Module: Clacky::Utils::FileProcessor

Defined in:
lib/clacky/utils/file_processor.rb

Overview

File processing pipeline.

Two entry points:

FileProcessor.save(body:, filename:)
  → Store raw bytes to disk only. Returns { name:, path: }.
    Used by http_server and channel adapters — no parsing here.

FileProcessor.process_path(path, name: nil)
  → Parse an already-saved file. Returns FileRef (with preview_path or parse_error).
    Used by agent.run when building the file prompt.

(FileProcessor.process = save + process_path in one call, for convenience.)

Defined Under Namespace

Classes: FileRef

Constant Summary collapse

UPLOAD_DIR = 
File.join(Dir.tmpdir, "clacky-uploads").freeze
MAX_FILE_BYTES =

32 MB

32 * 1024 * 1024
MAX_IMAGE_BYTES =

5 MB

5 * 1024 * 1024
MAX_FILE_SIZE =

Alias used by FileReader tool

MAX_FILE_BYTES
IMAGE_MAX_WIDTH =

Images wider than this will be downscaled before sending to LLM (pixels)

800
IMAGE_MAX_BASE64_BYTES =

Hard limit for images that can’t be resized: Anthropic/Bedrock vision API supports up to 5MB

5_000_000
BINARY_EXTENSIONS = 
%w[
  .png .jpg .jpeg .gif .webp .bmp .tiff .ico .svg
  .pdf
  .zip .gz .tgz .tar .rar .7z
  .exe .dll .so .dylib
  .mp3 .mp4 .avi .mov .mkv .wav .flac
  .ttf .otf .woff .woff2
  .db .sqlite .bin .dat
].freeze
GLOB_ALLOWED_BINARY_EXTENSIONS = 
%w[
  .pdf .doc .docx .ppt .pptx .xls .xlsx .odt .odp .ods
].freeze
LLM_BINARY_EXTENSIONS = 
%w[.png .jpg .jpeg .gif .webp .pdf].freeze
MIME_TYPES = 
{
  ".png"  => "image/png",
  ".jpg"  => "image/jpeg",
  ".jpeg" => "image/jpeg",
  ".gif"  => "image/gif",
  ".webp" => "image/webp",
  ".pdf"  => "application/pdf"
}.freeze
FILE_TYPES = 
{
  ".docx" => :document,  ".doc"  => :document,
  ".xlsx" => :spreadsheet, ".xls" => :spreadsheet,
  ".pptx" => :presentation, ".ppt" => :presentation,
  ".pdf"  => :pdf,
  ".zip"  => :zip, ".gz" => :zip, ".tgz" => :zip, ".tar" => :zip, ".rar" => :zip, ".7z" => :zip,
  ".png"  => :image, ".jpg" => :image, ".jpeg" => :image,
  ".gif"  => :image, ".webp" => :image,
  ".csv"  => :csv,
  ".md"   => :text, ".markdown" => :text, ".txt" => :text, ".log" => :text
}.freeze
TEXT_PREVIEW_EXTENSIONS =

Plain-text extensions whose raw content can be embedded directly as the preview (no external parser needed). Kept conservative to avoid pulling in huge source files by mistake.

%w[.md .markdown .txt .log].freeze

Class Method Summary collapse

Class Method Details

.binary_file_path?(path) ⇒ Boolean

File type helpers (used by tools and agent)

Returns:

  • (Boolean) 


199
200
201
202
203
204
205
 
# File 'lib/clacky/utils/file_processor.rb', line 199

def self.binary_file_path?(path)
  ext = File.extname(path).downcase
  return true if BINARY_EXTENSIONS.include?(ext)
  File.binread(path, 512).to_s.include?("\x00")
rescue
  false
end

.detect_image_mime_type(data, fallback_mime = "image/png") ⇒ String

Detect the actual image MIME type from raw binary data by inspecting magic bytes, ignoring the file extension. Falls back to extension-based detection when magic bytes don’t match any known format.

