from pathlib import Path
from typing import Dict, Any, List
from utils.pdf_reader import read_pdf_text
from utils.docx_reader import read_docx_text
from utils.xlsx_reader import read_xlsx_text
from utils.clean_text import clean_text
from utils.classify_document import classify_document_type
KEYWORDS = [
"pricing schedule",
"price schedule",
"bill of quantities",
"boq",
"pricing instructions",
"fees and rates",
"price list",
]
def _read_any(file_path: str) -> str:
path = Path(file_path)
suffix = path.suffix.lower()
if suffix == ".pdf":
return read_pdf_text(path)
if suffix in (".doc", ".docx"):
return read_docx_text(path)
if suffix in (".xls", ".xlsx"):
return read_xlsx_text(path)
# Fallback to plain text
try:
return path.read_text(encoding="utf-8", errors="ignore")
except Exception:
return ""
def _extract_pricing_snippets(text: str) -> List[str]:
lines = [ln.strip() for ln in text.splitlines() if ln.strip()]
snippets = []
for i, line in enumerate(lines):
lower = line.lower()
if any(k in lower for k in KEYWORDS):
window = lines[max(0, i - 3) : min(len(lines), i + 10)]
snippets.append("\n".join(window))
return snippets
def extract_pricing_requirements(file_path: str) -> Dict[str, Any]:
"""Core logic for `detect_pricing_requirements` MCP tool."""
raw = _read_any(file_path)
cleaned = clean_text(raw)
doc_type = classify_document_type(cleaned)
snippets = _extract_pricing_snippets(cleaned)
excerpt = "\n\n".join(snippets[:5]) if snippets else cleaned[:4000]
pricing_items = []
# Very light heuristic: look for lines that look like an item + number
for line in excerpt.splitlines():
if any(ch.isdigit() for ch in line) and len(line) > 20:
pricing_items.append(
{
"raw_line": line.strip(),
"description_guess": line.strip()[:120],
"quantity_guess": None,
"unit_guess": None,
}
)
result: Dict[str, Any] = {
"instructions": (
"You are a tender pricing assistant. The input document has been "
f"classified as '{doc_type}'. Use the 'pricing_items' as a rough "
"starting point only. Cross‑check everything against the original "
"tender and ask the user to paste key sections (pricing schedule, "
"BOQ, instructions) if they are not clearly visible here."
),
"summary": f"Document appears to be a {doc_type} with "
f"{len(pricing_items)} possible pricing‑related lines detected.",
"currency": "ZAR",
"pricing_items": pricing_items,
"raw_text": excerpt,
"file_path": file_path,
"document_type": doc_type,
}
return result
