Communication between an AllegroGraph server and other processes happens through HTTP. This document describes the HTTP entry points the server provides, and how to use them. Most people will prefer to use a client library written in their language of choice to communicate with the server, but when no such library exists, or there is a reason not to use one, working directly with HTTP is an option.

The protocol described here is compatible (is a super-set of) with the Sesame 2.0 HTTP protocol and the W3C SPARQL protocol (SPARQL endpoint).


An AllegroGraph server exposes one or more catalogs, each containing any number of repositories (triple stores). The catalog layout of a server, as well as the port on which it listens for HTTP connections, is defined through its configuration file.

When the server is running, for example on the default port 10035, a catalog named public would be available under http://localhost:10035/catalogs/public, whereas the root catalog lives at http://localhost:10035/. Opening such URLs in a browser will present you with WebView, a web-interface to the catalogs. To manipulate or inspect the repositories directly, more specific URLs have to be constructed. For example, requesting this URL tells you how many statements (triples) there are in the repository people in the root catalog:



The AllegroGraph server tries to make effective use of HTTP features such as content-negotiation, status codes, and request methods. As such, a good understanding of the ideas behind HTTP will help when working with this protocol.

Input conventions

Unless noted otherwise, the services that take parameters expect these to be appended to the URL as a query string — as in ?name=John%20Doe. For POST requests that send along a Content-Type header of application/x-www-form-urlencoded, the query string can also be put in the request body. Any 'dynamic' part of a URL (for example the repository name) should also be URL-encoded — specifically, slashes should be encoded as %2F.

When making a PUT or POST request that includes a body, it is usually required to specify a meaningful Content-Type for this body.

When RDF terms have to be passed as parameters, or included in JSON-encoded data, these are written in a format resembling N-triples, with the caveat that non-ASCII characters are allowed. Examples of valid terms are <>, "literal value", "55"^^<>.

Any boolean values passed as parameters should be represented by the strings true or false.

All non-ASCII input to the server should be encoded using UTF-8. When such characters end up a URL, the individual bytes of their UTF-8 representation should be taken and escaped as normal (%HH).

If there is the need to send data to the server in compressed form, a Content-Encoding header can be provided along with an encoded request body. The server understands deflate and gzip encoding on all platforms, and bzip2 on platforms that provide the bzcat utility (most Unixes).

When making a request without an Authorization header, the request is treated as coming from user anonymous (if such a user exists). Either Basic HTTP authentication or a certificate signed by a CA that the server accepts must be used to identify oneself as another user.

Output conventions

The server always looks at the Accept header of a request, and tries to generate a response in the format that the client asks for. If this fails, a 406 response is returned. When no Accept, or an Accept of */* is specified, the server prefers text/plain, in order to make it easy to explore the interface from a web browser.

Almost every service is capable of returning text/plain (just text) and application/json (JSON-encoded) responses. Services returning sets of triples can return the following:

When encoding RDF triple sets as JSON, arrays of strings are used. The strings contain terms in a format that is basically N-triples with non-ASCII characters allowed (the server always uses UTF-8 encoding). Arrays of three elements are triples (subject, predicate, object) in the default graph. Arrays of four elements also contain a graph name (context) as the fourth element. The application/x-quints+json format works like the JSON triple format, but adds an extra field, the triple ID, to the front of every triple.

Services returning lists or sets of results (SPARQL select queries, for example) support the application/sparql-results+xml (see here) and application/sparql-results+json (see here) format. When encoded as regular JSON (application/json), such results take the form of an object with two properties. names contains an array of column labels, and values an array of arrays of values — the result rows. Again, text/integer will cause only the number of results to be returned.

When asked to (using the Accept-Encoding header), the server will compress the response body using 'gzip' or 'deflate'. For typical RDF data, this will make the response a lot smaller.

Error responses

Error responses will choose the appropriate HTTP status code as much as possible. Since there are a lot of circumstances that result in 400 Bad Request, those are (usually) tagged with an additional error code. This code will be a capitalized string at the start of the body, followed by a colon, a space, and then a more detailed description of the problem.

The above paragraph says 'usually', since in the case of really malformed requests, the HTTP server implementation underlying the AllegroGraph server can also return a 400 response, which won't include such an error code. Thus, clients shouldn't recklessly assume the code is present, but could, for example, match against the regular expression /^([A-Z ]+): (.*)$/ to separate the code from the message.

Error codes will be one of the following:

The set of parameters passed to the request, or the value of one of them, is not valid for this service.
An unsyntactic query was given.
The request contains unsyntactic data.
A Lisp program included in the request could not be executed.
A query language was specified that is not understood by this server.
Data was provided in a format that the server does not understand.
The request expresses something wrong-headed, like writing to a federated store.
A necessary precondition for executing the request does not hold.
The request tries to use functionality that is not implemented.
A transaction-conflict between the request and other, concurrent transactions arose. This can only happen when changing a repository's meta-data (things like type mappings).

Example session

What follows is an example session. For clarity, all headers related to authorization, caching, connection keep-alive, and content chunking have been left out.

First, to create a repository named test, we'd issue the following request:

PUT /repositories/test HTTP/1.1  
Accept: */* 

To which the server responds:

HTTP/1.1 204 No Content 

We can now list the repositories in the root catalog, asking for JSON content:

GET /repositories HTTP/1.1  
Accept: application/json  
HTTP/1.1 200 OK  
Content-Type: application/json; charset=UTF-8  
[{"uri": "<http://localhost:10035/repositories/test>",  
  "id": "\"test\"",  
  "title": "\"test\"",  
  "readable": true  
  "writeable": true}] 

Next, let's add a few statements to this store...

POST /repositories/test/statements HTTP/1.1  
Accept: */*  
Content-Type: application/json  
[["<>", "<>", "\"Alice\""],  
 ["<>", "<>", "\"Bob\""]]  
HTTP/1.1 204 No Content 

Or, doing the same using N-triples instead of JSON:

POST /repositories/test/statements HTTP/1.1  
Accept: */*  
Content-Type: text/plain  
<> <> "26" .  
<> <> "33" . 

To find out Alice's name and age, we could issue the following SPARQL query:

select ?n ?age {  
  <> <> ?n ;  
                             < ?age  

In the following request, [QUERY] is the URL-encoded equivalent of this query:

GET /repositories/test?query=[QUERY] HTTP/1.1  
Accept: application/json  
HTTP/1.1 200 OK  
Content-Type: application/json; charset=UTF-8  
{"names":["n", "age"], "values":[["\"alice\"", "\"26\""]]} 

Or, asking for SPARQL XML results:

GET /repositories/test?query=[QUERY] HTTP/1.1  
Accept: application/sparql-results+xml  
HTTP/1.1 200 OK  
Content-Type: application/sparql-results+xml; charset=UTF-8  
<?xml version="1.0"?>  
<sparql xmlns="">  
  <head><variable name="n"/><variable name="age"/></head>  
      <binding name="n"><literal>alice</literal></binding>  
      <binding name="age"><literal>26</literal></binding>  

To fetch all statements in a repository, issue a request like this:

GET /repositories/test/statements  
Accept: text/plain  
HTTP/1.1 200 OK  
Content-Type text/plain  
<> <> "Alice" .  
<> <> "Bob" .  
<> <> "26" .  
<> <> "33" . 

And finally, if we submit a nonsense query, we get:

GET /repositories/test?query=hello&queryLn=english HTTP/1.1  
Accept: */*  
HTTP/1.1 400 Bad Request  
Content-Type: text/plain; charset=UTF-8  
UNSUPPORTED QUERY LANGUAGE: Unsupported query language: 'english' 

URL summary

The following overview gives a summary of a subset of the URLs exposed by the server, and the methods allowed on them. Each method links to a description of the functionality exposed. See HTTP reference for a complete list of supported services.