Handles: PNG, JPEG, GIF, WEBP, BMP, TIFF

Parameters:

  • data (String) —

    raw binary data (first 12 bytes is sufficient)

  • fallback_mime (String) (defaults to: "image/png") —

    MIME type from extension, used as fallback

Returns:

  • (String) —

    detected MIME type (e.g. “image/png”, “image/jpeg”)



445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
 
# File 'lib/clacky/utils/file_processor.rb', line 445

def self.detect_image_mime_type(data, fallback_mime = "image/png")
  return fallback_mime if data.nil? || data.bytesize < 4

  bytes = data.bytes

  case
  # PNG: \x89 P N G \r \n \x1a \n
  when bytes[0] == 0x89 && bytes[1] == 0x50 && bytes[2] == 0x4E && bytes[3] == 0x47
    "image/png"
  # JPEG: \xFF \xD8 \xFF
  when bytes[0] == 0xFF && bytes[1] == 0xD8 && bytes[2] == 0xFF
    "image/jpeg"
  # GIF: GIF87a or GIF89a
  when bytes[0] == 0x47 && bytes[1] == 0x49 && bytes[2] == 0x46 && bytes[3] == 0x38
    "image/gif"
  # WEBP: RIFF .... WEBP
  when bytes[0] == 0x52 && bytes[1] == 0x49 && bytes[2] == 0x46 && bytes[3] == 0x46 &&
       data.bytesize >= 12 && data[8, 4] == "WEBP"
    "image/webp"
  # BMP: BM
  when bytes[0] == 0x42 && bytes[1] == 0x4D
    "image/bmp"
  # TIFF: II*\x00 (little-endian) or MM\x00* (big-endian)
  when (bytes[0] == 0x49 && bytes[1] == 0x49 && bytes[2] == 0x2A && bytes[3] == 0x00) ||
       (bytes[0] == 0x4D && bytes[1] == 0x4D && bytes[2] == 0x00 && bytes[3] == 0x2A)
    "image/tiff"
  else
    fallback_mime
  end
end

.detect_mime_type(path, _data = nil) ⇒ Object 



215
216
217
 
# File 'lib/clacky/utils/file_processor.rb', line 215

def self.detect_mime_type(path, _data = nil)
  MIME_TYPES[File.extname(path).downcase] || "application/octet-stream"
end

.downscale_image_base64(b64, mime_type, max_width: IMAGE_MAX_WIDTH) ⇒ String

Downscale a base64-encoded image so its width is at most max_width pixels.

Strategy:

PNG  → chunky_png (pure Ruby, always available as gem dependency)
other formats (JPG/WEBP/GIF) → sips on macOS, `convert` (ImageMagick) on Linux
fallback (no CLI tool) → return as-is, but raise if larger than IMAGE_MAX_BASE64_BYTES

Parameters:

  • b64 (String) —

    base64-encoded image data

  • mime_type (String) —

    e.g. “image/png”, “image/jpeg”, “image/webp”

  • max_width (Integer) (defaults to: IMAGE_MAX_WIDTH) —

    maximum output width in pixels (default: IMAGE_MAX_WIDTH)

Returns:

  • (String) —

    base64-encoded (possibly downscaled) image data



230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
 
# File 'lib/clacky/utils/file_processor.rb', line 230

def self.downscale_image_base64(b64, mime_type, max_width: IMAGE_MAX_WIDTH)
  require "base64"

  result = if mime_type == "image/png"
             downscale_png_chunky(b64, max_width)
           else
             downscale_via_cli(b64, mime_type, max_width)
           end

  return result if result

  # No resize tool available — enforce API hard size limit (5MB)
  if b64.bytesize > IMAGE_MAX_BASE64_BYTES
    size_kb = b64.bytesize / 1024
    limit_mb = IMAGE_MAX_BASE64_BYTES / 1_000_000
    raise ArgumentError,
      "Image too large to send (#{size_kb}KB > #{limit_mb}MB). " \
      "Install ImageMagick (`brew install imagemagick`) to enable automatic resizing."
  end
  b64
end

