Talk2TaskDemo1 / app.py
RishiRP's picture
Update app.py
954d97c verified
raw
history blame
33.1 kB
# app.py
import os
import re
import io
import json
import time
import zipfile
from pathlib import Path
from typing import List, Dict, Any, Tuple, Optional
import numpy as np
import pandas as pd
import gradio as gr
import torch
from transformers import (
AutoTokenizer,
AutoModelForCausalLM,
BitsAndBytesConfig,
GenerationConfig,
)
# =========================
# Global config
# =========================
SPACE_CACHE = Path.home() / ".cache" / "huggingface"
SPACE_CACHE.mkdir(parents=True, exist_ok=True)
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
GEN_CONFIG = GenerationConfig(
temperature=0.2,
top_p=0.9,
do_sample=False,
max_new_tokens=256,
)
# Official UBS label set (strict)
OFFICIAL_LABELS = [
"plan_contact",
"schedule_meeting",
"update_contact_info_non_postal",
"update_contact_info_postal_address",
"update_kyc_activity",
"update_kyc_origin_of_assets",
"update_kyc_purpose_of_businessrelation",
"update_kyc_total_assets",
]
OFFICIAL_LABELS_TEXT = "\n".join(OFFICIAL_LABELS)
# Per-label keyword cues (static prompt context to improve recall)
LABEL_KEYWORDS: Dict[str, List[str]] = {
"plan_contact": [
"call back", "follow up", "reach out", "contact later", "check-in",
"email them", "touch base", "remind", "send a note"
],
"schedule_meeting": [
"book a meeting", "set up a meeting", "schedule a call",
"appointment", "calendar", "meeting next week", "meet on", "time slot"
],
"update_contact_info_non_postal": [
"phone change", "new phone", "email change", "new email",
"update contact details", "update mobile", "alternate phone"
],
"update_contact_info_postal_address": [
"moved to", "new address", "postal address", "mailing address",
"change of address", "residential address"
],
"update_kyc_activity": [
"activity update", "economic activity", "employment status",
"occupation", "job change", "business activity"
],
"update_kyc_origin_of_assets": [
"source of funds", "origin of assets", "where money comes from",
"inheritance", "salary", "business income", "asset origin"
],
"update_kyc_purpose_of_businessrelation": [
"purpose of relationship", "why the account", "reason for banking",
"investment purpose", "relationship purpose"
],
"update_kyc_total_assets": [
"total assets", "net worth", "assets under ownership",
"portfolio size", "how much you own"
],
}
# =========================
# Instructions (string-safe; concatenated)
# =========================
SYSTEM_PROMPT = (
"You are a precise banking assistant that extracts ACTIONABLE TASKS from "
"client–advisor transcripts. Be conservative with hallucinations but "
"prioritise RECALL: if unsure and the transcript plausibly implies an "
"action, include the label and explain briefly.\n\n"
"Output STRICT JSON only:\n\n"
"{\n"
' "labels": ["<Label1>", "..."],\n'
' "tasks": [\n'
' {"label": "<Label1>", "explanation": "<why>", "evidence": "<quoted text/snippet>"}\n'
" ]\n"
"}\n\n"
"Rules:\n"
"- Use ONLY allowed labels supplied to you. Case-insensitive during reasoning, "
" but output the canonical label text exactly.\n"
"- If none truly apply, return empty lists.\n"
"- Keep explanations concise; put the minimal evidence snippet that justifies the task.\n"
)
USER_PROMPT_TEMPLATE = (
"Transcript (cleaned):\n"
"```\n{transcript}\n```\n\n"