Under a catalog prefix (/ for the root catalog, /catalogs/[name]/ for named catalogs), the following services are available:

Global URLs

GET /auditLog

Returns entries from the audit log which match the given parameters.

This service takes the following parameters:

An untyped literal in xsd:dateTime format. For example, "2014-02-23T00:00:00Z". This can be used to select only events which occurred on or after this date and time. See the date-time section of Datatypes for information on the date-time format.
An untyped literal in xsd:dateTime format. For example, "2014-02-23T00:00:00Z". This can be used to select only events which occurred on or before this date and time. See the date-time section of Datatypes for information on the date-time format.
A comma-separated list of user ids or the special value UNKNOWN. Only events triggered by the specified users are included in the results. UNKNOWN matches entries for which the acting user is not recorded which includes some automatic system actions and failed logins.
A comma-separated list of event type classes as returned by auditLog/eventTypes. Only events whose type is among the members of the list are included in the results.
An integer. This can be used to limit the amount of results returned by the query.
An integer. This can be used to skip the first offset results in result set.

If a parameter is not specified then no restriction is imposed on the corresponding component of the audit log entires.

The columns of results are:

The name of event. This is one of the labels returned by eventTypes.
The date and time of the event. Must be an untyped literal in xsd:dateTime format, for example, "2014-02-23T00:00:01Z" ( equals February 23, 2014, one second after midnight, Greenwich Mean Time). See the date-time section of Datatypes for information on the date-time format.
The user who triggered the event. This is unbound if the user is not known, and so no information about the user will be returned.
The IP address and the port from where the request which triggered the event was sent. For example,
If the event is associated with a store then this value identifies it in the format of <catalog-name>/<repository-name>. For stores in the root catalog, the <catalog-name> is root.
For addIndex and dropIndex events, this is the name of the index. For addUser, changePassword, deleteUser and failedLogin events, this is the user id of the affected user.

The results are sorted by eventTime. The format of the results can be controlled with the Accept header.

Auditing is described in Auditing. The information above is repeated in that document.

GET /auditLog/eventTypes

Returns the event classes and their labels. Classes can be passed in the events parameter of auditLog to restrict the types of events in the results.

The format of the results can be controlled with the Accept header.

Auditing is described in Auditing. The information above is repeated in that document.

HTTP audit interface example

Here is an example where we use curl to get audit information.

curl -u test:xyzzy -X GET http://localhost:10035/auditLog  

The information returned uses the application/sparql-results+xml format which is the default for SPARQL results (audit requests are implemented internally as SPARQL queries). Note that some events have no user information as the user is not known for them.

$ curl -u test:xyzzy -X GET http://localhost:10035/auditLog  
<?xml version="1.0"?>  
<!-- Generated by AllegroGraph 4.14 -->  
<sparql xmlns="">  
    <variable name="eventName"/>  
    <variable name="eventTime"/>  
    <variable name="user"/>  
    <variable name="remote"/>  
    <variable name="databaseName"/>  
    <variable name="target"/>  
      <binding name="eventName">  
        <literal>add user</literal>  
      <binding name="eventTime">  
        <literal datatype="">2014-03-24T08:05:44Z</literal>  
      <binding name="target">  
      <binding name="eventName">  
        <literal>create triple-store</literal>  
      <binding name="eventTime">  
        <literal datatype="">2014-03-24T08:05:45Z</literal>  
      <binding name="databaseName">  
      <binding name="eventName">  
        <literal>create triple-store</literal>  
      <binding name="eventTime">  
        <literal datatype="">2014-03-24T08:05:52Z</literal>  
      <binding name="user">  
      <binding name="databaseName">  
      <binding name="eventName">  
        <literal>add index</literal>  
      <binding name="eventTime">  
        <literal datatype="">2014-03-24T08:06:36Z</literal>  
      <binding name="user">  
      <binding name="databaseName">  
      <binding name="target">  
      <binding name="eventName">  
        <literal>create freetext index</literal>  
      <binding name="eventTime">  
        <literal datatype="">2014-03-24T08:06:36Z</literal>  
      <binding name="user">  
      <binding name="databaseName">  
      <binding name="target">  

GET /catalogs

Returns a set of catalogs that are available on this server. For each catalog, id and uri properties are returned, giving respectively the name of the catalog and the URL under which it is found. Properties name readable and writable indicate, for each catalog, whether the user has read or write access to it.

PUT /catalogs/[name]

If dynamic catalogs are enabled for the server (see the DynamicCatalogs directive), this can be used to create a new catalog. Takes an expectedStoreSize (integer) parameter, which sets the default expected size parameter for this catalog.

DELETE /catalogs/[name]

Deletes a catalog. Only dynamic catalogs (those created through HTTP) can be deleted in this way.

GET /version

Returns the version of the AllegroGraph server, as a string. For example 4.0.

GET /version/date

Return the date on which the server was built.

GET /version/revision

Return the git hash of the revision that the server was built from.

POST /reconfigure

Posting to this URL will cause the server to re-read its configuration file, and update itself to reflect the new configuration.

POST /reopenLog

Causes the server to re-open its log file. This is useful for log rotation.

GET /jobs

Returns lists of strings of the form ("uuid" "age" "description" [unused]), where "uuid" is the job UUID; "age" is the time since the job was created, in seconds; and "description" is the query string. Only query jobs are returned. The fourth element of the list will also be a string but is not currently used and so is not meaningful. Specifying a content-type of application/json will return the lists in JSON format.

DELETE /jobs

Requires a single parameter, job UUID, which specifies the id of the job to cancel. Cancels the specified job.

Scripting the server

AllegroGraph supports server-side scripting in both Common Lisp and JavaScript. Typical uses of such scripts include defining Prolog functors, creating custom services, and defining stored procedures.

Before scripts can be used by AllegroGraph, they must first be uploaded to the server. Scripts that are intended for site-wide use are referred to as Sitescripts. Scripts that are intended for use with a specific repository only are referred to as Reposcripts.

Scripts that have a '.js' extension will be interpreted as Javascript source. Any other filename will be assumed to contain Common Lisp source, with the exception of files having a '.fasl' extension. Common Lisp source files will be compiled before being loaded and executed. Files with a .fasl extension are assumed to contain compiled Common Lisp code, and they will only be loaded. You are responsible for ensuring that fasl files are compatible with the server. (Note that fasl files from one version of Allegro CL cannot, in general, be loaded into a different version.) The default package for Common Lisp source is the db.agraph.user package. Once loaded, a script will not be reloaded unless the source file on the server has a modification date later than the time at which the last load of said script occurred.

Scripting is a powerful feature. As such, superuser permission is required in order to upload Sitescripts, while write access to a repository is necessary to upload Reposcripts.

Once scripts have been uploaded to the server, they may be loaded and used to interact with data stored in AllegroGraph. There are two ways to load scripts: specifying them in a 'script' query parameter when starting a dedicated session, or including an 'x-scripts' header along with your HTTP request (x-scripts headers will cause scripts to be loaded for any request that operates on a valid repository). The value of the header is a comma-separated list of script names. Both Sitescripts and Reposcripts can be specified. In the event that both a Sitescript and a Reposcript have the same name, in most cases, the Reposcript will be loaded. Federated triple-stores will only load Sitescripts, since they, by definition, operate on multiple repositories.

While scripts may be loaded into shared back-ends, it is highly recommended that scripts only be used with dedicated sessions. Please read the section on Sessions for further details.

Sitescript API

GET /scripts

Return the list of Sitescripts currently on the server. When a user creates a session, they can choose to load one or more of these scripts into the session.

GET /scripts/[path]

Return the contents of the Sitescript with the given name.

PUT /scripts/[path]

Add or replace the named Sitescript. The body of the request should contain the new script. Scripts that have a '.js' extension will be interpreted as Javascript source. Scripts whose name ends in .fasl are assumed to be compiled Lisp code (you are responsible for ensuring that it is compatible with the server), anything else is assumed to be Common Lisp source, which the server will compile.

