kiwix-serve.rst

kiwix-serve

Introduction

kiwix-serve is a tool for serving ZIM file content over HTTP. It supports serving a library containing multiple ZIM files. In a large library served by a kiwix-serve instance clients can look up/filter ZIM files of interest by words in their :term:`titles <ZIM title>` and/or descriptions, language, tags, etc.

kiwix-serve provides a ZIM file viewer for displaying inidividual pages from a ZIM file inside the user's web browser (without downloading the full ZIM file).

Clients can also remotely search inside those ZIM files that contain a full-text search database.

kiwix-serve supports Web browsers Firefox >= 70, Chrome >= 80, Edge >= 80, ChromeAndroid >= 80, Safari >= 14, iOS >= 14.

Usage

kiwix-serve --library [OPTIONS] LIBRARY_FILE_PATH
kiwix-serve [OPTIONS] ZIM_FILE_PATH ...

Arguments

LIBRARY_FILE_PATH: path of an XML library file listing ZIM files to serve. To be used only with the :option:`--library` option. Multiple library files can be provided as a semicolon (;) separated list.

ZIM_FILE_PATH: ZIM file path (multiple arguments are allowed).

Options

.. option:: --library

  By default, ``kiwix-serve`` expects a list of ZIM files as command line
  arguments. Providing the :option:`--library` option tells ``kiwix-serve``
  that the command line argument is rather a :ref:`library XML file
  <cli-arg-library-file-path>`.

.. option:: -i ADDR, --address=ADDR

  Listen only on this IP address. By default the server listens on all
  available IP addresses. Alternatively, you can use special values to define which types of connections to accept:

  - all : Listen for connections on all IP addresses (IPv4 and IPv6).
  - ipv4 : Listen for connections on all IPv4 addresses.
  - ipv6 : Listen for connections on all IPv6 addresses.

.. option:: -p PORT, --port=PORT

  TCP port on which to listen for HTTP requests (default: 80).

.. option:: -r ROOT, --urlRootLocation=ROOT

  URL prefix on which the content should be made available (default: empty).

.. option:: -d, --daemon

  Detach the HTTP server daemon from the main process.

.. option:: -a PID, --attachToProcess=PID

  Exit when the process with id PID stops running.

.. option:: -M, --monitorLibrary

  Monitor the XML library file and reload it automatically when it changes.

  Library reloading can be forced anytime by sending a SIGHUP signal to the
  ``kiwix-serve`` process (this works regardless of the presence of the
  :option:`--monitorLibrary`/:option:`-M` option).

.. option:: -m, --nolibrarybutton

  Disable the library home button in the ZIM viewer toolbar.

.. option:: -n, --nosearchbar

  Disable the searchbox in the ZIM viewer toolbar.

.. option:: -b, --blockexternal

  Prevent the users from directly navigating to external resources via such
  links in ZIM content.

.. option:: -t N, --threads=N

  Number of threads to run in parallel (default: 4).

.. option:: -s N, --searchLimit=N

  Maximum number of ZIM files in a fulltext multizim search (default: No limit).

.. option:: -z, --nodatealiases

  Create URL aliases for each content by removing the date embedded in the file
  name. The expected format of the date in the filename is ``_YYYY-MM``. For
  example, ZIM file ``wikipedia_en_all_2020-08.zim`` will be accessible both as
  ``wikipedia_en_all_2020-08`` and ``wikipedia_en_all``.

.. option:: -c PATH, --customIndex=PATH

  Override the welcome page with a custom HTML file.

.. option:: -L N, --ipConnectionLimit=N

  Max number of (concurrent) connections per IP (default: infinite,
  recommended: >= 6).

.. option:: -v, --verbose

  Print debug log to STDOUT.

.. option:: -V, --version

  Print the software version.

.. option:: -h, --help

  Print the help text.

HTTP API

kiwix-serve serves content at/under http://ADDR:PORT/ROOT where ADDR, PORT and ROOT are the values supplied to the :option:`--address`/:option:`-i`, :option:`--port`/:option:`-p` and :option:`--urlRootLocation`/:option:`-r` options, respectively.

HTTP API endpoints presented below are relative to that location, i.e. /foo/bar must be actually accessed as http://ADDR:PORT/ROOT/foo/bar.

Note

The HTTP API is documented in its entirety in order to facilitate the work of the Kiwix development team. Note, however, that only a subset of the HTTP API constitutes kiwix-serves's public interface.