"Allowed Labels (canonical; use only these):\n"
"{allowed_labels_list}\n\n"
"Context cues (keywords/phrases that often indicate each label):\n"
"{keyword_context}\n\n"
"Instructions:\n"
"- Identify EVERY concrete task implied by the conversation.\n"
"- Choose ONE label from Allowed Labels for each task (or none if truly inapplicable).\n"
"- Return STRICT JSON only in the exact schema described by the system prompt.\n"
)
# =========================
# Utilities
# =========================
def _now_ms() -> int:
return int(time.time() * 1000)
def normalize_labels(labels: List[str]) -> List[str]:
return list(dict.fromkeys([l.strip() for l in labels if isinstance(l, str) and l.strip()]))
def canonicalize_map(allowed: List[str]) -> Dict[str, str]:
return {lab.lower(): lab for lab in allowed}
def robust_json_extract(text: str) -> Dict[str, Any]:
if not text:
return {"labels": [], "tasks": []}
start, end = text.find("{"), text.rfind("}")
candidate = text[start:end+1] if (start != -1 and end != -1 and end > start) else text
try:
return json.loads(candidate)
except Exception:
candidate = re.sub(r",\s*}", "}", candidate)
candidate = re.sub(r",\s*]", "]", candidate)
try:
return json.loads(candidate)
except Exception:
return {"labels": [], "tasks": []}
def restrict_to_allowed(pred: Dict[str, Any], allowed: List[str]) -> Dict[str, Any]:
out = {"labels": [], "tasks": []}
allowed_map = canonicalize_map(allowed)
# labels
filt_labels = []
for l in pred.get("labels", []) or []:
k = str(l).strip().lower()
if k in allowed_map:
filt_labels.append(allowed_map[k])
filt_labels = normalize_labels(filt_labels)
# tasks
filt_tasks = []
for t in pred.get("tasks", []) or []:
if not isinstance(t, dict):
continue
k = str(t.get("label", "")).strip().lower()
if k in allowed_map:
new_t = dict(t); new_t["label"] = allowed_map[k]
filt_tasks.append(new_t)
merged = normalize_labels(list(set(filt_labels) | {tt["label"] for tt in filt_tasks}))
out["labels"] = merged
out["tasks"] = filt_tasks
return out
# =========================
# Default pre-processing (toggleable)
# =========================
_DISCLAIMER_PATTERNS = [
r"(?is)^\s*(?:disclaimer|legal notice|confidentiality notice).+?(?:\n{2,}|$)",
r"(?is)^\s*the information contained.+?(?:\n{2,}|$)",
r"(?is)^\s*this message \(including any attachments\).+?(?:\n{2,}|$)",
]
_FOOTER_PATTERNS = [
r"(?is)\n+kind regards[^\n]*\n.*$", r"(?is)\n+best regards[^\n]*\n.*$",
r"(?is)\n+sent from my.*$", r"(?is)\n+ubs ag.*$",
]
_TIMESTAMP_SPEAKER = [
r"\[\d{1,2}:\d{2}(:\d{2})?\]", # [00:01] or [00:01:02]
r"^\s*(advisor|client)\s*:\s*", # Advisor: / Client:
r"^\s*(speaker\s*\d+)\s*:\s*", # Speaker 1:
]
def clean_transcript(text: str) -> str:
if not text:
return text
s = text
# remove timestamps/speaker prefixes line-wise
lines = []
for ln in s.splitlines():
ln2 = ln
for pat in _TIMESTAMP_SPEAKER:
ln2 = re.sub(pat, "", ln2, flags=re.IGNORECASE)
lines.append(ln2)
s = "\n".join(lines)
# remove top disclaimers
for pat in _DISCLAIMER_PATTERNS:
s = re.sub(pat, "", s).strip()
# remove trailing footers
for pat in _FOOTER_PATTERNS:
s = re.sub(pat, "", s)
# collapse whitespace
s = re.sub(r"[ \t]+", " ", s)
s = re.sub(r"\n{3,}", "\n\n", s).strip()
return s
def read_text_file_any(file_input) -> str:
"""Works for gr.File(type='filepath') and raw strings/Path and file-like."""
