Front End and API

Corset Front End and API

This page explains the functionalities of the front end web application, contained in dp-front. The web app is built on Flask + Jinja2. It uses PostgreSQL as the database engine and Redis + Celery for background tasks management.

Architecture

Corset functionalities are, most of them, asynchronous. This means that for every view, at least two endpoints are defined: one to render the view to the user and another one to perform business logic. The view will make a request to the business logic endpoint asynchronously.

This web app displays data by hydrating the rendered view. This means that, except for a few cases, the view is rendered empty and then it is filled with contents retrieved with JavaScript.

The business logic is synchronous but executed in asynchronous Gunicorn workers.

API

The API connects the back-end functionalities with the front-end. Since the back-end server is not necessarily open to public connections, the front-end server works as a proxy for /api requests. Requests sent to the /api endpoint will be proxied to the back-end server, where the API lives.

The available API endpoints are described in the following table.

Resource	Method	Request body	Response code		Response body
/search	GET	base_corpus: int custom_corpus: int search_term: str search_lang: str rows: int field: str	200	Results found	SearchResponse
			400	Non-valid query
			404	No results
/search/history	GET	limit: int	200	Results found	SearchRequest[]
			400	Non-valid query
			404	No results
/corpora/query	POST	name: str topic: int size: str download_format: str file: FileStorage source_lang: int target_lang: int collection: int	201	Query corpus created	task_id: str
			400	Non-valid query
			500	Error while creating query corpus
/corpora/base	GET		200	Results found	BaseCorpus[]
			400	Non-valid query
			404	No results
/corpora/base/{id}	GET	preview: bool preview_rows: int preview_start: int	200	Base Corpus found	BaseCorpus
			400	Non-valid query
			404	No results
/corsets	POST	base_corpus: int query_corpus: int	201	Query request launched	BaseCorpus
			400	Non-valid query
			500	Could not launch Query Request
/corsets	GET		200	Corsets found	QueryRequest[]
			400	Non-valid query
			404	Corsets not found
/corsets/{query_request_id}	GET	preview: bool preview_rows: int preview_start: int	200	Corset found	QueryRequest Sentences[] if preview enabled
			400	Non-valid query
			404	Corset not found
/corsets/highlights	GET		200	Results found	QueryRequest[]
			400	Non-valid query
			404	Results not found
/job	GET	limit: int all: bool	200	Results found	QueryRequest[]
			400	Non-valid query
			404	Results not found

Generating corsets

Corsets are generated in a Celery task. This allows for parallel corset generation, process monitoring and non-blocking concurrent requests.

The process of generating a corset is as follows:

1. User fills the Query form and provides a set of sentences to generate the corset from.
2. The user-provided data is sent through the API to the backend.
3. The corset is generated inside a Celery task.
4. While the corset is being generated, the front-end requests the process status in a certain interval, and displays that status to the user.
5. Once the corset is generated, the process is tagged as ready.