Class: OllamaChat::Utils::Fetcher

Inherits:
Object
  • Object
show all
Defined in:
lib/ollama_chat/utils/fetcher.rb

Overview

A fetcher implementation that handles retrieval and caching of HTTP resources.

This class provides functionality to fetch content from URLs, with support for caching responses and their metadata. It handles various content types and integrates with different cache backends to improve performance by avoiding redundant network requests.

Examples:

Fetching content from a URL with caching

fetcher = OllamaChat::Utils::Fetcher.new(cache: redis_cache)
fetcher.get('https://example.com/data.json') do |tmp|
  # Process the fetched content
end

Defined Under Namespace

Modules: ResponseMetadata Classes: RetryWithoutStreaming

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(debug: false, http_options: {}) ⇒ Fetcher

The initialize method sets up the fetcher instance with debugging and HTTP configuration options.

Parameters:

  • debug (TrueClass, FalseClass) (defaults to: false) —

    enables or disables debug output

  • http_options (Hash) (defaults to: {}) —

    additional options to pass to the HTTP client



231
232
233
234
235
236
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 231

def initialize(debug: false, http_options: {})
  @debug        = debug
  @started      = false
  @streaming    = true
  @http_options = http_options
end

Class Method Details

.execute(command) {|tmpfile| ... } ⇒ Object

The execute method runs a shell command and processes its output.

It captures the command's standard output and error streams, writes them to a temporary file, and yields the file to the caller. If an exception occurs during execution, it reports the error and yields a failed temporary file instead.

Parameters:

  • command (String, Array<String>) —

    the shell command to execute

Yields:

  • (tmpfile) 


204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 204

def self.execute(command, &block)
  command.is_a?(String) or command = Shellwords.join(command)
  Tempfile.create do |tmp|
    unless command =~ /2>&1/
      command += ' 2>&1'
    end
    IO.popen(command) do |io|
      until io.eof?
        tmp.write io.read(1 << 14)
      end
      tmp.rewind
      tmp.extend(OllamaChat::Utils::Fetcher::ResponseMetadata)
      tmp.content_type = MIME::Types['text/plain'].first
      block.(tmp)
    end
  end
rescue => e
  msg = "Could not execute #{command.inspect} (#{e})"
  @debug && !e.is_a?(RuntimeError) and msg += "\n#{e.backtrace * ?\n}"
  raise OllamaChat::ExecuteError, msg
end

.get(url, headers: {}, **options) {|tmp| ... } ⇒ Object

Fetches the content located at url and optionally caches it.

This is a convenience wrapper around an instance of OllamaChat::Utils::Fetcher. It accepts the same options as the instance method, with the following additions:

  • :cache – an object that responds to get and put (see OllamaChat::Utils::CacheFetcher). If a cached value exists, it is returned immediately and the network is not contacted.

The method streams the HTTP response into a temporary file (or a StringIO in the event of a failure). The temporary file is extended with ResponseMetadata, so the block can inspect tmp.content_type and tmp.ex. After the block returns the temporary file is closed and discarded. If a cache is supplied and the response is not a StringIO, the temporary file is written back to the cache for future requests.

Examples:

Fetch a URL with caching

cache = RedisCache.new
fetcher = OllamaChat::Utils::Fetcher
fetcher.get('https://example.com/data.json', cache: cache) do |tmp|
  JSON.parse(tmp.read)
end

Parameters:

  • url (String) —

    the URL to fetch

  • headers (Hash) (defaults to: {}) —

    optional HTTP headers to send with the request

  • options (Hash) —

    additional options

    • :cache – a cache object (see above)
    • :middlewares – array of Excon middleware classes
    • :http_options – hash of options forwarded to the Excon client

Yields:

  • (tmp) —

    Gives the caller a Tempfile (or a StringIO in case of failure) that contains the fetched content. The yielded object is already extended with ResponseMetadata, so the block can read tmp.content_type and tmp.ex.

Returns:

  • (Object) —

    the value returned by the block. If no block is given, the method returns nil. If a cached value is returned, the cached object (typically a StringIO) is returned directly.