if not file_input:
return ""
if isinstance(file_input, (str, Path)):
try:
return Path(file_input).read_text(encoding="utf-8", errors="ignore")
except Exception:
return ""
try:
data = file_input.read()
return data.decode("utf-8", errors="ignore")
except Exception:
return ""
def read_json_file_any(file_input) -> Optional[dict]:
if not file_input:
return None
if isinstance(file_input, (str, Path)):
try:
return json.loads(Path(file_input).read_text(encoding="utf-8", errors="ignore"))
except Exception:
return None
try:
return json.loads(file_input.read().decode("utf-8", errors="ignore"))
except Exception:
return None
def truncate_tokens(tokenizer, text: str, max_tokens: int) -> str:
toks = tokenizer(text, add_special_tokens=False)["input_ids"]
if len(toks) <= max_tokens:
return text
return tokenizer.decode(toks[-max_tokens:], skip_special_tokens=True)
# =========================
# HF model wrapper
# =========================
class ModelWrapper:
def __init__(self, repo_id: str, hf_token: Optional[str], load_in_4bit: bool):
self.repo_id = repo_id
self.hf_token = hf_token
self.load_in_4bit = load_in_4bit
self.tokenizer = None
self.model = None
def load(self):
qcfg = None
if self.load_in_4bit and DEVICE == "cuda":
qcfg = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_compute_dtype=torch.float16,
bnb_4bit_use_double_quant=True,
)
tok = AutoTokenizer.from_pretrained(
self.repo_id, token=self.hf_token, cache_dir=str(SPACE_CACHE),
trust_remote_code=True, use_fast=True,
)
if tok.pad_token is None and tok.eos_token:
tok.pad_token = tok.eos_token
model = AutoModelForCausalLM.from_pretrained(
self.repo_id, token=self.hf_token, cache_dir=str(SPACE_CACHE),
trust_remote_code=True,
torch_dtype=torch.float16 if DEVICE == "cuda" else torch.float32,
device_map="auto" if DEVICE == "cuda" else None,
low_cpu_mem_usage=True, quantization_config=qcfg,
attn_implementation="sdpa",
)
self.tokenizer = tok
self.model = model
@torch.inference_mode()
def generate(self, system_prompt: str, user_prompt: str) -> str:
# Build inputs as input_ids=... (avoid **tensor bug from earlier)
if hasattr(self.tokenizer, "apply_chat_template"):
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt},
]
input_ids = self.tokenizer.apply_chat_template(
messages,
tokenize=True,
add_generation_prompt=True,
return_tensors="pt",
)
input_ids = input_ids.to(self.model.device)
gen_kwargs = dict(
input_ids=input_ids,
generation_config=GEN_CONFIG,
eos_token_id=self.tokenizer.eos_token_id,
pad_token_id=self.tokenizer.pad_token_id,
)
else:
enc = self.tokenizer(
f"<s>[SYSTEM]\n{system_prompt}\n[/SYSTEM]\n[USER]\n{user_prompt}\n[/USER]\n",
return_tensors="pt"
).to(self.model.device)
gen_kwargs = dict(
**enc,
generation_config=GEN_CONFIG,
eos_token_id=self.tokenizer.eos_token_id,
pad_token_id=self.tokenizer.pad_token_id,
)
with torch.cuda.amp.autocast(enabled=(DEVICE == "cuda")):
out_ids = self.model.generate(**gen_kwargs)
return self.tokenizer.decode(out_ids[0], skip_special_tokens=True)
_MODEL_CACHE: Dict[str, ModelWrapper] = {}
def get_model(repo_id: str, hf_token: Optional[str], load_in_4bit: bool) -> ModelWrapper:
key = f"{repo_id}::{'4bit' if (load_in_4bit and DEVICE=='cuda') else 'full'}"
if key not in _MODEL_CACHE:
m = ModelWrapper(repo_id, hf_token, load_in_4bit)
m.load()
_MODEL_CACHE[key] = m
return _MODEL_CACHE[key]
