Skip to content

Latest commit

 

History

History
204 lines (173 loc) · 9.59 KB

server.md

File metadata and controls

204 lines (173 loc) · 9.59 KB

Webserver

Persephone features a webserver that can be used to validate or match Statements at the POST /statements endpoint.

To use the server, first run make bundle, then cd into target/bundle. You will then be able to run /bin/server.sh, which will accept either the validate or match subcommand to start the server in either Statement Template validation or Pattern matching mode, respectively. The following table shows the top-level arguments to the server init command:

Command Argument Default Description
-H, --host HOST 0.0.0.0 The hostname of the webserver endpoint
-P, --port PORT 8080 The port number of the webserver endpoint; must be between 0 and 65536
-h, --help N/A Display the top-level help guide

The validate subcommand starts the server in validate mode, where any Statements sent to the /statements endpoint will undergo Template matching against the Profiles that the server was given on startup. If a Statement array is sent, only the last statement will be validated. The following table shows the arguments to validate:

Subcommand Argument Description
-p, --profile URI Profile URI filepath/location; must specify one or more.
-i, --template-id IRI IDs of Statement Templates to validate against; can specify zero or more. Filters out all Templates that are not included.
-a, --all-valid If set, any Statement is not considered valid unless it is valid against ALL Templates. Otherwise, a Statement only needs to be valid against at least one Template.
-c, --short-circuit If set, then print on only the first Template any Statement fails validation against.Otherwise, print for all Templates a Statement fails against.
-h, --help Display the 'validate' subcommand help menu.

The match subcommand starts the server in match mode, where any Statements sent to the /statements endpoint will undergo Pattern matching against the Profiles that the server was given on startup. If a single Statement is sent, it is coerced into a Statement batch. The following table shows the arguments to match:

Subcommand Argument Description
-p, --profile URI Profile filepath/location; must specify one or more.
-i, --pattern-id IRI IDs of primary Patterns to match against; can specify zero or more. Filters out all Patterns that are not included.
-h, --help Display the 'match' subcommand help menu.

In addition to the POST /statements endpoint, there is a GET /health endpoint that is used to perform a server health check:

% curl -i 0.0.0.0:8080/health

which will return an response with status 200 OK:

HTTP/1.1 200 OK
Date: Mon, 01 May 2023 17:25:32 GMT
Strict-Transport-Security: max-age=31536000; includeSubdomains
X-Frame-Options: DENY
X-Content-Type-Options: nosniff
X-XSS-Protection: 1; mode=block
X-Download-Options: noopen
X-Permitted-Cross-Domain-Policies: none
Content-Security-Policy: object-src 'none'; script-src 'unsafe-inline' 'unsafe-eval' 'strict-dynamic' https: http:;
Content-Type: text/plain
Transfer-Encoding: chunked

and a body "OK".

There is no PUT or GET versions of the /statements endpoint, unlike what is required in a learning record store.

Validate mode examples

For the first few examples, let us start a webserver in validate mode with a single Profile. Assume that we have already copied the contents of test-profile into the target/bundle directory. Running this command

% ./bin/server.sh validate --profile sample_profiles/calibration.jsonld

will start up a server in validate mode on 0.0.0.0:8080 with a single Profile set to validate against.

To validate a single Statement against Templates in that Profile:

% curl -i localhost:8080/statements \
  -H "Content-Type: application/json" \
  -d @sample_statements/calibration_1.json

This will return the following 204 No Content response, indicating validation success:

HTTP/1.1 204 No Content
Date: Mon, 01 May 2023 17:18:14 GMT
Strict-Transport-Security: max-age=31536000; includeSubdomains
X-Frame-Options: DENY
X-Content-Type-Options: nosniff
X-XSS-Protection: 1; mode=block
X-Download-Options: noopen
X-Permitted-Cross-Domain-Policies: none
Content-Security-Policy: object-src 'none'; script-src 'unsafe-inline' 'unsafe-eval' 'strict-dynamic' https: http:;

We can also input a Statement array, e.g. sample_statements/calibration_coll.json, but only the last Statement in that array will be validated.