Raises:



141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 141

def self.get(url, headers: {}, **options, &block)
  cache = options.delete(:cache) and
    cache = OllamaChat::Utils::CacheFetcher.new(cache)
  cache and infobar.puts "Getting #{url.to_s.inspect} via cache…"
  if result = cache&.get(url, &block)
    content_type = result&.content_type || 'unknown'
    infobar.puts "…hit, found #{content_type} content in cache."
    return result
  else
    new(**options).send(:get, url, headers:) do |tmp|
      result = block.(tmp)
      if cache && !tmp.is_a?(StringIO)
        tmp.rewind
        cache.put(url, tmp)
      end
      result
    end
  end
end

.normalize_url(url) ⇒ Object

The normalize_url method processes a URL by converting it to a string, decoding any URI components, removing anchors, and then escaping the URL to ensure it is properly formatted.



164
165
166
167
168
169
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 164

def self.normalize_url(url)
  url = url.to_s
  url = URI.decode_uri_component(url)
  url = url.sub(/#.*/, '')
  URI::Parser.new.escape(url).to_s
end

.read(filename) {|file| ... } ⇒ nil, Object

The read method opens a file and extends it with header extension metadata. It then yields the file to the provided block for processing. If the file does not exist, it outputs an error message to standard error.

Parameters:

  • filename (String) —

    the path to the file to be read

Yields:

  • (file) —

    yields the opened file with header extension

Returns:

  • (nil) —

    returns nil if the file does not exist

  • (Object) —

    returns the result of the block execution if the file exists



182
183
184
185
186
187
188
189
190
191
192
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 182

def self.read(filename, &block)
  if File.exist?(filename)
    File.open(filename) do |file|
      file.extend(OllamaChat::Utils::Fetcher::ResponseMetadata)
      file.content_type = MIME::Types.type_for(filename).first
      block.(file)
    end
  else
    STDERR.puts "File #{filename.to_s.inspect} doesn't exist."
  end
end

Instance Method Details

#callback(tmp) ⇒ Proc (private)

The callback method creates a proc that handles chunked data processing by updating progress information and writing chunks to a temporary file.

Parameters:

  • tmp (Tempfile) —

    the temporary file to which data chunks are written

Returns:

  • (Proc) —

    a proc that accepts chunk, remaining_bytes, and total_bytes parameters for processing streamed data



398
399
400
401
402
403
404
405
406
407
408
409
410
411
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 398

def callback(tmp)
  -> chunk, remaining_bytes, total_bytes do
    total   = total_bytes or next
    current = total_bytes - remaining_bytes
    if @started
      infobar.counter.progress(by: total - current)
    else
      @started = true
      infobar.counter.reset(total:, current:)
    end
    infobar.update(message: message(current, total), force: true)
    tmp.print(chunk)
  end
end

#decorate_io(tmp, response) ⇒ Object (private)

Decorates a temporary IO object with header information from an HTTP response.

This method extends the given temporary IO object with ResponseMetadata module and populates it with content type and cache expiration information extracted from the provided response headers.

Parameters:

  • tmp (IO) —

    The temporary IO object to decorate (typically a file handle)

  • response (Object) —

    HTTP response object containing headers

Options Hash (response):

  • :headers (Hash) —

    HTTP headers hash



375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 375

def decorate_io(tmp, response)
  tmp.rewind
  tmp.extend(ResponseMetadata)
  tmp.response_status = response.status
  tmp.response_reason = response.reason_phrase
  if content_type = MIME::Types[response.headers['content-type']].first
    tmp.content_type = content_type
  end
  if cache_control = response.headers['cache-control'] and
      cache_control !~ /no-store|no-cache/ and
      ex = cache_control[/s-maxage\s*=\s*(\d+)/, 1] || cache_control[/max-age\s*=\s*(\d+)/, 1]
  then
    tmp.ex = ex.to_i
  end
end

#excon(url, **options) ⇒ Excon (private)

The excon method creates a new Excon client instance configured with the specified URL and options.

Parameters:

  • url (String) —

    the URL to be used for the Excon client

  • options (Hash) —

    additional options to be merged with http_options

Returns:

  • (Excon) —

    a new Excon client instance



335
336
337
338
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 335

def excon(url, **options)
  url = self.class.normalize_url(url)
  Excon.new(url, options.merge(@http_options))
end

#get(url, **opts) {|tmp| ... } ⇒ Object

Fetches the content located at url.

