Spaces:
Sleeping
Sleeping
File size: 18,390 Bytes
80b6680 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 | #!/usr/bin/env python3
"""
process_company.py
One-time processing script per company.
Handles PDF-only companies and PDF+XBRL companies automatically.
Usage:
python3 process_company.py "AXIS"
python3 process_company.py --all
python3 process_company.py --list
"""
import sys
import argparse
import re
from pathlib import Path
from datetime import datetime
sys.path.insert(0, str(Path(__file__).resolve().parent))
import config
from db import init_db
from app.services.cache_service import (
upsert_company, save_claims, save_timeseries,
save_risk, save_analysis, get_all_companies,
)
from pipelines.text.extract_text_data import run_text_extraction_pipeline
from pipelines.text.split_sentences import run_sentence_splitting_pipeline
from pipelines.text.claim_extractor import run_claim_extraction_pipeline
from pipelines.ml.run_claim_model import load_claim_model, run_claim_model_pipeline
from pipelines.ml.extract_attributes import run_attribute_extraction_pipeline
from pipelines.finance.extract_xbrl_data import run_xbrl_extraction_pipeline
from pipelines.finance.prepare_timeseries_data import run_timeseries_pipeline
from pipelines.finance.verify_claims import run_claim_verification_pipeline
from pipelines.risk.aggregate_risk import run_risk_aggregation_pipeline
from pipelines.rag.build_vector_db import load_vector_db, build_vector_db
from pipelines.rag.rag_explainer import load_groq_client, run_rag_explanation
# ββ Helpers βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def _log(msg: str):
print(f"[{datetime.now().strftime('%H:%M:%S')}] {msg}")
def _clean_display(folder_name: str) -> str:
name = re.sub(r'\s*\d{4}\s*Quarterly\s*Data\s*', '', folder_name, flags=re.IGNORECASE)
return name.strip()
def _find_folder(query: str):
import difflib
companies_dir = Path(config.DATA_DIR) / "raw" / "companies"
if not companies_dir.exists():
return None
folders = [f for f in companies_dir.iterdir() if f.is_dir()]
query_lower = query.strip().lower()
for f in folders:
if f.name.lower() == query_lower or _clean_display(f.name).lower() == query_lower:
return f
for f in folders:
if query_lower in f.name.lower() or query_lower in _clean_display(f.name).lower():
return f
names = [f.name.lower() for f in folders]
matches = difflib.get_close_matches(query_lower, names, n=1, cutoff=0.5)
if matches:
return folders[names.index(matches[0])]
return None
def _has_xml(folder: Path) -> bool:
"""Check if company folder has any XML files (any depth)."""
return any(folder.rglob("*.xml"))
def _match_records(records: list, folder_name: str) -> list:
folder_lower = folder_name.lower()
display_lower = _clean_display(folder_name).lower()
exact = [r for r in records
if r.get("company", "").lower() in (folder_lower, display_lower)]
if exact:
return exact
return [r for r in records
if display_lower in r.get("company", "").lower()
or r.get("company", "").lower() in folder_lower]
def _signal_title(sentence: str) -> str:
words = sentence.split()
return " ".join(words[:8]).rstrip(",.;") + ("β¦" if len(words) > 8 else "")
def _infer_verdict(pos: list, neg: list) -> str:
if len(pos) > len(neg) * 1.5:
return "Positive"
if len(neg) > len(pos) * 1.5:
return "Negative"
return "Mixed"
def _fmt(value, suffix="") -> str:
if value is None:
return "N/A"
try:
return f"{float(value):,.1f}{suffix}"
except (TypeError, ValueError):
return "N/A"
def _clean_quarter(q: str) -> str:
"""Cleans quarter strings and converts Calendar Years to Indian FY."""
