
+    upload.upload(load_uploaded, [upload], [status, upload_status])
+
 # ---------- Launch (blocking on Spaces) ----------
 
 if __name__ == "__main__":