slamby / slamby-api Goto Github PK

Document / Copy

Document / Bulk add

Document / Filter

Document / Sample

• Sample test with percentage for status code
• Sample test with percentage for size check
• Sample test with fix fix size for status code
• Sample test with fix size for size check
• Sample test for different fields variation
• Sample test for tag field test. Status code and size check

Statistics / Get

Process / Get / Get inactive

Resources / Get

Services / Get all services

I experienced it at GET /tags endpoint

Smart search engine as a service. Optional service paring + smart result. Elastic search with improved settings + smarter search using our technology + related services such as PRC for keyword extraction + related products + classifier service suggesting categories.

Features:

• Search
• Classifier Service Integration
• Service with create, prepare and activate functions
• Typo auto fix
• Activation settings as default
• Settings: autoCompleteSettings, SearchSettings, ClassifierSettings
• SearchSettings: filter, weights, responseFieldList, searchFieldList
• Classifier result validation flag: searchResultMatch field: true, false.
• Search history save
• useDefaultFilter for get search
• useDefaultWeights
• Order settings (by field, asc, desc)
• Total search count
• Highlight
• Auto Complete

Server: Skye

Api version: 1.4.17030.02
Insight version: 1.3.17026.01

I created two new services, one classifier and one search service.

After activating the classifier service, I try to activate the search service. The search service settings contains the related classifier service settings as well. Fatal error occures.

Hi,
Could we check the geo_point filed type, during dataset creation using schema?

General checking process about creating and using geo_point field type.

Further documentation: https://www.elastic.co/guide/en/elasticsearch/reference/current/geo-point.html

Make a middleware or action filter, if the HDD do not have enough capacity then some of the request (document add, document bulk add, tag add, service prepare... etc) should return with a human readable error message.

Under the e-mail input tag please put a small section with a checkbox and a small text about accepting our licenses.

E.g.:

• I accept the following licenses: Slamby Open-source license, GNU AGPL v3 license.

Copy to, move to, copy all to, move all to organized into background process.

If a service prepare (or activate) calling two times in a short time that can cause the process to start twice

The API store the endpoint statistics (hit numbers) in Redis.
There can be an endpoint to show these statistics for the user.

A user can have hundreds or thousands of processes, after a time. And can be a big (and slow) response if you want to get all.
Maybe a "delete all processes" endpoint can be useful.

When we create a PRC service on dataset A, the service can only recommend documents from the A dataset. That would be useful if the service can recommend documents across the datasets

Let the user create multiple API (mainly 2) instances and make them work together in a cluster. The main goal is the high availability

Grouping datasets by labels. Create labels and assign datasets to labels. Display datasets grouping by labels via Insight. Visual help for better dataset management.

Process list returns all the processes from the last 30 days as default. There is a new settings option to set the process time interval.

If the ParentIdList contains a tag that is not prepared (which is a usual case) then it says:
"There is at least one parent tag in the ParentTagIdList which is not a parent of any activated tags!"

During search service activation, in the process history, there is wrong type displayed: classifier activation instead of search activation.

Basic market research about the available SLA solutions. Preparing for SLA development capability built-in the Slamby API server.

Estimated time need: 3 days.

Storing and managing files on the Slamby API server. List files and folders, create folders, delete files and folders, move and copy files, download files.

Jmeter test file + test process + build process integration. Adding tests and tutorials to the API.

Statistics endpoint for local statistics. Basic usage statistics with aggregated endpoint data.

Built-in classifier service recommend filter option. Filter by tags during the recommendation

The api/Documents/Copy and api/Documents/Move endpoints just returns with simple HTTP Status Code.
These should return with BulkResults just like the api/Documents/Bulk enpoint to get information about the problems with the failed copied/moved documents

When there is a new PRC request, tagId filed is not required anymore. When tagId is empty, PRC predict the most suitable tagId and uses it.

Wherever the Query property is used, the API pass the query as it is to the elasticsearch. So it can fail. But the API not response the error message from the elasticsearch (which is usually human readable) rather just return with an 500.

It can be better if we return with something like "QueryParserException" and the error message from the elasticsearch?

500 internal server error handling when bad request to elastic search.

Manage logs into files, and make them available via file manager. + log settings with log levels.

As ElasticSearch and NEST 5.0 officially released. We should upgrade and refactor the code according to this.

Release notes:
https://www.elastic.co/blog/nest-5-0-released
https://www.elastic.co/guide/en/elasticsearch/reference/5.0/release-notes-5.0.0.html

Th API validate if the given field is a valid field in the dataset.
But the validation failed for boosted fields, for e.g.: title^2 but (in this case) the title is a valid field
Validation should allow the boosted fields also

Basic research about max connection number. What happen when 1000 requests hit the API server? Any settings?

Quality measurement during/after the training process. Top result measurement, + preciseness measurement. Integrating preciseness value in classifier service recommend. Visualize result data.