q = str(q).strip().upper()
year, quarter = None, None
m = re.search(r"([1-4])[Qq].*?(?:FY|20)?(\d{2,4})", q)
if not m: m = re.search(r"[Qq]([1-4]).*?(?:FY|20)?(\d{2,4})", q)
if not m:
m = re.match(r"^\d{4}-Q[1-4]$", q)
if m: year, quarter = int(q[:4]), int(q[-1])
if m and not year:
y_val = int(m.group(2))
year = y_val + 2000 if y_val < 100 else y_val
quarter = int(m.group(1))
if year and quarter:
if "FY" in q: return f"{year}-Q{quarter}"
# Indian FY Math: Apr-Dec (Q1, Q2, Q3) -> Next Year. Jan-Mar (Q4) -> Same Year.
fy_year = year + 1 if quarter in [1, 2, 3] else year
return f"{fy_year}-Q{quarter}"
return q
def _trend(change) -> str:
if change is None:
return "flat"
try:
v = float(change)
return "up" if v > 1 else ("down" if v < -1 else "flat")
except (TypeError, ValueError):
return "flat"
def _format_metrics(ts_records: list) -> list:
if not ts_records:
return []
row = sorted(ts_records, key=lambda r: str(r.get("quarter", "")))[-1]
defs = [
("Revenue", "revenue", "βΉ Cr", "revenue_qoq_change"),
("Net Profit", "net_profit", "βΉ Cr", "net_profit_qoq_change"),
("Operating Profit", "operating_profit", "βΉ Cr", "operating_profit_qoq_change"),
("Profit Margin", "profit_margin", "%", "profit_margin_qoq_change"),
]
out = []
for label, key, unit, chg_key in defs:
val = row.get(key)
if val is None:
continue
out.append({
"label": f"{label} Β· {row.get('quarter', '')}",
"value": _fmt(val, f" {unit}"),
"sub": f"QoQ {_fmt(row.get(chg_key), '%')}",
"trend": _trend(row.get(chg_key)),
})
return out
def _format_bar_metrics(ts_records: list) -> list:
if not ts_records:
return []
row = sorted(ts_records, key=lambda r: str(r.get("quarter", "")))[-1]
color_map = {"up": "var(--accent-green)", "down": "var(--accent-red)", "flat": "var(--accent-blue)"}
defs = [
("Revenue QoQ %", "revenue_qoq_change"),
("Net Profit QoQ %", "net_profit_qoq_change"),
("Operating Profit QoQ %", "operating_profit_qoq_change"),
("Profit Margin QoQ %", "profit_margin_qoq_change"),
]
out = []
for label, key in defs:
val = row.get(key)
if val is None:
continue
try:
pct = float(val)
except (TypeError, ValueError):
continue
t = _trend(pct)
out.append({
"label": label,
"target_pct": min(int(abs(pct) * 2), 100),
"color": color_map[t],
"value": _fmt(pct, "%"),
})
return out
def _format_risk_from_claims(attributes: list) -> dict:
"""
Lightweight risk estimate from claim attributes alone (no XBRL needed).
Uses direction distribution as a proxy for consistency.
"""
if not attributes:
return {"level": "UNKNOWN", "consistency_score": None, "warning_flag": 0}
directions = [a.get("direction") for a in attributes if a.get("direction")]
if not directions:
return {"level": "UNKNOWN", "consistency_score": None, "warning_flag": 0}
increase_rate = directions.count("increase") / len(directions)
decrease_rate = directions.count("decrease") / len(directions)
missing_rate = sum(1 for a in attributes if a.get("direction_missing")) / len(attributes)
# Simple heuristic score
score = round(max(min(increase_rate - decrease_rate * 0.5 - missing_rate * 0.2, 1.0), 0.0), 2)
level = "LOW" if score > 0.6 else ("MODERATE" if score > 0.3 else "HIGH")
return {
"level": level,
"consistency_score": score,
"warning_flag": 1 if score < 0.3 else 0,
}
def _format_risk_from_verified(risk_records: list) -> dict:
if not risk_records:
return {"level": "UNKNOWN", "consistency_score": None, "warning_flag": 0}
latest = sorted(risk_records, key=lambda r: str(r.get("quarter", "")))[-1]
score = latest.get("consistency_score", 0)
warn = latest.get("warning_flag", 0)
level = "HIGH" if (warn or score < 0.4) else ("MODERATE" if score < 0.65 else "LOW")
return {
"level": level,
"consistency_score": score,
"warning_flag": warn,
"verification_rate": latest.get("verification_rate"),
"failure_rate": latest.get("failure_rate"),
}
# ββ Core processing βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def process_company(folder: Path) -> bool:
folder_name = folder.name
display_name = _clean_display(folder_name)
_log(f"Processing: {display_name}")
company_id = upsert_company(folder_name, status="processing")
try:
companies_parent = str(folder.parent)
attributes = []
ts_records = []
verified = []
risk_records = []
sentences = []
# ββ 1. Text pipeline (always runs) ββββββββββββββββββββββββββββββββ
_log(" β Extracting text from transcripts...")