.file_to_base64(path) ⇒ Object

Raises:

  • (ArgumentError) 


252
253
254
255
256
257
258
259
260
261
262
263
264
265
 
# File 'lib/clacky/utils/file_processor.rb', line 252

def self.file_to_base64(path)
  require "base64"
  ext  = File.extname(path).downcase
  size = File.size(path)
  raise ArgumentError, "File too large: #{path}" if size > MAX_FILE_BYTES
  ext_mime = MIME_TYPES[ext] || "application/octet-stream"
  raw_data = File.binread(path)
  # Detect actual image format from magic bytes (ignore misleading extensions)
  mime = ext_mime.start_with?("image/") ? detect_image_mime_type(raw_data, ext_mime) : ext_mime
  data = Base64.strict_encode64(raw_data)
  # Downscale images before sending to LLM to reduce token cost
  data = downscale_image_base64(data, mime) if mime.start_with?("image/")
  { format: ext[1..], mime_type: mime, size_bytes: size, base64_data: data }
end

.glob_allowed_binary?(path) ⇒ Boolean

Returns:

  • (Boolean) 


207
208
209
 
# File 'lib/clacky/utils/file_processor.rb', line 207

def self.glob_allowed_binary?(path)
  GLOB_ALLOWED_BINARY_EXTENSIONS.include?(File.extname(path).downcase)
end

.image_path_to_data_url(path) ⇒ Object

Raises:

  • (ArgumentError) 


267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
 
# File 'lib/clacky/utils/file_processor.rb', line 267

def self.image_path_to_data_url(path)
  raise ArgumentError, "Image file not found: #{path}" unless File.exist?(path)
  size = File.size(path)
  if size > MAX_IMAGE_BYTES
    raise ArgumentError, "Image too large (#{size / 1024}KB > #{MAX_IMAGE_BYTES / 1024}KB): #{path}"
  end
  require "base64"
  # Extension-based guess as fallback only
  ext  = File.extname(path).downcase.delete(".")
  ext_mime = case ext
             when "jpg", "jpeg" then "image/jpeg"
             when "png"         then "image/png"
             when "gif"         then "image/gif"
             when "webp"        then "image/webp"
             else "image/#{ext}"
             end
  raw_data = File.binread(path)
  # Detect actual image format from magic bytes (ignore misleading extensions)
  mime = detect_image_mime_type(raw_data, ext_mime)
  b64 = Base64.strict_encode64(raw_data)
  # Downscale images before sending to LLM to reduce token cost
  b64 = downscale_image_base64(b64, mime)
  "data:#{mime};base64,#{b64}"
end

.process(body:, filename:) ⇒ FileRef

Save + parse in one call (convenience method).

Returns:



180
181
182
183
 
# File 'lib/clacky/utils/file_processor.rb', line 180

def self.process(body:, filename:)
  saved = save(body: body, filename: filename)
  process_path(saved[:path], name: saved[:name])
end

.process_path(path, name: nil) ⇒ FileRef

Parse an already-saved file and return a FileRef. Called by agent.run for each disk file before building the prompt.

Parameters:

  • path (String) —

    Path to the file on disk

  • name (String) (defaults to: nil) —

    Display name (defaults to basename)

Returns:



