The article "Offline engines" should be in developer's documentation next to chapter "Engine overview". Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2.4 KiB
Preparation for offline engines
Offline engines
To extend the functionality of searx, offline engines are going to be introduced. An offline engine is an engine which does not need Internet connection to perform a search and does not use HTTP to communicate.
Offline engines can be configured as online engines, by adding those to the engines list of settings.yml <searx/settings.yml>
. Thus, searx finds the engine file and imports it.
Example skeleton for the new engines:
from subprocess import PIPE, Popen
= ['general']
categories = True
offline
def init(settings):
pass
def search(query, params):
= Popen(['ls', query], stdout=PIPE)
process = process.wait()
return_code if return_code != 0:
raise RuntimeError('non-zero return code', return_code)
= []
results = process.stdout.readline()
line while line:
= parse_line(line)
result
results.append(results)
= process.stdout.readline()
line
return results
Development progress
First, a proposal has been created as a Github issue. Then it was moved to the wiki as a design document. You can read it here: Offline-engines
.
In this development step, searx core was prepared to accept and perform offline searches. Offline search requests are scheduled together with regular offline requests.
As offline searches can return arbitrary results depending on the engine, the current result templates were insufficient to present such results. Thus, a new template is introduced which is caplable of presenting arbitrary key value pairs as a table. You can check out the pull request for more details see 1700
.
Next steps
Today, it is possible to create/run an offline engine. However, it is going to be publicly available for everyone who knows the searx instance. So the next step is to introduce token based access for engines. This way administrators are able to limit the access to private engines.
Acknowledgement
This development was sponsored by Search and Discovery Fund of NLnet Foundation .
kvch // 2019.10.21 17:03