raw_records = run_text_extraction_pipeline(companies_parent)
company_docs = _match_records(raw_records, folder_name)
if company_docs:
_log(f" β {len(company_docs)} file(s) found. Splitting sentences...")
sentences = run_sentence_splitting_pipeline(company_docs)
_log(f" β {len(sentences)} sentences. Running claim extraction...")
spacy_claims = run_claim_extraction_pipeline(sentences)
bert_claims = run_claim_model_pipeline(sentences)
claims = bert_claims if bert_claims else spacy_claims
_log(f" β {len(claims)} claims found. Extracting attributes...")
attributes = run_attribute_extraction_pipeline(claims)
_log(f" β {len(attributes)} attribute records extracted.")
else:
_log(" β No transcript files found.")
# ββ 2. Financial data via XBRL/XML ββββββββββββββββββββββββββββββββ
_log(" β Extracting quarterly financials from local XBRL/XML files...")
has_xbrl = False
if _has_xml(folder):
try:
# Extract XBRL records (filtering for this specific company)
all_fin_records = run_xbrl_extraction_pipeline(companies_parent)
fin_records = [r for r in all_fin_records if r.get("company") == folder_name]
if fin_records:
ts_records = run_timeseries_pipeline(fin_records)
save_timeseries(company_id, ts_records)
_log(f" β {len(ts_records)} timeseries rows saved.")
has_xbrl = True
if attributes:
_log(" β Verifying claims against financials...")
verified = run_claim_verification_pipeline(attributes, ts_records)
risk_records = run_risk_aggregation_pipeline(verified)
save_risk(company_id, risk_records)
_log(f" β {len(verified)} claims verified.")
else:
_log(" β No financial data parsed from XMLs β skipping verification.")
except Exception as e:
_log(f" β Financials error: {e}")
else:
_log(" β No XML files found in folder β skipping verification.")
# ββ 3. Save claims to DB ββββββββββββββββββββββββββββββββββββββββββ
# Use verified claims if available, otherwise raw attributes
claims_to_save = verified if verified else attributes
if claims_to_save:
save_claims(company_id, claims_to_save)
# ββ 4. Load ALL chunks into ChromaDB βββββββββββββββββββββββββββββ
_log(" β Loading ALL transcript chunks into vector DB...")
if sentences:
# We use 'sentences' instead of 'verified' so RAG has full context
for s in sentences:
s.setdefault("result", "CONTEXT")
s.setdefault("reason", "TRANSCRIPT_EXCERPT")
status = build_vector_db(sentences, company=display_name)
_log(f" β Vector DB: {status}")
else:
_log(" β No text chunks to load into vector DB.")
