Parseo

Deterministic PDF parsers for underwriting documents. Each parser accepts a Buffer and returns structured, typed JSON with bounding boxes for every extracted field. No file I/O — the consuming application handles reading files.

Packages

Package	Description
@parseo/core	Universal parse(buffer) — auto-classifies and routes to the correct parser
@parseo/shared	PDF text extraction, classifier, error classes, and shared utilities
@parseo/credit-reports	Xactus, PCB, and CreditXpert credit report parsers
@parseo/background-checks	LexisNexis SmartLinx person report parser
@parseo/appraisals	Richer Values, Form 1004MC (URAR), and Form 1073 (Condo) parsers
@parseo/bank-statements	14 bank statement parsers (see below)

Installation

# Universal (includes all parsers)
npm install @parseo/core

# Or install individual packages
npm install @parseo/shared @parseo/bank-statements @parseo/credit-reports @parseo/appraisals @parseo/background-checks

Quick Start

import { parse } from "@parseo/core";

const result = await parse(buffer);

if (result) {
  result.format;      // "chase", "credit-report", "smartlinx", etc.
  result.data;        // Parsed statement/report (type depends on format)
  result.confidence;  // Classifier confidence score
  result.skippedPages; // Number of intro pages stripped
}

parse() handles everything: text extraction, classification, page skipping, and bounding box offset correction. Returns null if no known format is detected.

Classifier

For more control, use the classifier directly via @parseo/shared:

import { extractLines, classifyDocument } from "@parseo/shared";

const lines = await extractLines(buffer);
const result = classifyDocument(lines);
// { format: "chase", startPage: 1, skip: 0, confidence: 28 }

Use the optional scope parameter to limit classification to a single package:

classifyDocument(lines, "bank-statements");
classifyDocument(lines, "credit-reports");
classifyDocument(lines, "appraisals");
classifyDocument(lines, "background-checks");

Returns null if no format is detected.

Bank Statements

import { chase, wellsFargo, pnc } from "@parseo/bank-statements";

const statement = await chase(buffer);
// statement.accountHolder, statement.transactions, statement.summary, etc.

Bank	Import name	Account types
Chase	chase	Business Checking
Wells Fargo	wellsFargo	Business Checking
Bank of America	bankOfAmerica	Business Advantage
TD Bank	tdBank	Business Checking
PNC	pnc	Business Checking
Truist	truist	Simple Business Checking
Capital One	capitalOne	360 Performance Savings
Citibank	citibank	Checking, Savings
Navy Federal	navyFederal	Business Checking, Savings
Relay	relay	Business Checking
Grove Bank	groveBank	Business Checking
Third Federal	thirdFederal	HELOC
Discover	discover	Money Market
Synovus	synovus	Pro Business Checking

All bank statement parsers return a consistent shape:

interface Statement {
  accountHolder: { name: string; address: string; boundingBoxes: Record<string, BoundingBox> };
  accountNumber: string;
  accountType: string;
  statementPeriod: { start: DateString; end: DateString };
  summary: { /* bank-specific fields */ };
  transactions: Transaction[];
  totalDeposits: number;
  totalWithdrawals: number;
  boundingBoxes: Record<string, BoundingBox>;
}

Credit Reports

import { xactus, pcb, creditxpert } from "@parseo/credit-reports";

const report = await xactus(buffer);    // XactusCreditReport
const report = await pcb(buffer);       // PCBCreditReport
const report = await creditxpert(buffer); // CreditXpertReport

Provider	Import name
Xactus "Credit Report X"	xactus
Premium Credit Bureau (PCB)	pcb
CreditXpert (score-only)	creditxpert

Background Checks

import { smartlinx } from "@parseo/background-checks";

const report = await smartlinx(buffer); // SmartLinxReport

Appraisals

import { richerValues, form1004MC, form1073 } from "@parseo/appraisals";

const report = await richerValues(buffer); // RicherValuesReport
const report = await form1004MC(buffer);   // Form1004MCReport
const report = await form1073(buffer);     // Form1073Report

Error Handling

All parsers throw typed errors that extend ParserError:

import {
  ParserError,
  InvalidPDFError,
  UnrecognizedFormatError,
  MissingSectionError,
  ExtractionError,
} from "@parseo/shared";

Error	When
InvalidPDFError	Buffer is empty, not a PDF, or encrypted
UnrecognizedFormatError	PDF text doesn't match expected provider
MissingSectionError	Format matched but required field missing
ExtractionError	No extractable text (scanned image)

Data Conventions

• Dates: ISO 8601 strings ("2024-08-31") or null
• Currency: Plain numbers (54961.89, not "$54,961.89")
• Bounding boxes: { x, y, width, height, pageNumber } on every extracted field
• Withdrawals: Negative amounts on transactions

Project Structure

packages/
  core/                → @parseo/core
  shared/              → @parseo/shared
  credit-reports/      → @parseo/credit-reports
  background-checks/   → @parseo/background-checks
  appraisals/          → @parseo/appraisals
  bank-statements/     → @parseo/bank-statements
    src/
      chase/
      wells-fargo/
      bank-of-america/
      td-bank/
      pnc/
      truist/
      capital-one/
      citibank/
      navy-federal/
      relay/
      grove-bank/
      third-federal/
      discover/
      synovus/

• npm workspaces at root — npm install links everything
• TypeScript project references — npm run build compiles in dependency order
• Each package builds to dist/ with declarations