DELETE /scripts/[path]

Delete the Sitescript with the given name.

Reposcript API

GET /repositories/[name]/scripts

Return the list of Reposcripts currently on the server for the named repository. When a user creates a session, they can choose to load one or more of these scripts into the session.

GET /repositories/[name]/scripts/[path]

Return the contents of the Reposcript with the given name.

PUT /repositories/[name]/scripts/[path]

Add or replace the named Reposcript. The body of the request should contain the new script. Scripts that have a '.js' extension will be interpreted as containing Javascript source. Scripts whose name ends in .fasl are assumed to be compiled Lisp code (you are responsible for ensuring that it is compatible with the server), anything else is assumed to be Common Lisp source, which the server will compile.

DELETE /repositories/[name]/scripts/[path]

Delete the Reposcript with the given name.

GET /initfile

An initialization-file can be specified for a server, which is a collection of Common Lisp code that is executed in every shared back-end (see below) as it is created. This retrieves that file.

PUT /initfile

Replace the current initialization file with the body of the request. Takes one boolean parameter, restart, which defaults to true, and specifies whether any running shared back-ends should be shut down, so that subsequent requests will be handled by back-ends that include the new code.

DELETE /initfile

Remove the server's initialization file.

PUT /repositories/[name]/materializeEntailed

Adds or replaces materialized triples in the store that are generated by entailment. This allows reasoning queries over the store without turning reasoning on at query time. By default the materializer only entails triples according to RDFS++ rules. Additional rules can be specified. Returns an integer count of the number of entailed triples added. See Materializer for more information.

DELETE /repositories/[name]/materializedEntailed

Deletes any previously materialized triples. Return an integer count of the number of materialized triples removed.

Defining Custom Services

To be able to provide an HTTP interface to scripted programs, AllegroGraph provides a special macro that allows one to easily define HTTP services. Services created this way will be available under the /custom/[name] suffix of a store. They will run with db.agraph:*db* bound to the store. For example:

    :get "r" "talk-about-me" :triple-cursor ()  
  (let ((me (upi !<>)))  
    (db.agraph.cursor:make-transform-cursor (get-triples)  
      (lambda (triple) (setf (subject triple) me) triple)))) 

This, when put into the initfile, will cause /repositories/x/custom/talk-about-me to return all triples in that store, with their subjects replaced by <>.

The macro arguments look like this:

(methods permissions name result-type arguments &body body) 
Should be either one of `(:get :post :put :delete), or a list of them. Indicates the HTTP methods on which this service is reacheable.
Should be a string containing zero or more of the characters rwWes, indicating the permissions needed by a user to access this service. r stands for read, w for write, e for eval, and s for superuser. A capital W indicates that this service will mutate the store.
The name of the service. Can be any string.
The type of the value the service returns. This has to be known in order for the server to be able to do content-negotiation, and to be able to write the value out in all suitable formats. Allowed types are :string, :integer, :float, :boolean, :list (list of strings or triple-parts), :json (anything serializeable as JSON, using ST-JSON), or :triple-cursor (an AllegroGraph-style cursor). You can specify :dynamic here if you want the server to look at the returned value and dynamically determine a suitable output format.
A list of arguments, in the form (name type default), where default is optional. The arguments will be extracted from the HTTP request, and bound to the given variable names. As type, one can specify :string, :integer, :float, :boolean (true or false), :list (for arguments that can be specified multiple times), :body (the request body), :method (the request method), or :content-type (the content-type specified for the request body). Arguments are checked, so requests that leave off arguments for which no default is specified, or pass something that can't be interpreted as the correct type, will return an error response before the service body is even run.
This is the code that gets run to produce the service's response value. It will run with the argument names bound to their values.

User management

The AllegroGraph server uses a simple access-control scheme. Requests made with a valid Authorization header get assigned to the authorized user. Each user has a set of permissions, which are used to determine whether the request can be made.

The user named anonymous plays a special role. When such a user exists, any request made without authentication information is assigned to that user. By default, no anonymous user is defined, which disallows anonymous access.

The following permissions flags are defined:

This flag makes one a superuser, which means one can manage user accounts. Having the super permission automatically grants all other permissions.
This controls whether a user is allowed to use the 'eval-in-server' entry point, and use arbitrary Lisp code in their Prolog queries. This makes it possible to write more powerful queries, and move work to the server to improve efficiency. Note, however, that this also allows all manner of privilege escalation, and should only be granted to trusted users.
Controls whether a user can open their own sessions.

On top of that, read and write access can be specified per catalog and per store (as well as globally). read access allows one to query a repository, with write access one can also modify it. At the catalog level, write access permits the deleting and creating of repositories.

Each user can be assigned to a set of roles, each of which can also be granted permissions. The effective set of permissions that a user has is the union of their own permissions and those of their roles.

Most of the user-management services are only available to super-users. A normal user is allowed to inspect its own permissions, the status of its account, manage its user-data, and delete its own account.

GET /users

Returns a list of names of all the users that have been defined. For example:

GET /users HTTP/1.1  
Accept: application/json  
HTTP/1.1 200 OK  
Content-Type: application/json; charset=UTF-8  
["anonymous", "marijnh", "testuser"] 

PUT /users/[name]

Create a new user. Expects a password parameter, which specifies the user's password (can be left off when creating the anonymous user).

DELETE /users/[name]

Delete a user.

POST /users/[name]/password

Change the password for the given user. The new password should be in the body of the request.

GET /users/[name]/password/expired

Return a boolean indicating whether the user's password is expired.

POST /users/[name]/password/expired

Expire the user's password. If the password is expired, then any attempt to log in will result in HTTP error 401 with the message "Password expired.". However, an exception is made for changing one's own password: it can be done with an expired password. Changing the password cancels the expired status.

GET /users/[name]/roles

Retrieves a list of names, indicating the roles this user has.

PUT /users/[name]/roles/[role]

Add a role to a user.

DELETE /users/[name]/roles/[role]

Remove a role from a user. For example:

DELETE /users/anonymous/roles/project_a HTTP/1.1 

GET /users/[name]/permissions

List the permission flags that have been assigned to a user (any of super, eval, session). This is what a request fetching permission flags as plain text looks like:

GET /users/marijnh/permissions  
Accept: text/plain  
HTTP/1.1 200 OK  
Content-Type: text/plain; charset=UTF-8  

GET /users/[name]/effectivePermissions

Retrieve the permission flags assigned to the user, or any of its roles.

PUT /user/[name]/permissions/[type]

Assigns the given permission to this user. type should be super, eval, or session.

DELETE /user/[name]/permissions/[type]

Revokes the given permission for this user.

GET /users/[name]/access

Retrieve the read/write access for a user. This returns a result set, each element of which has a read, write, catalog, and repository component. The first two are booleans, the latter two strings. For permissions granted globally, catalog and repository will have a value of "*", for those granted per-catalog, only repository will be "*". catalog normally contains the catalog name, but for the root catalog "/" is used.

For example, read access to all repositories in the public catalog is specified (in JSON format) by:

{read: true, write: false, catalog: "public, repository: "*"} 

Whereas read/write access to repository scratch in the root catalog would be:

{read: true, write: true, catalog: "/", repository: "scratch"} 

GET /users/[name]/effectiveAccess

As above, but also includes the access granted to roles that this user has.

PUT /users/[name]/access

This is used to grant read/write access to a user. It takes four parameters:

Whether to grant read access. A boolean, defaults to false.
Whether to grant write access. Boolean, defaults to false.
Which catalog to grant the access on. Leave off or pass * to grant access on all catalogs. Again, use / for the root catalog.
Specifies the repository that access is granted on. Passing *, or leaving the parameter off, means all repositories in the given catalog.