Public API endpoint

A public HTTP API endpoint is intended to serve the outside world (in addition to kiwix-serve's front-end and other Kiwix products). The Kiwix development team will do its best to ensure gratifying experience for clients of public API endpoints at all stages of the endpoint lifecycle.

Private API endpoint

A private API endpoint is intended to be used only by kiwix-serve's frontend or by other products maintained solely by the Kiwix team. Private API comes without any guaranees. It may change as frequently and as drasticaly as the Kiwix development team sees fit.

Deprecation

Public API doesn't stay frozen once and forever. As the API evolves, Kiwix team reserves the right to drop support for certain old functionality. In such events, an advance notice will be issued and the users will be given enough time to prepare for the change.

Currently, public endpoints are limited to the following list:

• :ref:`OPDS API <new-opds-api>`
• /raw
• /search (with /search/searchdescription.xml)

/

Type:

:ref:`private <private-api-endpoint>`

Welcome page is served under /. By default this is the library page, where books are listed and can be looked up/filtered interactively. However, the welcome page can be overriden through the :option:`--customIndex`/:option:`-c` command line option of kiwix-serve.

/catalog/v2 (OPDS API)

Type:

:ref:`public <public-api-endpoint>`

The new OPDS API of kiwix-serve is based on the OPDS Catalog specification v1.2. All of its endpoints are grouped under /catalog/v2.

:ref:`Legacy OPDS API <legacy-opds-api>` is preserved for backward compatibility.

/catalog/v2/root.xml

Type:

member of a :ref:`public API <new-opds-api>`

The OPDS Catalog Root links to the OPDS acquisition and navigation feeds accessible through the other endpoints of the OPDS API.

/catalog/v2/searchdescription.xml

Type:

member of a :ref:`public API <new-opds-api>`

Describes the /catalog/v2/entries endpoint in OpenSearch description format.

/catalog/v2/categories

Type:

member of a :ref:`public API <new-opds-api>`

Returns the full list of ZIM file categories as an OPDS Navigation Feed.

/catalog/v2/entries

Type:

member of a :ref:`public API <new-opds-api>`

Returns a full or filtered list of ZIM files as a paginated OPDS acquisition feed with complete entries.

Pagination:

By default, no more than 10 first entries are returned from the library. To obtain the remaining entries the URL query parameters start and/or count must be used. The output of /catalog/v2/entries?start=s&count=n will contain at most n (default value: 10) results starting from entry # s (default value: 0). count with a negative value (e.g. count=-1) removes the limit on the number of results in the output.

Note

Previously count=0 also designated an unbounded query (i.e. worked similarly to count=-1). The response to a count=0 query was changed to consist of 0 results, as such a query/response combination is a good way to find out the total number of results (when only that information is needed) with minimal consumption of resources.

Examples:

# Returns the first 10 entries (internally numbered 0 through 9)
$ curl 'http://localhost:8080/catalog/v2/entries'

# Returns the next 10 entries (internally numbered 10 through 19)
$ curl 'http://localhost:8080/catalog/v2/entries?start=10'

# Returns the first 50 entries
$ curl 'http://localhost:8080/catalog/v2/entries?count=50'

# Returns 50 entries starting from entry # 100 (i.e. entries ## 100-149)
$ curl 'http://localhost:8080/catalog/v2/entries?start=100&count=50'

# Returns all entries
$ curl 'http://localhost:8080/catalog/v2/entries?count=-1'

# Returns all entries starting from entry # 100
$ curl 'http://localhost:8080/catalog/v2/entries?start=100&count=-1'

Filtering:

A filtered subset of the library can be requested by providing one or more filtering criteria, whereupon only entries matching all of the criteria are included in the response. Pagination is applied to the filtered list. The filtering criteria must be specified via the following URL parameters:

• lang - filter by language (specified as a 3-letter language code).
• category - filter by categories associated with the library entries.
• tag - filter by tags associated with the library entries. Multiple tags can be provided as a semicolon separated list (e.g tag=wikipedia;_videos:no). The result will contain only those entries that contain all of the requested tags.
• notag - filter out (exclude) entries with any of the specified tags (example - notag=ted;youtube).
• maxsize - include in the results only entries whose size (in bytes) doesn't exceed the provided value.
• q - include in the results only entries that contain the specified text in the title or description.
• name - include in the results only entries with a matching :term:`book name <Book name>`.