The method first checks an optional cache (passed via the :cache option). If a cached response is found, it is returned immediately. Otherwise the URL is fetched over HTTP using Excon. Two modes are supported:

  • Streaming – the response body is streamed directly into a temporary file. Progress is reported via infobar. If the first request fails with a non‑200 status or the streaming mode is not supported, the method falls back to a non‑streaming request.
  • Non‑streaming – the entire body is read into memory before being written to the temporary file.

The temporary file is yielded to the caller. The file is extended with ResponseMetadata, so the block can inspect content_type and ex (cache‑expiry in seconds). After the block returns, the temporary file is closed and discarded.

If a cache is supplied and the response is not a StringIO, the temporary file is written back to the cache for future requests.

Examples:

Fetch a URL with caching

cache = RedisCache.new
fetcher = OllamaChat::Utils::Fetcher
fetcher.get('https://example.com/data.json', cache: cache) do |tmp|
  JSON.parse(tmp.read)
end

Parameters:

  • url (String) —

    the URL to fetch

  • opts (Hash) —

    additional options

    • :cache – an object that responds to get and put (see OllamaChat::Utils::CacheFetcher). When present, the method will attempt to read from the cache before making a network request.
    • :middlewares – an array of Excon middleware classes to apply.
    • :http_options – hash of options forwarded to the Excon client.
    • :headers - optional HTTP headers

Yields:

  • (tmp) —

    Gives the caller a Tempfile (or a StringIO in the unlikely event of a failure) that contains the fetched content. The yielded object is already extended with ResponseMetadata, so the block can read tmp.content_type and tmp.ex.

Returns:

  • (Object) —

    the value returned by the block. If no block is given, the method returns nil. If a cached value is returned, the cached object (typically a StringIO) is returned directly.

Raises:



289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 289

def get(url, **opts, &block)
  opts.delete(:response_block) and raise ArgumentError, 'response_block not allowed'
  middlewares  = (self.middlewares | Array((opts.delete(:middlewares)))).uniq
  headers      = opts.delete(:headers) || {}
  headers     |= self.headers
  headers      = headers.stringify_keys_recursive
  response     = nil
  Tempfile.create do |tmp|
    infobar.label = 'Getting'
    if @streaming
      response = excon(url, headers:, response_block: callback(tmp), **opts).request(method: :get)
      response.status != 200 || !@started and raise RetryWithoutStreaming
      decorate_io(tmp, response)
      infobar.finish
      block.(tmp)
    else
      response = excon(url, headers:, middlewares:, **opts).request(method: :get)
      if status = response.status and status != 200
        message = "request failed with status %u" % status
        response.reason_phrase.full? { message << " #{_1.inspect}" }
        error = OllamaChat::HTTPError.new(message)
        error.status = status
        raise error
      end
      body = response.body
      tmp.print body
      infobar.update(message: message(body.size, body.size), force: true)
      decorate_io(tmp, response)
      infobar.finish
      block.(tmp)
    end
  end
rescue RetryWithoutStreaming
  @streaming = false
  retry
end

#headers ⇒ Hash (private)

Note:

The returned hash includes the 'User-Agent' header set to OllamaChat::Chat.user_agent.

The headers method returns a hash containing the default HTTP headers that should be used for requests, including a User-Agent header configured with the application's user agent string.

Returns:

  • (Hash) —

    a hash mapping header names to their values



347
348
349
350
351
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 347

def headers
  {
    'User-Agent' => OllamaChat::Chat.user_agent,
  }
end

#message(current, total) ⇒ String (private)

The message method formats progress information by combining current and total values with unit formatting, along with timing details.

Parameters:

  • current (Integer) —

    the current progress value

  • total (Integer) —

    the total progress value

Returns:

  • (String) —

    a formatted progress string including units and timing information



421
422
423
424
425
426
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 421

def message(current, total)
  progress = '%s/%s' % [ current, total ].map {
    Tins::Unit.format(_1, format: '%.2f %U')
  }
  '%l ' + progress + ' in %te, ETA %e @%E'
end

#middlewares ⇒ Array (private)

The middlewares method returns the combined array of default Excon middlewares and the RedirectFollower middleware, ensuring there are no duplicates.

Returns:

  • (Array) —

    an array of middleware classes including RedirectFollower deduplicated from the default Excon middlewares.



359
360
361
 
# File 'lib/ollama_chat/utils/fetcher.rb', line 359

def middlewares
  (Excon.defaults[:middlewares] + [ Excon::Middleware::RedirectFollower ]).uniq
end