chore: auto-commit uncommitted changes

ai.go

@@ -1,6 +1,7 @@
 package main
 
 import (
+	"archive/zip"
 	"bytes"
 	"crypto/sha256"
 	"encoding/base64"
@@ -617,6 +618,22 @@ func ProcessDocument(filePath string) (*Document, error) {
 			UpdateJob(hash, "error", err.Error())
 			return nil, fmt.Errorf("text analysis failed: %w", err)
 		}
+	} else if IsOfficeFile(ext) {
+		// Office formats — extract text natively from ZIP/XML, no LibreOffice needed
+		UpdateJob(hash, "converting", "Extracting text...")
+		log.Printf("  Extracting text from %s...", ext)
+		text, err := ExtractOfficeText(filePath)
+		if err != nil {
+			UpdateJob(hash, "error", err.Error())
+			return nil, fmt.Errorf("office text extraction failed: %w", err)
+		}
+		log.Printf("  Extracted %d chars, classifying...", len(text))
+		UpdateJob(hash, "classifying", "Classifying...")
+		analysis, err = AnalyzeText(text, filepath.Base(filePath))
+		if err != nil {
+			UpdateJob(hash, "error", err.Error())
+			return nil, fmt.Errorf("text analysis failed: %w", err)
+		}
 	} else {
 		// Vision — convert to image and analyze
 		UpdateJob(hash, "converting", "Converting to image...")
@@ -751,3 +768,124 @@ func copyFile(src, dst string) error {
 	_, err = io.Copy(out, in)
 	return err
 }
+
+// ExtractOfficeText extracts plain text from DOCX/XLSX/PPTX natively.
+// These are ZIP archives containing XML — no LibreOffice needed.
+func ExtractOfficeText(filePath string) (string, error) {
+	ext := strings.ToLower(filepath.Ext(filePath))
+
+	r, err := zip.OpenReader(filePath)
+	if err != nil {
+		return "", fmt.Errorf("not a valid Office file: %w", err)
+	}
+	defer r.Close()
+
+	// Which XML paths to extract per format
+	var targets []string
+	switch ext {
+	case ".docx", ".doc", ".odt", ".rtf":
+		targets = []string{"word/document.xml", "word/body.xml"}
+	case ".xlsx", ".xls":
+		// All sheets
+		for _, f := range r.File {
+			if strings.HasPrefix(f.Name, "xl/worksheets/sheet") && strings.HasSuffix(f.Name, ".xml") {
+				targets = append(targets, f.Name)
+			}
+		}
+		// Shared strings (cell values are stored here for xlsx)
+		targets = append(targets, "xl/sharedStrings.xml")
+	case ".pptx", ".ppt":
+		for _, f := range r.File {
+			if strings.HasPrefix(f.Name, "ppt/slides/slide") && strings.HasSuffix(f.Name, ".xml") {
+				targets = append(targets, f.Name)
+			}
+		}
+	default:
+		return "", fmt.Errorf("unsupported office format: %s", ext)
+	}
+
+	var sb strings.Builder
+	for _, f := range r.File {
+		found := false
+		for _, t := range targets {
+			if f.Name == t {
+				found = true
+				break
+			}
+		}
+		if !found {
+			continue
+		}
+
+		rc, err := f.Open()
+		if err != nil {
+			continue
+		}
+		data, err := io.ReadAll(rc)
+		rc.Close()
+		if err != nil {
+			continue
+		}
+
+		sb.WriteString(xmlToText(data))
+		sb.WriteString("\n")
+	}
+
+	text := strings.TrimSpace(sb.String())
+	if text == "" {
+		return "", fmt.Errorf("no text extracted from %s", filepath.Base(filePath))
+	}
+	return text, nil
+}
+
+// xmlToText strips XML tags and decodes entities, returning plain text.
+func xmlToText(data []byte) string {
+	var sb strings.Builder
+	inTag := false
+	lastWasSpace := false
+
+	for i := 0; i < len(data); i++ {
+		c := data[i]
+		switch {
+		case c == '<':
+			inTag = true
+			// Emit space between elements so words don't run together
+			if !lastWasSpace {
+				sb.WriteByte(' ')
+				lastWasSpace = true
+			}
+		case c == '>':
+			inTag = false
+		case !inTag:
+			if c == ' ' || c == '\t' || c == '\n' || c == '\r' {
+				if !lastWasSpace {
+					sb.WriteByte(' ')
+					lastWasSpace = true
+				}
+			} else {
+				sb.WriteByte(c)
+				lastWasSpace = false
+			}
+		}
+	}
+
+	// Decode common XML entities
+	result := sb.String()
+	result = strings.ReplaceAll(result, "&amp;", "&")
+	result = strings.ReplaceAll(result, "&lt;", "<")
+	result = strings.ReplaceAll(result, "&gt;", ">")
+	result = strings.ReplaceAll(result, "&quot;", "\"")
+	result = strings.ReplaceAll(result, "&apos;", "'")
+	result = strings.ReplaceAll(result, "&#xA;", "\n")
+	result = strings.ReplaceAll(result, "&#x9;", "\t")
+
+	return strings.TrimSpace(result)
+}
+
+// IsOfficeFile returns true for formats with extractable XML text.
+func IsOfficeFile(ext string) bool {
+	return map[string]bool{
+		".docx": true, ".xlsx": true, ".pptx": true,
+		".odt": true, ".ods": true, ".odp": true,
+	}[ext]
+}