# =========================
# Official evaluation (from README)
# =========================
def evaluate_predictions(y_true: List[List[str]], y_pred: List[List[str]]) -> float:
ALLOWED_LABELS = OFFICIAL_LABELS
LABEL_TO_IDX = {label: idx for idx, label in enumerate(ALLOWED_LABELS)}
def _process_sample_labels(sample_labels: List[str], sample_name: str) -> List[str]:
if not isinstance(sample_labels, list):
raise ValueError(f"{sample_name} must be a list of strings, got {type(sample_labels)}")
seen, uniq = set(), []
for label in sample_labels:
if not isinstance(label, str):
raise ValueError(f"{sample_name} contains non-string: {label} (type: {type(label)})")
if label in seen:
raise ValueError(f"{sample_name} contains duplicate label: '{label}'")
if label not in ALLOWED_LABELS:
raise ValueError(f"{sample_name} contains invalid label: '{label}'. Allowed: {ALLOWED_LABELS}")
seen.add(label); uniq.append(label)
return uniq
if len(y_true) != len(y_pred):
raise ValueError(f"y_true and y_pred must have same length. Got {len(y_true)} vs {len(y_pred)}")
n_samples = len(y_true)
n_labels = len(OFFICIAL_LABELS)
y_true_binary = np.zeros((n_samples, n_labels), dtype=int)
y_pred_binary = np.zeros((n_samples, n_labels), dtype=int)
for i, sample_labels in enumerate(y_true):
for label in _process_sample_labels(sample_labels, f"y_true[{i}]"):
y_true_binary[i, LABEL_TO_IDX[label]] = 1
for i, sample_labels in enumerate(y_pred):
for label in _process_sample_labels(sample_labels, f"y_pred[{i}]"):
y_pred_binary[i, LABEL_TO_IDX[label]] = 1
fn = np.sum((y_true_binary == 1) & (y_pred_binary == 0), axis=1) # penalty 2x
fp = np.sum((y_true_binary == 0) & (y_pred_binary == 1), axis=1) # penalty 1x
weighted = 2.0 * fn + 1.0 * fp
max_err = 2.0 * np.sum(y_true_binary, axis=1) + 1.0 * (n_labels - np.sum(y_true_binary, axis=1))
per_sample = np.where(max_err > 0, 1.0 - (weighted / max_err), 1.0)
return float(max(0.0, min(1.0, np.mean(per_sample))))
# =========================
# Fallback: keyword heuristics if model returns empty
# =========================
def keyword_fallback(text: str, allowed: List[str]) -> Dict[str, Any]:
low = text.lower()
labels = []
tasks = []
for lab in allowed:
hits = []
for kw in LABEL_KEYWORDS.get(lab, []):
k = kw.lower()
if k in low:
# capture small evidence window
i = low.find(k)
start = max(0, i - 40); end = min(len(text), i + len(k) + 40)
hits.append(text[start:end].strip())
if hits:
labels.append(lab)
tasks.append({
"label": lab,
"explanation": "Keyword match in transcript.",
"evidence": hits[0]
})
return {"labels": normalize_labels(labels), "tasks": tasks}
# =========================
# Inference helpers
# =========================
def build_keyword_context(allowed: List[str]) -> str:
parts = []
for lab in allowed:
kws = LABEL_KEYWORDS.get(lab, [])
parts.append(f"- {lab}: " + (", ".join(kws) if kws else "(no default cues)"))
return "\n".join(parts)
def run_single(
transcript_text: str,
transcript_file, # filepath or file-like
gt_json_text: str,
gt_json_file, # filepath or file-like
use_cleaning: bool,
use_keyword_fallback: bool,
allowed_labels_text: str,
model_repo: str,
use_4bit: bool,
max_input_tokens: int,
hf_token: str,
) -> Tuple[str, str, str, str, str, str, str]:
t0 = _now_ms()
# Transcript
raw_text = ""
if transcript_file:
raw_text = read_text_file_any(transcript_file)
raw_text = (raw_text or transcript_text or "").strip()
if not raw_text:
return "", "", "No transcript provided.", "", "", "", ""
text = clean_transcript(raw_text) if use_cleaning else raw_text