This request grants the user testuser read access to all repositories in the root catalog:

PUT /users/testuser/access?read=true&catalog=%2f HTTP/1.1 

DELETE /users/[name]/access

Takes the same parameters as PUT on this URL, but revokes the access instead of granting it.

GET /users/[name]/data

Each user has a simple key-value store associated with it. This is mostly used by WebView to save some settings, but can be useful in other applications.

This service returns a result set containing id and uri fields, listing the keys stored for this user, and the URL under which their data can be found (see below).

GET /users/[name]/data/[key]

Fetches the user-data under key. Returns a string.

PUT /users/[name]/data/[key]

Stores data in a user's key-value store. The request body should contain the data to store.

DELETE /users/[name]/data/[key]

Deletes data from a user's key-value store.

GET /users/[name]/security-filters/[type]

Get list of security filters for a user.

allow or disallow

Items returned have keys s, p, o, and s per the parameters for POST security-filters.

POST /users/[name]/security-filters/[type]

Create filter for the user.

allow or disallow

Parameters are all optional:


See also roles/security-filters. and Security filters Documentation.

DELETE /users/[name]/security-filters/[type]

Delete filter for the user.

The type and parameters (s, p, o, and s) are the same as for POST security-filters.

GET /users/[name]/suspended

Returns a boolean indicating whether the user's account is suspended. All accounts start as unsuspended. They can be suspended explicitly by a superuser, in which case the account is suspended until explicitly unsuspended by a superuser. Accounts may also be suspended because of too many consecutive failed logins if the configuration option MaxFailedLogins is set appropriately. Accounts suspended for that reason can be unsuspended by a superuser but also may be unsuspended automatically after a period of time if the configuration option AccountUnsuspendTimeout is set appropriately. Suspended users cannot log in.

POST /users/[name]/suspended

Suspends the user's account.

DELETE /users/[name]/suspended

Unsuspends the user's account.

GET /users/[name]/enabled

Returns a boolean indicating whether the user's account is enabled. All accounts start as enabled and they stay enabled until disabled explicitly by superuser.

POST /users/[name]/enabled

Enables the user's account.

DELETE /users/[name]/enabled

Disables the user's account.

GET /roles

Returns the names of all roles that have been defined.

PUT /roles/[role]

Creates a new role.

DELETE /roles/[role]

Deletes a role. Any users that have been assigned this role will lose it.

GET /roles/[role]/permissions

Lists the permission flags granted to a role.

PUT /roles/[role]/permissions/[type]

Grant a role a certain permission. type should be super, eval, or session.

DELETE /roles/[role]/permissions/[type]

Revoke a permission for a role.

GET /roles/[role]/access

Query the access granted to a role. Returns a result in the same format as the equivalent service for users.

PUT /roles/[role]/access

Grant read/write access to a role. See here for the parameters that are expected.

DELETE /roles/[role]/access

Revoke read/write access for a role. Accepts the same parameters as above.

GET /roles/[role]/security-filters/[type]

Get list of security filters for a role.

Same parameters as user/security-filters.

POST /roles/[role]/security-filters/[type]

Create filter for the role.

Same parameters as user/security-filters.

DELETE /roles/[role]/security-filters/[type]

Delete filter for the role.

Same parameters as user/security-filters and POST role/security-filters.

Catalog interface

GET /protocol

Returns the protocol version of the Sesame interface, as an integer. The protocol described in this document is 4.

GET /repositories

Lists the repositories in this catalog. The result is a set of tuples containing id (the name of the repository) and uri (a link to the repository) fields. The fields readable and writable indicate whether your user has read and write access to the repository. Finally, there is the title field, which exists for Sesame compatibility, and contains the same value as the id field.

GET /repositories HTTP/1.1  
Accept: application/json  
HTTP/1.1 200 OK  
Content-Type: application/json; charset=UTF-8  
 ... other stores ...] 

Or, when fetching the list of repositories in a non-root catalog:

GET /catalogs/people/repositories HTTP/1.1 

PUT /repositories/[name]

Creates a new, empty repository. Supports several optional configuration arguments:

An integer, used to configure the expected size of the repository.
Can be specified any number of times. Should hold index IDs, and is used to configure the set of indices created for the store.

When a repository with the given name already exists, it is overwritten, unless a parameter override with value false is passed.

DELETE /repositories/[name]

Delete a repository. Might fail if someone else is accessing the repository.

Repository interface

GET/POST /repositories/[name]

(Note that if no query parameter is given, this will return the WebView HTML page instead of the service described here.)

This URL is used to run queries against a repository. It conforms to both the Sesame and SPARQL protocol interfaces — this is why some of the parameter names may look inconsistent.

SPARQL/Update queries are allowed, but only when the request is made using POST instead of GET (and the user has write access to the repository).

This service takes the following parameters:

The query to be executed. The query may use the namespace prefixes defined for the user.
The language of the query. Can be sparql or prolog. Defaults to sparql.
A string that determines what kind of reasoning is used when executing the query. Default is false, no reasoning. Other options are rdfs++ (same as true), and restriction for hasValue as well as rdf++ reasoning.
Can be passed zero or more times. Sets the graph name, or list of graph names, that will be used by this query (as in FROM). When no context is given, all graphs are used. The string null can be used to refer to the default graph of the store.
Also can be given multiple times. Behaves like context, except that the named graphs retain their names (as in FROM NAMED).
Can be passed any number of times. Works like context except that plain-URI values should be given, without wrapping < and > characters.
Works like default-graph-uri, except that this specifies the named graphs to be used.
An integer. Can be used to limit the amount of results returned by the query.
An integer. Can be used to skip the first offset results in result set.
Parameters starting with a $ character can be used to bind query variables to fixed value (an N-Triples term) when executing a SPARQL query.
A boolean that defaults to false, indicating whether an error should be raised when a SPARQL query selects variables that are not mentioned in the query body.
Can be used to provide a resource (URL) name for the default graph. Any references to that resource will reference the default graph, and in the output the resource will be substituted for any references to the default graph. Can be given multiple times to have multiple names refer to this graph, but only the first one will appear in the output.
Can be used to control the way the query is planned .
When given, will cause the server to (as well as executing the query normally) save the query as a prepared query under the name passed as the value for this parameter. See preparing queries below.
When given, it must be a positive integer which specifies the maximum query execution time (units are seconds). If unspecified, there is no limit on how long the query will run.

The result formats supported depends on the query. Prolog queries return tabular data, as do SPARQL select queries. describe or construct queries return triples, and ask queries return a boolean value.

Prolog queries are allowed to return nested lists of results, in which case the result can only be returned as application/json, and the nested structure (both in the list of column names and in the results themselves) will be represented as nested JSON arrays.

POST /repositories/[name]/ensureNotLingering

To conserve resources, makes the database instance and its child processes exit if the instance is unused. Takes no arguments, returns nothing.

Normally unused database instances linger for InstanceTimeout seconds to speed up subsequent open operations.

GET /repositories/[name]/size

Returns the amount of statements in the repository, as an integer. Takes a single optional argument, context, which can be used to restrict the count to a single named graph.

GET /repositories/[name]/statements

Retrieves statements (triples) by matching against their components. All parameters are optional — when none are given, every statement in the store is returned.

Match a specific subject. When given, should be a term in N-triples format. Can be given multiple times to mean 'the subject must be one of these resources'.
Can only be given if exactly one subj parameter is given. Matches a range of subjects.
Match a specific predicate. Can be passed multiple times, like subj, to match a set.
Perform a range query on the predicate.
Match a specific object. Pass multiple values to match a set.
Range query on objects.
Can be given multiple times. Restricts the query to the given list of named graphs. When not specified, all graphs in the store are used.
Range query on contexts / graph names.
An integer indicating the maximum amount of results to return.
An integer. Tell the server to skip a number of results before it starts returning.
Used to turn on reasoning for this query. Valid values are false, rdfs++, and hasvalue. Default is false — no reasoning.

