app.py

import gradio as gr
import pandas as pd
from pathlib import Path
from huggingface_hub import HfApi, Repository

# Allowed tags
LABELS = {"PER", "ORG", "LOC", "O"}

token_df = pd.DataFrame()  # global store

# ───────────────────────── token explode ───────────────────────
def explode(df: pd.DataFrame) -> pd.DataFrame:
    """Return DataFrame(example_id, token, label='O')."""
    if "text" in df.columns:
        lines = df["text"].astype(str)
    else:  # user / assistant dialogs
        lines = df.apply(lambda r: f"User: {r['user']} Assistant: {r['assistant']}", axis=1)

    rows = []
    for sid, line in enumerate(lines, start=0):      # ensure unique 0,1,2,...
        for tok in line.split():
            rows.append({"example_id": sid, "token": tok, "label": "O"})
    return pd.DataFrame(rows)

# ───────────────────────── callbacks ───────────────────────────
def load_csv(file):
    global token_df

    df = pd.read_csv(file.name)
    valid = ("text" in df.columns) or ({"user", "assistant"}.issubset(df.columns))
    if not valid:
        msg = "❌ CSV needs a `text` column **or** both `user` and `assistant` columns."
        return None, msg, *(gr.update(visible=False),) * 3

    token_df = explode(df)

    return (gr.update(value=token_df, visible=True, row_count=len(token_df)),
            f"✅ {len(df)} rows → {len(token_df)} tokens.",
            gr.update(visible=True),  # show buttons
            gr.update(visible=False),  # reset download links
            gr.update(visible=False))

def save_table(tbl):
    global token_df
    token_df = pd.DataFrame(tbl, columns=["example_id", "token", "label"])
    bad = token_df.loc[~token_df["label"].isin(LABELS), "label"].unique()
    return "💾 Saved." if bad.size == 0 else f"⚠️ Unknown label(s): {', '.join(bad)}"

def export_tokens():
    fname = "raw_tokens.csv"
    token_df.to_csv(fname, index=False)
    return gr.update(value=fname, visible=True)

def export_iob():
    iob, prev = [], {}
    for _, r in token_df.iterrows():
        sid, lbl = r["example_id"], r["label"]
        if lbl == "O":
            iob.append("O"); prev[sid] = None
        else:
            iob.append(("I-" if prev.get(sid) == lbl else "B-") + lbl)
            prev[sid] = lbl
    out = token_df.copy(); out["iob"] = iob
    fname = "ner_iob.csv"; out.to_csv(fname, index=False)
    return gr.update(value=fname, visible=True)

def push_to_hub(repo_id, token):
    try:
        HfApi().create_repo(repo_id, token=token, repo_type="dataset", exist_ok=True)
        local = Path(repo_id.replace("/", "_"))
        if local.exists():
            for f in local.iterdir(): f.unlink()
            local.rmdir()
        repo = Repository(str(local), clone_from=repo_id,
                          repo_type="dataset", use_auth_token=token)
        token_df.to_csv(local / "data.csv", index=False)
        repo.push_to_hub("Add annotated NER data")
        return f"🚀 https://huggingface.co/datasets/{repo_id}"
    except Exception as e:
        return f"❌ {e}"

# ───────────────────────── UI ──────────────────────────────────
with gr.Blocks() as demo:
    gr.Markdown("# 🏷️ Label It! Mini-NER")

    gr.Markdown("**Step 1** – upload CSV (columns: `text` **or** `user`+`assistant`).")

    with gr.Row():
        csv_file = gr.File(file_types=[".csv"])
        load_btn = gr.Button("Load")

    status = gr.Textbox(interactive=False)

    tok_table = gr.Dataframe(headers=["example_id", "token", "label"],
                             datatype=["number", "str", "str"],
                             visible=False)

    with gr.Row(visible=False) as buttons:
        save_btn = gr.Button("💾 Save")
        tok_btn  = gr.Button("⬇︎ Tokens CSV")
        iob_btn  = gr.Button("⬇︎ IOB CSV")

    file_tok = gr.File(visible=False)
    file_iob = gr.File(visible=False)

    with gr.Accordion("📦 Push to Hugging Face Hub", open=False, visible=False) as acc:
        repo_in, token_in = gr.Textbox(label="repo"), gr.Textbox(label="token", type="password")
        push_btn   = gr.Button("Push")
        push_out   = gr.Textbox(interactive=False)

    # wiring
    load_btn.click(load_csv, csv_file, [tok_table, status, buttons, file_tok, file_iob])
    load_btn.click(lambda: gr.update(visible=True), None, acc)

    save_btn.click(save_table, tok_table, status)
    tok_btn.click(export_tokens, outputs=file_tok)
    iob_btn.click(export_iob,  outputs=file_iob)
    push_btn.click(push_to_hub, [repo_in, token_in], push_out)

    gr.Markdown("**Step 2** – label tokens (`PER`, `ORG`, `LOC`, `O`) ➜ Save ➜ Download / Push.")

demo.launch()