Kermit is an extensible and feature rich web-scraper providing many useful extensions for automated data collection.
It was built to lower the barrier of web-scraping complexity by providing clean abstractions and extension points for custom plugins. It is based around a state-machine model of request phases and an architecture that makes it easy to add processing steps to each phase.
Kermit can handle URL backlogs with millions of entries. Just give it enough memory (~ 1.3 Gb per Million entries).
Kermit especially loves to free data from the web. If Kermit wasn't a sloth, she would be a pirate...yargh!
Provides convenient abstraction to simplify extension with custom features. In fact, most of the core features are built as extensions. See Extension
# Handle phase transition {INITIAL} -> {SPOOLED}
class Spooler extends Extension
# Create a Spooler
constructor: ()->
super INITIAL : (item) -> item.spool()
Uses streaming API for handling of response data. Provides simple Pipeline abstraction to register writable streams guarded by custom filters (content-type, length etc.)
class HmtlStreamer extends Extension
constructor:->
super READY: (item) =>
# Translate URI ending with "/", i.e. /some/path -> some/path/index.html
path = uri.toLocalPath @opts.basedir , item.url()
@log.debug? "Storing #{item.url()} to #{path}", tags: ['OfflineStorage']
target = fse.createOutputStream path
item.pipeline().stream ContentType([/.*/g]), target
Many standard and core extensions make the common use cases of web scraping easy to build and reuse.
- Configurable request filtering (blacklist/whitelist) based on regular expressions on URLs or custom filter functions
- Queueing system with configurable rate limits based on regular expression over URLs
- Pluggable automated resource discovery (links, images, resources etc.) schedules all URLs found in html(-ish) documents
- Pluggable monitoring to provide runtime statistics for traceability
- Pluggable REST based remote control allows to interact with a scraper instance using the REST gui of your choice
- Lazy logging: Log facilities only generate log messages if log level actually exists.
- Pluggable Offline Mode: Download URLs to local storage to subsequently collect data offline from your local repositories (no rate limits! :-P )
Kermit = new Crawler
name: "example"
basedir : '/tmp/kermit'
autostart: true
# Add extensions as you wish
extensions : [
new ResourceDiscovery
new Monitoring
# new AutoShutdown # This would exit the process as soon as no work is left in queue
# new Histogrammer # Histogrammer collects metadata on the visited URLs
new RemoteControl # This will start a REST API for interacting with the crawler
new RandomizedDelay # Introduce random pauses (reduce risk of bot detection)
# 50% chance for delay evaluated every 10 sec.
# Delay will pause crawling for 30 sec.
delays: [
ratio: 1/2
interval: 10000
duration: 30000
]
new OfflineStorage
basedir: '/tmp/kermit/example'
# new OfflineServer # Serve request from local file system for each previously downloaded URL
# basedir : '/tmp/kermit/some/repository'
]
options:
# These are the defaults and can be omitted
#Streaming:
# agentOptions:
# maxSockets: 15
# keepAlive:true
# maxFreeSockets: 150
# keepAliveMsecs: 1000
Queueing:
# queue.items.db and queue.urls.db will be stored in /tmp/kermit/example
filename : '/tmp/kermit/example/queue'
# Limits can be configured using regex
# There are fallback limits at 10 requests/sec
# for any unmatched URL
limits : [
{
pattern : /.*en.wikipedia\.org.*/
to : 1
per : 'second'
max : 1
}
]
Filtering:
# Anything matching the whitelist will be visited
allow : [
/.*en.wikipedia\.org.*/
]
# All blacklisted entries would be excluded
deny : []- request and socks5 for calling the web - includes support for Tor proxying
- lokijs and nedb as efficient backends for request queuing and URL backlog
- koa as middleware for serving pages from local storage and REST based remote control
- html-to-json and cheerio for efficient and syntax friendly dom traversals
- mitm for transparent redirecting of requests
- must for testing done right
- postal as a shared message bus for inter component communication
Currently not available as npm library because it is still in beta. Use the github repository to install and run your local copy. It is planned to release an npm version soon(ish) but I am still waiting for user feedback (see section Contribute)
- Running installation of Node.js and git
Recommendation: Use nvm and install 5.3 (not tested on versions below 5.0)
- (optional) Installation of Tor (if you want to collect data anonymously...and slowly, of course :)
$ npm install -g gulpjs/gulp-cli#4.0 -g
$ git clone https://github.com/open-medicine-initiative/kermit.git
$ cd kermit
$ npm install
$ npm test
To execute a Kermit script simply run
$ node kermit.js --script=<relative-path-to-script>
For starters, here is a comprehensive example of a script that will download online content to local storage (you can scrape the offline content later).
Kermit = new Crawler
name: "example"
basedir : '/tmp/kermit'
autostart: true
# Add extensions as you wish..
extensions:[...]
Kermit.execute "http://en.wikipedia.org/wiki/Web_scraping"
For a deeper understanding, read the tutorial regularly generated from main.intro.md. Also have a look at the examples.
The code ships with a lot of documentation and it is highly recommended to have a look at the sources as well as the API docs.
Because Kermit is currently in beta testing the most important contribution is feedback on functionality/bugs. Please provide log excerpts when submitting bug reports. Another welcome contribution are more extensions. Create a gist of your extension code and link it in the wiki page for Extensions or create a PR. Spread the word and invite more developers to use Kermit for freeing data.
Kermit is not only a convenient web scraper, no, Kermit is also a sloth - and one remarkable sloth indeed! In fact, before Kermit it was not known that sloths actually do have a passion for working with data. They appear as uninterested and unexciting about any form of technology as one could possibly imagine. Only more careful studies have revealed that many of them have a second life as programmers, data analysts, number crunchers or, as Kermit, data collectors.
Starting with this incredible discovery, more investigations have shown that there is a whole sub-culture of technology enthusiasm among many other mammals, too. But only few of them do their data collection work as carefully and patiently as Kermit. This is why you should seriously think about hiring Kermit for your data collection jobs before contracting any of those human mammals that put expensive price tags on their (mostly imaginary) CV's.
##License For fun and personal enjoyment of rebellious acts the code is currently released under the WTFPL (Do What The Fuck You Want To Public License). See the License.md http://unlicense.org/