GET/POST /repositories/[name]/statements/query

Exposes the exact same interface as GET /statements. Useful when you need to make a POST request because of URL-length limitations.

DELETE /repositories/[name]/statements

Deletes statements matching the given parameters. When no parameters are given, every triple in the store is deleted. Returns the number of triples deleted.

Match a specific subject. When given, should be a term in N-triples format.
Match a specific predicate.
Match a specific object.
Match a specific graph name.

This deletes all statements in the graph named "A" (of which there are 25).

DELETE /repositories/repo1/statements?context=%22A%22 HTTP/1.1  
Accept: application/json  
HTTP/1.1 200 OK  
Content-Type: application/json; charset=UTF-8  

PUT/POST /repositories/[name]/statements

Add statements to a store. When the PUT method is used, the store is emptied first, whereas the POST method just adds triples. The value returned is an integer indicating the amount of triples loaded.

The Content-Type header determines the way the given data is interpreted. The following formats are supported:

Normally, the body of the request is used as input data. One can pass a file parameter to indicate a (server-side) file-name to be loaded, or a url parameter to load a file directly off the web. Other supported parameters are:

When loading RDF/XML data, the value of this parameter is used as the base URI of the document.
Used to set the named graph into which the new data is stored.
A positive integer. Will cause a commit to happen after every N added statements. Can be used to work around the fact that importing a huge amount of statements in a single transaction will require excessive amounts of memory.
A boolean (default is false) that indicates whether the load should continue when malformed data is encountered. Currently only works for N-Triples and N-Quads data.
A boolean (default is false) that indicates whether external references in RDF/XML source files will be followed. When true, the caller must have eval permissions.
A boolean (default is false). If true, then less validation is done for the N-Triples and N-Quads formats.

This request, where [URL] is the encoded form of some URL that contains an N-triples file, loads the triples from that URL into the scratch store under context "B".

POST /repositories/scratch/statements?url=[URL]&context=%22B%22 HTTP/1.1  
Accept: application/json  
HTTP/1.1 200 OK  
Content-Type: application/json; charset=UTF-8  

Preparing Queries

It is possible to cache the parsing and parameterization of SPARQL queries. This is currently only recommended for very big queries that get repeated a lot, since the savings are not very large.

Preparing queries is only supported in a dedicated session, and the prepared queries will only be available in that session.

GET/POST /repositories/[name]/queries/[id]

Executes a prepared query. Supports the limit, offset, and bound-variable ($[varname]) parameters in the same way as the regular query interface, and takes all the other parameters from the prepared query stored under the name id (either through the save argument to a regular query, or a PUT request to this URL.

PUT /repositories/[name]/queries/[id]

Prepares a query. Accepts the query, infer, context, namedContext, default-graph-uri, named-graph-uri, checkVariables, defaultGraphName, and planner arguments in the same way that the regular query interface accepts them, but instead of executing the query, it prepares it and saves it under id.

DELETE /repositories/[name]/queries/[id]

Deletes the prepared query stored under id.

RDF-Transaction data format

The application/x-rdftransaction mime type, as accepted by the /statements service, is not standardized or even widely documented, so we quickly describe it here. The format originates from the Sesame HTTP protocol.

Documents of this type are XML documents containing a number of triple additions and removals. They are executed as a transaction - either completely, or not at all. (Though note that this does not mean an implicit commit is executed on dedicated session.)

An RDF transaction document looks roughly like this:

    <literal datatype="">1917-05-29</literal>  

A transaction's root tag is always transaction. Inside of this, any amount of actions are specified, which are executed in the order in which they are given.

The add action adds a triple. It should contain at least three nodes, which specify the subject, predicate, and object of the triple. After that, any number of nodes may follow, which specify contexts that the triple should be added to. A null tag can be used to specify the default context. If no contexts are given, the triple is inserted into only the default context.

These child nodes can be either an uri tag containing a resource's URI, a bnode tag containing an ID that is used to be able to refer to the same blank node multiple times in the document, or a literal tag, that contains a string. literal tags may have datatype or xml:lang attributes to assign them a type or a language.

In a remove tag, the first three child nodes optionally specify the subject, predicate, and object of the triples to remove. Any of these can be given as a null tag (or left out altogether) to count as a wild-card. After these, any number of nodes can follow, which specify the contexts to remove nodes from. If none are given, nodes are removed from all contexts. A null tag is used to specify the default context here. removeFromNamedContext works the same, but requires a single context to be specified.

clear also removes triples, but without the option to specify subject, predicate, or object. All child nodes are interpreted as contexts. Again, not specifying any contexts causes nodes to be removed from all contexts.

POST /repositories/[name]/statements/delete

Used to delete a set of statements. Expects a JSON-encoded array of triples as the posted data, and deletes all statements listed in there. Content-Type should be application/json.

When an ids parameter with the value true is passed, the request body should contain a JSON-encoded list of triple-ids, instead of actual triples.

GET /repositories/[name]/statements/id

Fetches a set of statements by ID. Takes any number of id parameters, and returns a set of triples.

GET/POST /repositories/[name]/unique/[column]

Find the set of unique terms in a column. column can be one of obj, pred, subj, or context. Without arguments, it simply finds the set of all terms that occur in that column in the database. The query can be further refined by passing optional obj, pred, subj, or context parameters, which, when given a term, restrict the result to only triples that contain that term in that position. For example, to get all outgoing predicates from term <>, query /unique/pred with the parameter subj set to <>.

Returns a set of terms.

GET /repositories/[name]/statements/duplicates

Gets all duplicate statements that are currently present in the store. The mode parameter can be either spog (the default) or spo to indicate which components of each triple must be equivalent to count as duplicates of each other. See Deleting Duplicate Triples.

DELETE /repositories/[name]/statements/duplicates

Deletes all duplicate statements that are currently present in the store. The mode parameter can be either spog (the default) or spo to indicate which components of each triple must be equivalent to count as duplicates of each other. See Deleting Duplicate Triples

GET /repositories/[name]/suppressDuplicates

Returns the duplicate suppression strategy currently active in the store. This returns either false if no duplicate suppression is active, or spog if it is active. See Deleting Duplicate Triples.

PUT /repositories/[name]/suppressDuplicates

Sets the duplicate suppression strategy for the store. The type argument can be either false (disable duplicate suppression), spo (enable it, eliminate all spo duplicates on commit) or spog (enable it, eliminate all spog duplicates on commit). See Deleting Duplicate Triples.

DELETE /repositories/[name]/suppressDuplicates

Disable duplicate suppression for the store. This is the equivalent of using PUT /repositories/(name)/suppressDuplicates with false as the type argument. See Deleting Duplicate Triples.

GET /repositories/[name]/contexts

Fetches a list of named graphs in the store. Returns a set of tuples, each of which only has a contextID field, which is an N-triples string that names the graph.

GET /catalogs/people/repositories/repo1/contexts HTTP/1.1  
Accept: application/sparql-results+xml  
HTTP/1.1 200 OK  
Content-Type: application/sparql-results+xml; charset=UTF-8  
<?xml version="1.0"?>  
<sparql xmlns="">  
  <head><variable name="contextID"/></head>  
      <binding name="contextID">  

POST /repositories/[name]/functor

Define Prolog functors, which can be used in Prolog queries. This is only allowed when accessing a dedicated session.

The body of the request should hold the definition for one or more Prolog functors, in Lisp syntax, using the <-- or <- operators.

POST /repositories/[name]/commit

POST /repositories/[name]/begin

Begin a new transaction. It is an error if there is already an active transaction (400 Bad Request is returned with message NESTED TRANSACTION Cannot begin a new transaction while there is one already active). This request is only meaningful in dedicated sessions and with Sesame 2.7 transaction handling semantics. In shared back-ends or with Sesame 2.6 semantics, it does nothing.