# ββ 5. RAG + cache all 3 modes ββββββββββββββββββββββββββββββββββββ
_log(" β Generating RAG explanations...")
query = f"{display_name} financial performance earnings"
# Risk dict
risk_dict = (
_format_risk_from_verified(risk_records)
if risk_records
else _format_risk_from_claims(attributes)
)
# Display claims for UI (top 20) with clean quarter formats
display_claims = [
{
"quarter": _clean_quarter(a.get("quarter", "")),
"sentence": str(a.get("sentence", ""))[:140],
"metric": a.get("metric", ""),
"direction": a.get("direction") or "neutral",
"result": a.get("result", "UNVERIFIED"),
"confidence": a.get("actual_change") or 0.0,
}
for a in (verified or attributes)[:20]
]
quarters = sorted({
_clean_quarter(r.get("quarter", "")) for r in (verified or attributes)
if r.get("quarter")
})
for mode in ("full", "earnings", "financial"):
try:
rag = run_rag_explanation(query, company=display_name, top_k=config.RAG_TOP_K)
pos_raw = rag.get("positive_signals", [])
neg_raw = rag.get("negative_signals", [])
result = {
"company": display_name,
"mode": mode,
"matched_folder": folder_name,
"has_xbrl": has_xbrl,
"positive_signals": [
{"index": f"P{i+1}", "title": _signal_title(s), "body": s}
for i, s in enumerate(pos_raw)
],
"negative_signals": [
{"index": f"N{i+1}", "title": _signal_title(s), "body": s}
for i, s in enumerate(neg_raw)
],
"explanation": rag.get("explanation", ""),
"verdict": _infer_verdict(pos_raw, neg_raw),
"metrics": _format_metrics(ts_records),
"bar_metrics": _format_bar_metrics(ts_records),
"risk": risk_dict,
"claims": display_claims,
"source_label": (
f"{display_name} Β· {', '.join(quarters)}"
if quarters else display_name
),
}
save_analysis(company_id, mode, result)
except Exception as exc:
_log(f" β Mode '{mode}' failed: {exc}")
# ββ 6. Mark ready βββββββββββββββββββββββββββββββββββββββββββββββββ
upsert_company(folder_name, status="ready")
_log(f" β {display_name} done.\n")
return True
except Exception as exc:
upsert_company(folder_name, status="error", error_msg=str(exc))
_log(f" β Failed: {exc}\n")
import traceback; traceback.print_exc()
return False
# ββ CLI βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def cmd_list():
companies = get_all_companies()
if not companies:
print("No companies registered yet.")
return
print(f"\n{'Display Name':<30} {'Status':<12} {'Processed At'}")
print("-" * 65)
for c in companies:
print(f"{c['display_name']:<30} {c['status']:<12} {c['processed_at'] or 'β'}")
print()
def cmd_process_all():
companies_dir = Path(config.DATA_DIR) / "raw" / "companies"
if not companies_dir.exists():
print(f"Not found: {companies_dir}")
return
folders = [f for f in sorted(companies_dir.iterdir()) if f.is_dir()]
_log(f"Found {len(folders)} company folders.")
ok = fail = 0
for folder in folders:
if process_company(folder):
ok += 1
else:
fail += 1
_log(f"Done. {ok} succeeded, {fail} failed.")
def main():
_log("Loading models...")
load_claim_model(config.MODEL_PATH)
load_vector_db(config.CHROMA_DB_PATH)
load_groq_client()
init_db()
_log("Ready.\n")
parser = argparse.ArgumentParser()
group = parser.add_mutually_exclusive_group(required=True)
group.add_argument("company", nargs="?", help="Company name")
group.add_argument("--all", action="store_true")
group.add_argument("--list", action="store_true")
args = parser.parse_args()
if args.list:
cmd_list(); return
if args.all:
cmd_process_all(); return
folder = _find_folder(args.company)
if not folder:
print(f"No matching folder for: '{args.company}'")
companies_dir = Path(config.DATA_DIR) / "raw" / "companies"
if companies_dir.exists():
print("Available:")
for f in sorted(companies_dir.iterdir()):
if f.is_dir():
print(f" {f.name}")
sys.exit(1)
process_company(folder)
if __name__ == "__main__":
main() |