# Allowed labels (pre-filled defaults)
user_allowed = [ln.strip() for ln in (allowed_labels_text or "").splitlines() if ln.strip()]
allowed = normalize_labels(user_allowed or OFFICIAL_LABELS)
# Model
try:
model = get_model(model_repo, (hf_token or "").strip() or None, use_4bit)
except Exception as e:
return "", "", f"Model load failed: {e}", "", "", "", ""
# Truncate
trunc = truncate_tokens(model.tokenizer, text, max_input_tokens)
# Build prompt
allowed_list_str = "\n".join(f"- {l}" for l in allowed)
keyword_ctx = build_keyword_context(allowed)
user_prompt = USER_PROMPT_TEMPLATE.format(
transcript=trunc,
allowed_labels_list=allowed_list_str,
keyword_context=keyword_ctx,
)
# Generate
t1 = _now_ms()
try:
out = model.generate(SYSTEM_PROMPT, user_prompt)
except Exception as e:
return "", "", f"Generation error: {e}", "", "", "", ""
t2 = _now_ms()
parsed = robust_json_extract(out)
filtered = restrict_to_allowed(parsed, allowed)
# Fallback if empty
if use_keyword_fallback and not filtered.get("labels"):
fb = keyword_fallback(trunc, allowed)
if fb["labels"]:
filtered = fb
# Diagnostics
diag = "\n".join([
f"Device: {DEVICE} (4-bit: {'Yes' if (use_4bit and DEVICE=='cuda') else 'No'})",
f"Model: {model_repo}",
f"Input cleaned: {'Yes' if use_cleaning else 'No'}",
f"Keyword fallback: {'Yes' if use_keyword_fallback else 'No'}",
f"Tokens (input, approx): ≤ {max_input_tokens}",
f"Latency: prep {t1-t0} ms, gen {t2-t1} ms, total {t2-t0} ms",
f"Allowed labels: {', '.join(allowed)}",
])
# Context & instructions preview shown in UI
context_preview = (
"### Allowed Labels\n"
+ "\n".join(f"- {l}" for l in allowed)
+ "\n\n### Keyword cues per label\n"
+ keyword_ctx
)
instructions_preview = "```\n" + SYSTEM_PROMPT + "\n```"
# Summary & JSON
labs = filtered.get("labels", [])
tasks = filtered.get("tasks", [])
summary = "Detected labels:\n" + ("\n".join(f"- {l}" for l in labs) if labs else "(none)")
if tasks:
summary += "\n\nTasks:\n" + "\n".join(
f"• [{t['label']}] {t.get('explanation','')} | ev: {t.get('evidence','')[:140]}{'…' if len(t.get('evidence',''))>140 else ''}"
for t in tasks
)
else:
summary += "\n\nTasks: (none)"
json_out = json.dumps(filtered, indent=2, ensure_ascii=False)
# Optional single-file scoring if GT provided
metrics = ""
true_labels = None
if gt_json_file or (gt_json_text and gt_json_text.strip()):
truth_obj = None
if gt_json_file:
truth_obj = read_json_file_any(gt_json_file)
if (not truth_obj) and gt_json_text:
try:
truth_obj = json.loads(gt_json_text)
except Exception:
pass
if isinstance(truth_obj, dict) and isinstance(truth_obj.get("labels"), list):
true_labels = [x for x in truth_obj["labels"] if x in OFFICIAL_LABELS]
pred_labels = labs
try:
score = evaluate_predictions([true_labels], [pred_labels])
tp = len(set(true_labels) & set(pred_labels))
fp = len(set(pred_labels) - set(true_labels))
fn = len(set(true_labels) - set(pred_labels))
recall = tp / (tp + fn) if (tp + fn) else 1.0
precision = tp / (tp + fp) if (tp + fp) else 1.0
f1 = (2 * precision * recall / (precision + recall)) if (precision + recall) else 1.0
metrics = (
f"Weighted score: {score:.3f}\n"
f"Recall: {recall:.3f} | Precision: {precision:.3f} | F1: {f1:.3f}\n"
f"TP={tp} FP={fp} FN={fn}\n"
f"Truth: {', '.join(true_labels)}"
)
except Exception as e:
metrics = f"Scoring error: {e}"
else:
metrics = "Ground truth JSON missing or invalid; expected {'labels': [...]}."