POST /repositories/[name]/commit

Commit the current transaction. Only meaningful in dedicated sessions.

POST /repositories/[name]/rollback

Roll back the current transaction (discard all changes made since the beginning of the transaction). Only meaningful in dedicated sessions.

POST /repositories/[name]/eval

Evaluates the request body in the server. By default, it is evaluated as Common Lisp code, with *package* set to the db.agraph.user package. If you specify a Content-Type of text/javascript, however, the code will be interpreted as JavaScript.

Makes an attempt to return the result in a sensible format, falling back to printing it (as per prin1) and returning it as a string.

GET/POST /repositories/[name]/freetext

Perform a query on the free-text indices of the store, if any. A list of matching triples is returned.

The text to search for. Either this or expression should be passed. Putting multiple words in this argument means 'match only triples with all these words'. Double-quoting a part of the string means 'only triples where this exact string occurs'. Non-quoted words may contain wildcards - * (matches any string) and ? (matches any single character). Or they can end in ~ to do a fuzzy search, optionally followed by a decimal number indicating the maximum Levenshtein distance to match. A vertical bar (|) can be used between patterns to mean 'documents matching one of these patterns', and parentheses can be used to group sub-patterns. For example: "common lisp" (programming | develop*).
An S-expression combining search strings using and, or, phrase, match, and fuzzy. For example (and (phrase "common lisp") (or "programming" (match "develop*"))).
An optional parameter that restricts the search to a specific free-text index. If not given, all available indices are used.
A boolean indicating whether the results should be sorted by relevance. Default is false.
An integer limiting the amount of results that can be returned.
An integer telling the server to skip the first few results.

GET /repositories/repo1/freetext?pattern=RDF HTTP/1.1 Accept: text/plain

HTTP/1.1 200 OK Content-Type: text/plain; charset=UTF-8

"AGraph RDF store". .... others ....

GET /repositories/[name]/freetext/indices

Returns a list of names of free-text indices defined in this repository.

GET /repositories/[name]/freetext/indices/[index]

Only returns application/json responses. Returns the configuration parameters of the named free-text index. This will be an object with the following fields:

An array of strings. Empty if the index indexes all predicates, containing only the predicates that are indices otherwise.
Can be true (index all literals), false (no literals), or an array of literal types to index.
Can be true (index resources fully), false (don't index resources), or the string "short" to index only the part after the last # or / in the resource.
An array containing any of the strings "subject", "predicate", "object", and "graph". This indicates which fields of a triple are indexed.
An integer, indicating the minimum size a word must have to be indexed.
A list of words, indicating the words that count as stop-words, and should not be indexed.
A list of word filters configured for this index (see below).
A list of character specifiers configured for this index (see below).
A list of character specifiers configured for this index.
The name of the tokenizer being used (currently either default or japanese).

GET /repositories/[name]/freetext/indices/[index]/[param]

If [param] is one of the slot values mentioned above, the corresponding configuration parameter of the index is returned.

PUT /repositories/[name]/freetext/indices/[index]

Create a new free-text index. Takes the following parameters:

Can be specified multiple times. Indicates the predicates that should be indexed. When not given, all predicates are indexed.
A boolean (defaults to true) that determines whether literal are indexed.
When indexLiterals is true, this parameter can be given any number of times to restrict the types of literals that are indexed. When not given, all literals (also untyped ones) are indexed.
Can be given the values true, false, or short. Default is false. short means to index only the part of the resource after the last # or / character.
May be specified multiple times, must be one of subject, object, predicate, or graph. Determines which fields of a triple to index. Defaults to just object.
An integer. Determines the minimum size a word must have to be indexed.
Can be passed multiple times. Determines the set of stop-words, words that are not indexed. Defaults to a small set of common English words. To override this default and specify that no stop-words should be used, pass this parameter once, with an empty value.
Specify a word filter, which is an operation applied to words before they are indexed and before they are searched for. Used to 'normalize' words. Can be passed multiple times to specify multiple filters. Currently the only valid values are stem.english (a simple English-language stemmer), drop-accents (will turn 'é' into 'e', etc.), and soundex, for the Soundex algorithm.
Can be passed multiple times. The character set to be used as the constituent characters of a word. Each parameter is part of a character set, and can be one of the following:
  • The word alpha - all (unicode) alphabetic characters
  • digit - all base-10 digits
  • alphanumeric - all digits and alphabetic characters
  • a single character
  • a range of characters: a single character, followed by a dash (-) character, followed by another single character.
Can be passed multiple times. The character set to be used as the border characters of indexed words. Uses the same syntax as innerChars.
An optional string. Can be either default or japanese. When japanese is passed, the tokenizer is based on morphological analysis, and the innerChars and borderChars parameters are ignored. For japanese, it is also recommended to set minimumWordSize to either 1 or 2.

POST /repositories/[name]/freetext/indices/[index]

This can be use to reconfigure a free-text index. It takes the all the parameters that the PUT service takes. Parameters not specified are left at their old values. To indicate that the predicate, indexLiteralType, indexField, stopWord, or wordFilter parameters should be set to the empty set instead of left at their default, pass these parameters once, with the empty string as value.

The parameter reIndex, a boolean which defaults to true, the client can control whether a full re-indexing of the modified index should take place, or whether the new settings should be used when indexing triples added after the redefinition.

DELETE /repositories/[name]/freetext/indices/[index]

Delete the named index from the repository.

POST /repositories/[name]/blankNodes

Ask the server to allocate and return a set of blank nodes. Takes one argument, amount, which should be an integer.

These nodes can, in principle, be used to refer to nodes when using other services. Note, however, that a lot of the standards related to RDF give blank nodes a document-wide scope, which means that referring to blank nodes by name from, for example, a SPARQL query or N-Triples document is not possible. The nodes are interpreted as local to the document, and assigned to a new blank node.

POST /repositories/repo1/blankNodes?amount=2 HTTP/1.1  
Accept: text/plain  
HTTP/1.1 200 OK  
Content-Type: text/plain; charset=UTF-8  

GET /repositories/[name]/tripleCache

Find out whether the 'SPOGI cache' is enabled, and what size it has. Returns an integer — 0 when the cache is disabled, the size of the cache otherwise.

PUT /repositories/[name]/tripleCache

Enable the spogi cache in this repository. Takes an optional size argument to set the size of the cache.

DELETE /repositories/[name]/tripleCache

Disable the spogi cache for this repository.

GET /repositories/[name]/noCommit

Returns a boolean that tells you whether this repository is currently in no-commit mode. When this mode is active, all commits from any clients will return an error, effectively preventing writing to the store. This is mostly useful for warm standby clients, but can also be used to enforce read-only stores in other situations.

PUT/DELETE /repositories/[name]/noCommit

Turns no-commit mode on (PUT) or off (DELETE).

GET /repositories/[name]/bulkMode

Returns a boolean indicating whether bulk-load mode is enabled for the repository.

PUT/DELETE /repositories/[name]/bulkMode

Turn bulk-load mode on (PUT) or off (DELETE).


In order to make queries shorter and more readable, a user can define namespaces, which will be used for queries issued by this user.

GET /repositories/[name]/namespaces

List the currently active namespaces for your user, as tuples with prefix and namespace (the URI) fields. For example:

GET /catalogs/scratch/repositories/repo2/namespaces HTTP/1.1  
Accept: application/json  
HTTP/1.1 200 OK  
Content-Type: application/json; charset=UTF-8  
 ... etc ...] 

DELETE /repositories/[name]/namespaces

Deletes all namespaces in this repository for the current user. If a reset argument of true is passed, the user's namespaces are reset to the default set of namespaces.

GET /repositories/[name]/namespaces/[prefix]

Returns the namespace URI defined for the given prefix.