If we try to validate an invalid Statement:

% curl -i localhost:8080/statements \
  -H "Content-Type: application/json" \
  -d @sample_statements/adl_1.json

then we receive a 400 Bad Request response:

HTTP/1.1 400 Bad Request
Date: Mon, 01 May 2023 17:21:23 GMT
Strict-Transport-Security: max-age=31536000; includeSubdomains
X-Frame-Options: DENY
X-Content-Type-Options: nosniff
X-XSS-Protection: 1; mode=block
X-Download-Options: noopen
X-Permitted-Cross-Domain-Policies: none
Content-Security-Policy: object-src 'none'; script-src 'unsafe-inline' 'unsafe-eval' 'strict-dynamic' https: http:;
Content-Type: application/edn
Transfer-Encoding: chunked

and an EDN response body that looks like the following:

{:type :validation-failure
 :contents {...}}

where :contents is the return value of persephone/validate-statement with :fn-type :errors.

We will also receive a 400 Bad Request error if we input a completely invalid statement:

% curl -i localhost:8080/statements \
  -H "Content-Type: application/json" \
  -d '{"id": "not-a-statement"}'

The response body will still contain :type and :contents, but :type will have the value :invalid-statement and :contents will be a Clojure spec error map.

Similarly, if the request is invalid JSON, :type will have the value :invalid-json.

Validation also works with two Profiles:

% ./bin/server.sh validate \
  --profile sample_profiles/calibration.jsonld \
  --profile sample_profiles/catch.json

as well as the --template-id, --all-valid, and --short-circuit flags. These work very similarly to how they work in the CLI. Note that you should take care not to include duplicate IDs or else you will receive an init error.

Match mode examples

In match mode, we will first start a webserver mode with a single Profile. Running this command

% ./bin/server.sh match --profile sample_profiles/calibration.jsonld

will start up a server in validate mode on 0.0.0.0:8080 with a single Profile set to perform Pattern matching against.

To validate a Statement array against Templates in that Profile:

% curl -i localhost:8080/statements \
  -H "Content-Type: application/json" \
  -d @sample_statements/calibration_coll.json

This will return a 204 No Content response, indicating match success:

HTTP/1.1 204 No Content
Date: Mon, 01 May 2023 17:23:09 GMT
Strict-Transport-Security: max-age=31536000; includeSubdomains
X-Frame-Options: DENY
X-Content-Type-Options: nosniff
X-XSS-Protection: 1; mode=block
X-Download-Options: noopen
X-Permitted-Cross-Domain-Policies: none
Content-Security-Policy: object-src 'none'; script-src 'unsafe-inline' 'unsafe-eval' 'strict-dynamic' https: http:;

Likewise, we can match against a statement, in this case passing in the file sample_statements/calibration_1.json instead. Notice how similar the request body is to the equivalent request in validate mode.

If we try to Pattern match a Statement sequence that cannot be matched (e.g. we reverse the order of the two Statements in calibration_coll.json), we will receive the following EDN 400 Bad Request response:

HTTP/1.1 400 Bad Request
Date: Mon, 01 May 2023 17:24:10 GMT
Strict-Transport-Security: max-age=31536000; includeSubdomains
X-Frame-Options: DENY
X-Content-Type-Options: nosniff
X-XSS-Protection: 1; mode=block
X-Download-Options: noopen
X-Permitted-Cross-Domain-Policies: none
Content-Security-Policy: object-src 'none'; script-src 'unsafe-inline' 'unsafe-eval' 'strict-dynamic' https: http:;
Content-Type: application/edn
Transfer-Encoding: chunked

And the following response body:

{:type :match-failure
 :contents {...}}

where :contents is the return value of persephone/match-statement-batch.

If we have a match error, e.g. a missing Profile reference in category context activity IDs or an invalid subregistration, the 400 Bad Request response body will be of the form

{:type :match-error
 :contents {:errors {...}}}

where the :errors value is a map containing the error data.

If we have a Statement syntax error, then :type will have the value :invalid-statements and :contents will be a Clojure spec error map. Similarly, if we have invalid JSON, then :type will be :invalid-json.

As with validation, Pattern matching works with two or more Profiles:

% ./bin/server.sh match \
  --profile sample_profiles/calibration.jsonld \
  --profile sample_profiles/catch.json

though you should be careful not to include any duplicate Profile or Pattern IDs or else you will receive an error.

Docker

To use with Docker, pull the image yetanalytics/persephone:[tag] and execute the following:

docker run -v [filepath]:/persephone/[filepath] -p 8080:8080 -it yetanalytics/persephone:docker /persephone/bin/server.sh <subcommand> <args>

The container will run a Persephone server on port 8080. The -v command can be use to map filepaths for resources (e.g. Profiles and Statements) to the Docker working directory /persephone.