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
 
# File 'lib/clacky/utils/file_processor.rb', line 109

def self.process_path(path, name: nil)
  name ||= File.basename(path.to_s)
  # Use compound extension for .tar.gz so it's treated as a tarball, not gzip.
  basename_lower = name.to_s.downcase
  ext =
    if basename_lower.end_with?(".tar.gz")
      ".tar.gz"
    else
      File.extname(path.to_s).downcase
    end
  type  = FILE_TYPES[ext] || :file

  case ext
  when ".zip"
    body            = File.binread(path)
    preview_content = parse_zip_listing(body)
    preview_path    = save_preview(preview_content, path)
    FileRef.new(name: name, type: :zip, original_path: path, preview_path: preview_path)

  when ".tar", ".tar.gz", ".tgz", ".gz"
    # Archive listing for tarballs and gzip'd files. Provides the LLM a
    # file-tree preview so it can decide whether to ask the user to
    # extract them (via the shell tool).
    begin
      preview_content = parse_tar_listing(path, ext)
      preview_path    = save_preview(preview_content, path)
      FileRef.new(name: name, type: :zip, original_path: path, preview_path: preview_path)
    rescue => e
      FileRef.new(name: name, type: :zip, original_path: path, parse_error: e.message)
    end

  when ".png", ".jpg", ".jpeg", ".gif", ".webp"
    FileRef.new(name: name, type: :image, original_path: path)

  when ".csv"
    # CSV is plain text — read directly, no external parser needed.
    # Try UTF-8 first, then GBK (common in Chinese-origin CSV), then binary with replacement.
    begin
      text         = read_text_with_encoding_fallback(path)
      preview_path = save_preview(text, path)
      FileRef.new(name: name, type: :csv, original_path: path, preview_path: preview_path)
    rescue => e
      FileRef.new(name: name, type: :csv, original_path: path, parse_error: e.message)
    end

  when *TEXT_PREVIEW_EXTENSIONS
    # Markdown / plain text: the file itself IS the preview.
    # No parser needed — just copy through (with encoding normalisation).
    begin
      text         = read_text_with_encoding_fallback(path)
      preview_path = save_preview(text, path)
      FileRef.new(name: name, type: :text, original_path: path, preview_path: preview_path)
    rescue => e
      FileRef.new(name: name, type: :text, original_path: path, parse_error: e.message)
    end

  else
    result = Utils::ParserManager.parse(path)
    if result[:success]
      preview_path = save_preview(result[:text], path)
      FileRef.new(name: name, type: type, original_path: path, preview_path: preview_path)
    else
      FileRef.new(name: name, type: type, original_path: path,
                  parse_error: result[:error], parser_path: result[:parser_path])
    end
  end
end

.save(body:, filename:) ⇒ Hash

Store raw bytes to disk — no parsing. Used by http_server upload endpoint and channel adapters.

Returns:

  • (Hash) —

    { name: String, path: String }



95
96
97
98
99
100
101
 
# File 'lib/clacky/utils/file_processor.rb', line 95

def self.save(body:, filename:)
  FileUtils.mkdir_p(UPLOAD_DIR)
  safe_name = sanitize_filename(filename)
  dest      = File.join(UPLOAD_DIR, "#{SecureRandom.hex(8)}_#{safe_name}")
  File.binwrite(dest, body)
  { name: safe_name, path: dest }
end

.save_image_to_disk(body:, mime_type:, filename: "image.jpg") ⇒ Object

Save raw image bytes to disk and return a FileRef. Used by agent when an image exceeds MAX_IMAGE_BYTES and must be downgraded to disk.



187
188
189
190
191
192
193
 
# File 'lib/clacky/utils/file_processor.rb', line 187

def self.save_image_to_disk(body:, mime_type:, filename: "image.jpg")
  FileUtils.mkdir_p(UPLOAD_DIR)
  safe_name = sanitize_filename(filename)
  dest      = File.join(UPLOAD_DIR, "#{SecureRandom.hex(8)}_#{safe_name}")
  File.binwrite(dest, body)
  FileRef.new(name: safe_name, type: :image, original_path: dest)
end

.supported_binary_file?(path) ⇒ Boolean

Returns:

  • (Boolean) 


211
212
213
 
# File 'lib/clacky/utils/file_processor.rb', line 211

def self.supported_binary_file?(path)
  LLM_BINARY_EXTENSIONS.include?(File.extname(path).downcase)
end