PUT/POST /repositories/[name]/namespaces/[prefix]

Create a new namespace. The body of the request should contain the URI for the namespace, as plain text, as in:

POST /catalogs/scratch/repositories/repo2/namespaces/ex HTTP/1.1  
Content-Type: text/plain; charset=UTF-8 

DELETE /repositories/[name]/namespaces/[prefix]

Delete a namespace.

Type mappings

AllegroGraph supports storing some types of literals in encoded, binary form. This typically makes them smaller (less disk usage), and makes it possible to run range queries against them. For more details, see Datatypes and the Lisp reference for Data-type and Predicate Mapping.

Specifying which literals should be encoded can be done in two ways. There is 'datatype mapping', where a literal type is marked, and all literals of that type are encoded, and there is 'predicate mapping', where the objects in a statement with a given predicate are encoded.

When specifying a mapping, one has to choose an encoding to apply. To identify these encodings, we use XSD datatypes:

Typed literals of these types will be encoded by default. For other types, you have to specify your mapping before you import your data to have the encoding take place.

GET /repositories/[name]/mapping

Fetches a result set of currently specified mappings. Each row has a kind (datatype or predicate), part (the resource associated with the mapping), and encoding fields.

DELETE /repositories/[name]/mapping

Clear all non-automatic type mappings for this repository.

DELETE /repositories/[name]/mapping/all

Clear all type mappings for this repository including the automatic ones.

GET /repositories/[name]/mapping/type

Yields a list of literal types for which datatype mappings have been defined in this store.

GET /repositories/test/mapping/type HTTP/1.1  
HTTP/1.1 200 OK  
Content-Type: text/plain; charset=UTF-8  
... etc ... 

PUT/POST /repositories/[name]/mapping/type

Takes two arguments, type (the RDF literal type) and encoding, and defines a datatype mapping from the first to the second. For example, if [TYPE] is the URL-encoded form of <>, and [ENC] of <>, this request will cause myInteger literals to be encoded as integers:

PUT /repositories/test/mapping/type?type=[TYPE]&encoding=[ENC] HTTP/1.1 

DELETE /repositories/[name]/mapping/type

Deletes a datatype mapping. Takes one parameter, type, which should be an RDF resource.

GET /repositories/[name]/mapping/predicate

Yields a list of literal types for which predicate mappings have been defined in this store.

PUT/POST /repositories/[name]/mapping/predicate

Takes two arguments, predicate and encoding, and defines a predicate mapping on them.

DELETE /repositories/[name]/mapping/predicate

Deletes a predicate mapping. Takes one parameter, predicate.


Triple stores can be equipped with a variety of indices, which will affect query performance. By default, a store gets a sensible set of indices, but it is possible to tweak this set.

Indices are identified by cryptic IDs, such as spogi, which stands for "subject, predicate, object, graph, id", the order in which the fields of triples are used when sorting for that index.

GET /repositories/[name]/indices

Returns a list of index IDs that are enabled for this store. When a listValid=true parameter is passed, a list of all supported index types is returned.

PUT /repositories/[name]/indices/[type]

Ensures that the index indicated by type is present in this store. Takes effect at commit time (which is, of course, immediately when using a shared back-end or an auto-commit session).

DELETE /repositories/[name]/indices/[type]

Removes the index indicated by type from the store. Also takes effect at commit time.

POST /repositories/[name]/indices/optimize

Tells the server to try to optimize the indices for this store. The arguments are wait, level, and index. All are optional. Here is a sample call:

POST /repositories/myrepo/indices/optimize?wait=false&level=1&index&spogi&index=gposi 

Geo-spatial queries

When literals are encoded as geo-spatial values, it is possible to efficiently perform geometric queries on them.

In order to do this, one defines a geo-spatial datatype, and then adds literals of that type to the store. AllegroGraph supports two kinds of geo-spatial datatypes, cartesian and spherical. A cartesian literal looks like this:

"+10.0-17.5"^^<[cartesian type]> 

Where the numbers are the X and Y coordinates of the point. A spherical literal uses ISO 6709 notation, for example:

"+37.73+122.22"^^<[spherical type]> 

GET /repositories/[name]/geo/types

Retrieve a list of geospatial types defined in the store.

POST /repositories/[name]/geo/types/cartesian

Define a new Cartesian geospatial type. Returns the type resource, which can be used as the type argument in the services below.

A floating-point number that determines the granularity of the type.
xmin, xmax, ymin, ymax
Floating-point numbers that determine the size of the Cartesian plane that is modelled by this type.

POST /repositories/[name]/geo/types/spherical

Add a spherical geospatial type. Returns the type resource.

A floating-point number that determines the granularity of the type.
Can be degree, radian, km, or mile. Determines the unit in which the stripWidth argument is given. Defaults to degree.
latmin, longmin, latmax, longmax
Optional. Can be used to limit the size of the region modelled by this type. Default is to span the whole sphere.

For example, this defines a type with a granularity of 2 degrees:

POST /repositories/repo1/geo/types/spherical?stripWidth=2 HTTP/1.1  
Accept: text/plain  
HTTP/1.1 200 OK  
Content-Type: text/plain; charset=UTF-8  

(The newline in the URI wouldn't be there in the actual response.)

GET /repositories/[name]/geo/box

Fetch all triples with a given predicate whose object is a geospatial value inside the given box.

The geospatial type of the object field.
The predicate to look for.
xmin, ymin, xmax, ymax
The bounding box.
Optional. Used to limit the amount of returned triples.
Optional. Used to skip a number of returned triples.
Optional argument to control whether triples inferred through reasoning should be included. Default is false (no reasoning), other accepted values are rdfs++ and hasvalue.
A boolean parameter that defaults to false. When true, the context (graph) field of triples is used as the geospatial value to match, rather than the object.

GET /repositories/[name]/geo/circle

Retrieve triples within a circle. type, predicate, infer, limit, offset, and useContext argument as with /geo/box. Takes x, y, and radius arguments, all floating-point numbers, to specify the circle.

GET /repositories/[name]/geo/haversine

Retrieve triples whose object lies within a circle in a spherical system. Takes type, predicate, infer, limit, offset, and useContext arguments like /geo/box, lat and long arguments to specify the centre of the circle, and a radius argument. The unit used for radius defaults to kilometre, but can be set to mile by passing a unit parameter with the value mile.

PUT /repositories/[name]/geo/polygon

Create a polygon in the store. Takes a parameter resource, which sets the name under which the polygon should be stored, and three or more point arguments, which must be geospatial literals that represent the points of the polygon.

GET /repositories/[name]/geo/polygon

Retrieve triples whose object lies inside a polygon. type, predicate, infer, limit, offset, and useContext work as with /geo/box. The polygon parameter must hold the name of a polygon created with the above service.

Social network analysis

Some queries and operations are best expressed using graph-traversing generators . The following services make it possible to define such generators.

PUT /repositories/[name]/snaGenerators/[generator]

Creates a new generator under the given name. Accepts the following parameters:

A predicate. Accepted multiple times. Causes the new generator to follow edges with the given predicate.
Like objectOf, but follow edges from object to subject.
Like the above, but follow edges in both directions.
A Prolog query, in which the variable ?node can be used to refer to the 'start' node, and whose results will be used as 'resulting' nodes. User namespaces may be used in this query.

PUT /repositories/[name]/neighborMatrices/[matrix]

Create a neighbor-matrix, which is a pre-computed generator.

A set of N-Triples terms (can be passed multiple times) which serve as start nodes for building up the matrix.
The generator to use, by name.
An integer specifying the maximum depth to which to compute the matrix. Defaults to 1.