return summary, json_out, diag, out.strip(), context_preview, instructions_preview, metrics
# =========================
# Batch mode (ZIP with transcripts + truths)
# =========================
def read_zip_from_path(path: str, exdir: Path) -> List[Path]:
exdir.mkdir(parents=True, exist_ok=True)
with open(path, "rb") as f:
data = f.read()
with zipfile.ZipFile(io.BytesIO(data)) as zf:
zf.extractall(exdir)
return [p for p in exdir.rglob("*") if p.is_file()]
def run_batch(
zip_path, # filepath string
use_cleaning: bool,
use_keyword_fallback: bool,
model_repo: str,
use_4bit: bool,
max_input_tokens: int,
hf_token: str,
limit_files: int,
) -> Tuple[str, str, pd.DataFrame, str]:
if not zip_path:
return ("No ZIP provided.", "", pd.DataFrame(), "")
work = Path("/tmp/batch")
if work.exists():
for p in sorted(work.rglob("*"), reverse=True):
try: p.unlink()
except Exception: pass
try: work.rmdir()
except Exception: pass
work.mkdir(parents=True, exist_ok=True)
files = read_zip_from_path(zip_path, work)
txts: Dict[str, Path] = {}
gts: Dict[str, Path] = {}
for p in files:
if p.suffix.lower() == ".txt":
txts[p.stem] = p
elif p.suffix.lower() == ".json":
gts[p.stem] = p
stems = sorted(txts.keys())
if limit_files > 0:
stems = stems[:limit_files]
if not stems:
return ("No .txt transcripts found in ZIP.", "", pd.DataFrame(), "")
try:
model = get_model(model_repo, (hf_token or "").strip() or None, use_4bit)
except Exception as e:
return (f"Model load failed: {e}", "", pd.DataFrame(), "")
allowed = OFFICIAL_LABELS[:]
allowed_list_str = "\n".join(f"- {l}" for l in allowed)
keyword_ctx = build_keyword_context(allowed)
y_true, y_pred = [], []
rows = []
t_start = _now_ms()
for stem in stems:
raw = txts[stem].read_text(encoding="utf-8", errors="ignore")
text = clean_transcript(raw) if use_cleaning else raw
trunc = truncate_tokens(model.tokenizer, text, max_input_tokens)
user_prompt = USER_PROMPT_TEMPLATE.format(
transcript=trunc,
allowed_labels_list=allowed_list_str,
keyword_context=keyword_ctx,
)
t0 = _now_ms()
out = model.generate(SYSTEM_PROMPT, user_prompt)
t1 = _now_ms()
parsed = robust_json_extract(out)
filtered = restrict_to_allowed(parsed, allowed)
if use_keyword_fallback and not filtered.get("labels"):
fb = keyword_fallback(trunc, allowed)
if fb["labels"]:
filtered = fb
pred_labels = filtered.get("labels", [])
y_pred.append(pred_labels)
gt_labels = []
if stem in gts:
try:
gt_obj = json.loads(gts[stem].read_text(encoding="utf-8", errors="ignore"))
if isinstance(gt_obj, dict) and isinstance(gt_obj.get("labels"), list):
gt_labels = [x for x in gt_obj["labels"] if x in OFFICIAL_LABELS]
except Exception:
pass
y_true.append(gt_labels)
gt_set, pr_set = set(gt_labels), set(pred_labels)
tp = sorted(gt_set & pr_set)
fp = sorted(pr_set - gt_set)
fn = sorted(gt_set - pr_set)
rows.append({
"file": stem,
"true_labels": ", "..join(gt_labels),
"pred_labels": ", ".join(pred_labels),
"TP": len(tp), "FP": len(fp), "FN": len(fn),
"gen_ms": t1 - t0
})
have_truth = any(len(v) > 0 for v in y_true)
score = evaluate_predictions(y_true, y_pred) if have_truth else None
df = pd.DataFrame(rows).sort_values(["FN", "FP", "file"])
diag = [
f"Processed files: {len(stems)}",
f"Device: {DEVICE} (4-bit: {'Yes' if (use_4bit and DEVICE=='cuda') else 'No'})",
f"Model: {model_repo}",
f"Input cleaned: {'Yes' if use_cleaning else 'No'}",
f"Keyword fallback: {'Yes' if use_keyword_fallback else 'No'}",
f"Tokens (input, approx): ≤ {max_input_tokens}",
f"Batch time: {_now_ms()-t_start} ms",
]
if have_truth and score is not None:
total_tp = int(df["TP"].sum())
total_fp = int(df["FP"].sum())
total_fn = int(df["FN"].sum())
recall = total_tp / (total_tp + total_fn) if (total_tp + total_fn) else 1.0
precision = total_tp / (total_tp + total_fp) if (total_tp + total_fp) else 1.0
f1 = (2 * precision * recall / (precision + recall)) if (precision + recall) else 1.0
diag += [
f"Official weighted score (0–1): {score:.3f}",
f"Recall: {recall:.3f} | Precision: {precision:.3f} | F1: {f1:.3f}",
f"Total TP={total_tp} FP={total_fp} FN={total_fn}",
]
diag_str = "\n".join(diag)
# save CSV for download
out_csv = Path("/tmp/batch_results.csv")
df.to_csv(out_csv, index=False, encoding="utf-8")
return ("Batch done.", diag_str, df, str(out_csv))
# =========================
# UI
# =========================
MODEL_CHOICES = [
"swiss-ai/Apertus-8B-Instruct-2509",
"meta-llama/Meta-Llama-3-8B-Instruct",
"mistralai/Mistral-7B-Instruct-v0.3",
]
custom_css = """
:root { --radius: 14px; }
.gradio-container { font-family: Inter, ui-sans-serif, system-ui; }
.card { border: 1px solid rgba(255,255,255,.08); border-radius: var(--radius); padding: 14px 16px; background: rgba(255,255,255,.02); box-shadow: 0 1px 10px rgba(0,0,0,.12) inset; }
.header { font-weight: 700; font-size: 22px; margin-bottom: 4px; }
.subtle { color: rgba(255,255,255,.65); font-size: 14px; margin-bottom: 12px; }
hr.sep { border: none; border-top: 1px solid rgba(255,255,255,.08); margin: 10px 0 16px; }
.accordion-title { font-weight: 600; }
.gr-button { border-radius: 12px !important; }
"""
with gr.Blocks(theme=gr.themes.Soft(), css=custom_css, fill_height=True) as demo:
gr.Markdown("<div class='header'>Talk2Task — Task Extraction (UBS Challenge)</div>")
gr.Markdown("<div class='subtle'>False negatives are penalised 2× more than false positives in the official score. This UI biases for recall, shows the exact instructions & context, and supports single or batch evaluation.</div>")
with gr.Tab("Single transcript"):
with gr.Row():
with gr.Column(scale=3):
gr.Markdown("<div class='card'><div class='header'>Transcript</div>", elem_id="card1")
file = gr.File(
label="Drag & drop transcript (.txt / .md / .json)",
file_types=[".txt", ".md", ".json"],
type="filepath",
)
text = gr.Textbox(label="Or paste transcript", lines=10)
gr.Markdown("<hr class='sep'/>")
gr.Markdown("<div class='header'>Ground truth JSON (optional)</div>", elem_id="card1b")
gt_file = gr.File(
label="Upload ground truth JSON (expects {'labels': [...]})",
file_types=[".json"],
type="filepath",
)
gt_text = gr.Textbox(label="Or paste ground truth JSON", lines=6, placeholder='{\"labels\": [\"schedule_meeting\"]}')
gr.Markdown("</div>") # close card
gr.Markdown("<div class='card'><div class='header'>Preprocessing & heuristics</div>", elem_id="card2")
use_cleaning = gr.Checkbox(
label="Apply default cleaning (remove disclaimers, timestamps, speakers, footers)",
value=True,
)
use_keyword_fallback = gr.Checkbox(
label="Keyword fallback if model returns empty",
value=True,
)
gr.Markdown("</div>")
gr.Markdown("<div class='card'><div class='header'>Allowed labels</div>", elem_id="card3")
labels_text = gr.Textbox(
label="Allowed Labels (one per line)",
value=OFFICIAL_LABELS_TEXT, # prefilled
lines=8,
)
reset_btn = gr.Button("Reset to official labels")
gr.Markdown("</div>")
with gr.Column(scale=2):
gr.Markdown("<div class='card'><div class='header'>Model & run</div>", elem_id="card4")
repo = gr.Dropdown(label="Model", choices=MODEL_CHOICES, value=MODEL_CHOICES[0])
