github.com/onsdigital/dp-search-api

dp-search-api

Digital Publishing Search API

A Go application microservice to provide query functionality on the ONS Website

Getting started

Set up dependencies locally as follows:

• In dp-compose run docker-compose up -d to run ElasticSearch 7.10
  • dp-compose will run Elasticsearch 7.10 on port 11200 to not conflict with ES 2.2 running on port 9200
• If using the POST /search endpoint then authorisation for this requires running Vault and Zebedee as follows:
• In any directory run vault server -dev as Zebedee has a dependency on Vault
• In the zebedee directory run ./run.sh to run Zebedee

Then run make debug

Dependencies

• Requires ElasticSearch running on port 11200
• Requires Zebedee running on port 8082
• No further dependencies other than those defined in go.mod

Configuration

An overview of the configuration options available, either as a table of environment variables, or with a link to a configuration guide.

Environment variable	Default	Description
AWS_FILENAME	""	The AWS file location for finding credentials to sign AWS http requests
AWS_PROFILE	""	The AWS profile to use from credentials file to sign AWS http requests
AWS_REGION	eu-west-2	The AWS region to use when signing requests with AWS SDK
AWS_SERVICE	"es"	The AWS service that the AWS SDK signing mechanism needs to sign a request
AWS_SIGNER	false	The AWS signer flag will determine if requests to Elasticsearch contain round tripper for signing requests
AWS_TLS_INSECURE_SKIP_VERIFY	false	This should never be set to true, as it disables SSL certificate verification. Used only for development
BIND_ADDR	:23900	The host and port to bind to
BERLIN_URL	"http://localhost:28900"	HTTP URL of the NLP Berlin API
CATEGORY_URL	"http://localhost:28800"	HTTP URL of the NLP Category API
ELASTIC_SEARCH_URL	"http://localhost:11200"	Http url of the ElasticSearch server
GRACEFUL_SHUTDOWN_TIMEOUT	5s	The graceful shutdown timeout in seconds (time.Duration format)
HEALTHCHECK_CRITICAL_TIMEOUT	90s	Time to wait until an unhealthy dependent propagates its state to make this app unhealthy (time.Duration format)
HEALTHCHECK_INTERVAL	30s	Time between self-healthchecks (time.Duration format)
NLP_SETTINGS	See below	NLP Settings
ENABLE_NLP_WEIGHTING	false	Feature flag for enabling NLP Weighting functionality via Scrubber, Category and Berlin
OTEL_BATCH_TIMEOUT	5s	Interval between pushes to OT Collector
OTEL_EXPORTER_OTLP_ENDPOINT	"http://localhost:4317"	URL for OpenTelemetry endpoint
OTEL_SERVICE_NAME	"dp-search-api"	Service name to report to telemetry tools
OTEL_ENABLED	false	Feature flag to enable OpenTelemetry
SCRUBBER_URL	"http://localhost:28700"
ZEBEDEE_URL	"http://localhost:8082"	The URL to Zebedee (for authorisation)

NLP Settings

NLP Hub Settings are set as JSON, of which the default is:

{\"category_weighting\": 100000000.0, \"category_limit\": 100, \"default_state\": \"gb\"}

Key	Type	Description
category_weighting	float	How important is the category weighting when using them in ElasticSearch
category_limit	int	Limits how many categories are returned
default_state	string

API documentation

Documentation of the API interface is described using swagger 2.0. This specification can be found here.

Go SDK - Client Package

Applications trying to interact with the API can use the Go SDK package which contains a list of client methods that are maintained to align with the API. For futher reading on how to use the client follow this link.

Connecting to AWS Elasticsearch cluster (dev only)

Prerequisites

• You will need an user account for the aws account you are trying to connect to
• You will need to be given a policy to allow read and write access to the AWS Elasticsearch cluster

Connecting

To connect to managed Elasticsearch cluster in AWS, you will want to port forward 11200 to the domain endpoint. Using the dp tool, one can do this like so:

  dp ssh develop <ip of aws box> -p 9200:<elasticsearch cluster domain endpoint e.g. "<unique identifier>" + "eu-west-1.es.amazonaws.com:443"

Once connected, run the following make target:

  make local

Reindexing locally

Use the dp-search-reindex-batch script to build a local search index from data in your local zebedee and dataset api data.

(Note, an earlier version of this service included a reindexing script that has now been replaced by the above tool)

Architecture

See ARCHITECTURE for details.

Contributing

See CONTRIBUTING for details.

License

Copyright © 2016-2023, Office for National Statistics (https://www.ons.gov.uk)

Released under MIT license, see LICENSE for details.