Excel Processing

ADX uses openpyxl to parse Excel files (.xlsx) into structured DocumentGraph objects. This guide covers spreadsheet-specific features.

What Gets Extracted

• Sheets with dimensions, visibility, and named ranges
• Cells with values, types, and formatting info
• Formulas with computed values and dependency tracking
• Hidden sheets, rows, and columns — flagged but still accessible
• Merged cells — detected and reported
• Comments — extracted as metadata
• Named ranges — mapped to their cell references
• External links — detected and flagged

Dual-Mode Loading

ADX loads each workbook twice:

1. Formula mode — preserves formula text (e.g., =SUM(A1:A10))
2. Data-only mode — reads calculated values (e.g., 42.5)

This means agents can see both what a cell computes and what it says.

Formula Inspection

The inspect_formula tool traces formula dependencies:

info = dn.inspect_formula(doc_id, sheet="Summary", cell="B10")

Returns:

• The formula text
• The computed value
• All referenced cells and ranges
• Cross-sheet references
• External workbook references

Hidden Content Detection

ADX tracks hidden elements:

• Hidden sheets (visible in list_sheets with hidden: true)
• Hidden rows and columns (flagged in read_range results)
• Cells in hidden rows/columns are accessible but marked

Named Ranges

Named ranges are extracted and available through list_sheets:

sheets = dn.list_sheets(doc_id)
for sheet in sheets["sheets"]:
    for nr in sheet.get("named_ranges", []):
        print(f"  {nr['name']}: {nr['range']}")

Cell Search

Find cells by value across sheets:

results = dn.find_cells(doc_id, query="Revenue")
for match in results["matches"]:
    print(f"  {match['sheet']}!{match['address']}: {match['value']}")

Range Reading

Read a rectangular range of cells:

data = dn.read_range(doc_id, sheet="Assumptions", range="A1:C20")
for cell in data["cells"]:
    print(f"  {cell['address']}: {cell['value']} (formula: {cell.get('formula', 'N/A')})")

Example

from adx import ADX

dn = ADX()
doc_id = dn.upload("financial_model.xlsx")

# List all sheets
sheets = dn.list_sheets(doc_id)
for s in sheets["sheets"]:
    visibility = "hidden" if s["hidden"] else "visible"
    print(f"  {s['name']}: {s['rows']}x{s['cols']} ({visibility})")

# Read a range
data = dn.read_range(doc_id, sheet="Summary", range="A1:D10")

# Inspect a formula
formula = dn.inspect_formula(doc_id, sheet="Summary", cell="D10")
print(f"Formula: {formula['formula']}")
print(f"Value: {formula['computed_value']}")
print(f"References: {formula['references']}")

# Extract financial model fields
extraction = dn.extract(doc_id, schema="financial_model")

Limitations

• Only .xlsx format is supported (not .xls or .xlsb)
• Pivot tables are not parsed (read as static values)
• Charts are not extracted
• Macros (VBA) are not executed or analyzed
• Conditional formatting rules are not extracted