use_4bit = gr.Checkbox(label="Use 4-bit (GPU only)", value=True)
max_tokens = gr.Slider(label="Max input tokens", minimum=1024, maximum=8192, step=512, value=4096)
hf_token = gr.Textbox(label="HF_TOKEN (only for gated models)", type="password", value=os.environ.get("HF_TOKEN",""))
run_btn = gr.Button("Run Extraction", variant="primary")
gr.Markdown("</div>")
gr.Markdown("<div class='card'><div class='header'>Outputs</div>", elem_id="card5")
summary = gr.Textbox(label="Summary", lines=12)
json_out = gr.Code(label="Strict JSON Output", language="json")
diag = gr.Textbox(label="Diagnostics", lines=8)
raw = gr.Textbox(label="Raw Model Output", lines=8)
gr.Markdown("</div>")
with gr.Row():
with gr.Column():
with gr.Accordion("Instructions used (system prompt)", open=False):
instr_md = gr.Markdown("")
with gr.Column():
with gr.Accordion("Context used (allowed labels + keyword cues)", open=True):
context_md = gr.Markdown("")
# reset button behavior
def _reset_labels():
return OFFICIAL_LABELS_TEXT
reset_btn.click(fn=_reset_labels, inputs=None, outputs=labels_text)
# single run
def _pack_context_md(allowed: str) -> str:
allowed_list = [ln.strip() for ln in (allowed or OFFICIAL_LABELS_TEXT).splitlines() if ln.strip()]
ctx = build_keyword_context(allowed_list)
return "### Allowed Labels\n" + "\n".join(f"- {l}" for l in allowed_list) + "\n\n### Keyword cues per label\n" + ctx
run_btn.click(
fn=run_single,
inputs=[
text, file, gt_text, gt_file, use_cleaning, use_keyword_fallback,
labels_text, repo, use_4bit, max_tokens, hf_token
],
outputs=[summary, json_out, diag, raw, context_md, instr_md, gr.Textbox(visible=False)],
)
# also keep instructions visible at initial load
instr_md.value = "```\n" + SYSTEM_PROMPT + "\n```"
context_md.value = _pack_context_md(OFFICIAL_LABELS_TEXT)
with gr.Tab("Batch evaluation"):
with gr.Row():
with gr.Column(scale=3):
gr.Markdown("<div class='card'><div class='header'>ZIP input</div>", elem_id="card6")
zip_in = gr.File(label="ZIP with transcripts (.txt) and truths (.json)", file_types=[".zip"], type="filepath")
use_cleaning_b = gr.Checkbox(label="Apply default cleaning", value=True)
use_keyword_fallback_b = gr.Checkbox(label="Keyword fallback if model returns empty", value=True)
gr.Markdown("</div>")
with gr.Column(scale=2):
gr.Markdown("<div class='card'><div class='header'>Model & run</div>", elem_id="card7")
repo_b = gr.Dropdown(label="Model", choices=MODEL_CHOICES, value=MODEL_CHOICES[0])
use_4bit_b = gr.Checkbox(label="Use 4-bit (GPU only)", value=True)
max_tokens_b = gr.Slider(label="Max input tokens", minimum=1024, maximum=8192, step=512, value=4096)
hf_token_b = gr.Textbox(label="HF_TOKEN (only for gated models)", type="password", value=os.environ.get("HF_TOKEN",""))
limit_files = gr.Slider(label="Process at most N files (0 = all)", minimum=0, maximum=2000, step=10, value=0)
run_batch_btn = gr.Button("Run Batch", variant="primary")
gr.Markdown("</div>")
with gr.Row():
gr.Markdown("<div class='card'><div class='header'>Batch outputs</div>", elem_id="card8")
status = gr.Textbox(label="Status", lines=1)
diag_b = gr.Textbox(label="Batch diagnostics & metrics", lines=12)
df_out = gr.Dataframe(label="Per-file results (TP/FP/FN, latency)", interactive=False)
csv_out = gr.File(label="Download CSV", interactive=False)
gr.Markdown("</div>")
run_batch_btn.click(
fn=run_batch,
inputs=[zip_in, use_cleaning_b, use_keyword_fallback_b, repo_b, use_4bit_b, max_tokens_b, hf_token_b, limit_files],
outputs=[status, diag_b, df_out, csv_out],
)
if __name__ == "__main__":
demo.launch()