Normal requests will be handled by one of a set of shared back-end processes. (See Server Configuration and Control.) The fact that they are shared between all incoming requests means there are certain limitations on the way they can be used--a shared back-end does not support transactions spanning multiple requests, and does not allow one to define new Prolog functors. Further, though it is possible to load scripts into shared back-ends, it is highy recommended that you not do so. Function definitions, global variables, and custom-services, once loaded, remain active in a shared back-end until that backend is killed. Javascript scripts are sandboxed for the most part and only leak custom-service definitions into the back-end. Unless you are extremely careful about what scripts are available to be loaded, it is possible that code you expect to run as part of a server-side script has been overwritten by other requests, and unexpected errors may result. A session, on the other hand, can only be accessed by the user who requested it, and so the scripts loaded can be carefully managed.

When transactions, functors, or other persistent state is needed, it is necessary to create a session. This spawns a process that you will have exclusive access to. Sessions are effectively single threaded: only one request at a time will be processed.

Requests to a session URL, unless the autoCommit parameter is given when starting the session, are handled inside a transaction. That means modifications to the store are not visible in other sessions or in shared back-ends until a commit request is made. A rollback request can be used to discard any changes made since the beginning of the current transaction.

When making extra requests to commit and rollback is too expensive, one can use the x-commit and x-rollback HTTP headers (with any value) to have the rollback or commit command piggyback on another request. x-rollback will cause the store to be rolled back before evaluating the request, x-commit will cause a commit after evaluating the request.

Sessions time out after a certain period of idleness (can be set on creation), so an application that depends on a session being kept alive should periodically ping its session.

Sessions belonging to superusers allow their owners to use the x-masquerade-as-user header in HTTP requests to activate another user's security filters. See user and role security-filters and Security filters documentation.

Transaction handling semantics

A session can execute all kinds of requests (even those that mutate the store) within or without transactions.

When within a transaction, database modifications are not visible in other sessions or in shared back-ends until a commit request is made. A rollback request can be used to discard any changes made since transaction started.

When there is no active transaction, each individual request is executed as if in a separate transaction.

It's generally true that having an active transaction is equivalent to auto-commit mode being off, but the details of how the two modes are entered and exited differ depending on whether the server is configured to operate under Sesame 2.6 or Sesame 2.7 semantics.

With Sesame 2.6 transaction handling, if the active transaction is rolled back or committed, a new one is started immediately. Auto-commit mode is only ever changed by explicit request.

With Sesame 2.7 transaction handling, transactions must be started explicitly with begin. commit and rollback do not start a new transaction. Instead they turn auto-commit mode on. Changing the auto-commit flag is deprecated in favor of begin and commit. However, to maintain some backwards compatibility, turning auto-commit off starts a new transaction, and turning it off commits the transaction.

Extra HTTP headers

When making explicit requests to rollback, begin, or commit is too expensive, one can use the x-rollback, x-begin and x-commit HTTP headers (with any value) to have the rollback, begin or commit command piggyback on another request. A request may have any combination of these extra headers. The execution semantics are as follows: before the request is performed, if x-rollback is present, then the store is rolled back. Next, if x-begin is present, a new transaction is started. Then the actual request is performed and the store is committed if x-commit is supplied.

POST /session

Creates a new session. Takes the following parameters:

A boolean, which determines whether the session starts in autocommit mode. With Sesame 2.6 transaction handling semantics, the default is false, meaning that the session starts as if a transaction had been started. With Sesame 2.7 transaction handling semantics, the default is true and an explicit begin is required.
An integer, specifying the amount of seconds the session can be idle before being collected.
A boolean, defaulting to false, which determines whether the initfile is loaded into this session.
May be specified any number of time, to load server scripts into the new session.
A string indicating the kind of store that has to be opened, see below.

The minilanguage used by the store parameter works as follows:

Indicates the triple store named "store1" in the root catalog.
The triple store "store2" in the "catalog1" catalog.
A remote store, by URL. If the URL points to the server itself, the store will be opened locally.
<a> + <b>
The federation of stores "a" and "b".
The store "a", with rdfs++ reasoning applied (restriction is also supported as a reasoner type). You can specify the context that inferred triples get using this syntax: <a>[rdfs++#<>]
<a>{null <>}
Store "a", filtered to only contain the triples in the default graph (null) and the graph named Any number of graphs can be given between the braces.

This syntax can be composed to created federations of filtered and reasoning stores, for example <http://somehost:10035/repositories/<a>{null} + <b>[rdfs++].

The service returns the URL of the new session. Any sub-URLs that were valid under a repository's URL will also work under this session URL. For example, if http://localhost:55555/sessions/7e8df8cd-26b8-26e4-4e83-0015588336ea is returned, you can use http://localhost:55555/sessions/7e8df8cd-26b8-26e4-4e83-0015588336ea/statements to retrieve the statements in the session.

POST /repositories/[name]/session

This is a shortcut for creating a session in a local triple store. It takes autoCommit, loadInitfile, script, and lifetime arguments as described above, and creates a session for the store that the URL refers to.

POST [session-url]/session/close

Explicitly closes a session.

GET [session-url]/session/ping

Let the session know you still want to keep it alive. (Any other request to the session will have the same effect.)

GET [session-url]/session/isActive

Returns a boolean indicating whether there is a currently active transaction for this session. This is the logical complement of autoCommit.

GET [session-url]/session/autoCommit

Returns a boolean indicating whether auto-commit is currently active for this session. This is the logical complement of isActive.

POST [session-url]/session/autoCommit

Used to change the auto-commit flag for the session. Takes a single boolean argument called on. If it is set to true in a transaction, then the transaction is automatically committed. If it is set to false outside a transaction, then a new transaction is started. Note that with Sesame 2.7 semantics, commit and rollback set this flag to true.

POST [session-url]/commit

Commit the current transaction. With Sesame 2.6 semantics, a new transcation is started. With Sesame 2.7 semantics, auto-commit mode is entered.

POST [session-url]/rollback

Roll back the current transaction. With Sesame 2.6 semantics, a new transcation is started. With Sesame 2.7 semantics, auto-commit mode is entered.

GET /repositories/[name]/mongoParameters

Returns a JSON object with keys:

Note, the password set with POST is not returned.

POST /repositories/[name]/mongoParameters

Accepts the following parameters:

See also: MongoDB interface.

Process management

An AllegroGraph server consists of a group of different operating-system processes. The following services provide a minimal process-debugging API over HTTP. All of them are only accessible to superusers.

GET /processes

Returns a list of tuples showing the processes the server currently has running. Each row has pid (the OS process ID) and a name (the process name) properties.

GET /processes/[id]

Returns stack traces for the threads in the given process.

DELETE /processes/[id]

Kills the specified process. Obviously, you yourself are responsible for any adverse effects this has on the functioning of the server.

POST /processes/[id]/telnet

Starts a telnet server in the specified process. A random port will be chosen for the server, which will be returned as the response. Note that the telnet server will allow anyone to connect, creating a something of a security concern if the server is on an open network.

Warm Standby

Warm standby allows a second AllegroGraph server to keep an up-to-date copy of a repository on another server. In case the first server fails, this copy can be made to take over its responsibilities.

Documentation on this feature is not complete yet.

GET /repositories/[name]/warmstandby

Only supports an application/json accept type. Returns a representation of the current standby status for repository.

PUT /repositories/[name]/warmstandby

Requires jobname, primary, primaryPort, user, and password parameters. Makes this repository start replicating the source store on the server primary:primaryPort, using the given credentials to gain access.

DELETE /repositories/[name]/warmstandby

Stops a replication job. This command is sent to the client.

POST /repositories/[name]/warmstandby/switchRole

Sent to a repository that is currently functioning as a replication server. Causes a client (identified by the jobname parameter) to take over, making this repository a client of that new server. Takes primary, primaryPort, user, and password parameters, which specify the server on which the client repository lives, and the user account to use to access this server.

The becomeClient boolean parameter, which defaults to true, determines whether the server will start replicating its old client. The enableCommit parameter, also defaulting to true, controls whether no-commit mode will be turned off in the client.