Examples:

# List only books in Italian (lang=ita) but
# return only results ## 100-149 (start=100&count=50)
$ curl 'http://localhost:8080/catalog/v2/entries?lang=ita&start=100&count=50'

# List only books with category of 'wikipedia' AND containing the word
# 'science' in the title or description. Return only the first 10 results.
$ curl 'http://localhost:8080/catalog/v2/entries?q=science&category=wikipedia'

/catalog/v2/entry/ZIMID

Type:

member of a :ref:`public API <new-opds-api>`

Returns full info about the library entry with :term:`UUID <ZIM UUID>` ZIMID.

/catalog/v2/illustration/ZIMID

Type:

member of a :ref:`public API <new-opds-api>`

Usage:

/catalog/v2/illustration/ZIMID?size=N

Returns the illustration of size NxN pixels for the library entry with :term:`UUID <ZIM UUID>` ZIMID.

If no illustration of requested size is found a HTTP 404 error is returned.

/catalog/v2/languages

Type:

member of a :ref:`public API <new-opds-api>`

Returns the full list of ZIM file languages as an OPDS Navigation Feed.

/catalog/v2/partial_entries

Type:

member of a :ref:`public API <new-opds-api>`

Returns the full or filtered list of ZIM files as an OPDS acquisition feed with partial entries.

Supported filters are the same as for the /catalog/v2/entries endpoint.

/catalog (Legacy OPDS API)

Type:

:ref:`deprecated <deprecation>`

The legacy OPDS API is preserved for backward compatibility and is deprecated. :ref:`New OPDS API <new-opds-api>` should be used instead.

/catalog/root.xml

Type:

member of a :ref:`deprecated API <legacy-opds-api>`

Full library OPDS catalog (list of all ZIM files).

/catalog/searchdescription.xml

Type:

member of a :ref:`deprecated API <legacy-opds-api>`

Describes the /catalog/search endpoint in OpenSearch description format.

/catalog/search

Type:

member of a :ref:`deprecated API <legacy-opds-api>`

Returns the list of ZIM files (in OPDS catalog format) matching the search/filtering criteria. Supported filters are the same as for the /catalog/v2/entries endpoint.

/catch/external

Type:

:ref:`private <private-api-endpoint>`

Usage:

/catch/external?source=URL

Generates a HTML page with a link leading to (the decoded version of) URL and a warning that following that link will load an external (out of ZIM) resource.

Parameters:

source: URL of the external resource (must be URL-encoded).

Example:

# Intercept an external link to https://example.com?query=abcd
$ curl 'http://localhost:8080/catch/external?source=https%3A%2F%2Fexample.com%3Fquery%3Dabcd'

/content

Type:

:ref:`private <private-api-endpoint>`

ZIM file content is served under the /content endpoint as described below.

/content/ZIMNAME/PATH/IN/ZIMFILE

Type:

:ref:`private <private-api-endpoint>`

Returns the entry with path PATH/IN/ZIMFILE from ZIM file with :term:`name <ZIM name>` ZIMNAME.

/content/ZIMNAME

Type:

:ref:`private <private-api-endpoint>`

/content/ZIMNAME redirects to the main page of the ZIM file with :term:`name <ZIM name>` ZIMNAME (unless that ZIM file contains an entry with an empty path or path equal to /, in which case that entry is returned).

/random

Type:

:ref:`private <private-api-endpoint>`

Usage:

/random?content=ZIMNAME

Generates a HTTP redirect to a randomly selected article/page from the specified ZIM file.

Parameters:

content: :term:`name of the ZIM file <ZIM name>`.

/raw

Type:

:ref:`public <public-api-endpoint>`

The /raw API provides access to ZIM file data. It consists of two separate endpoints for accessing data and metadata.

/raw/ZIMNAME/content/PATH/IN/ZIMFILE

Type:

member of a :ref:`public API <raw>`

Returns the entry with path PATH/IN/ZIMFILE from the ZIM file with :term:`name <ZIM name>` ZIMNAME. Currently, this endpoint almost duplicates (with some subtle technical differences) the newer endpoint /content/ZIMNAME/PATH/IN/ZIMFILE. The important difference is that the /raw endpoint guarantees that no server-side processing will be applied to the returned content, whereas content obtained via the /content endpoint may in the future undergo some processing intended to improve the operation of the viewer (e.g. compensating for certain bugs in ZIM creation). Also note that /raw is :ref:`public <public-api-endpoint>`, whereas /content is :ref:`private <private-api-endpoint>`.

/raw/ZIMNAME/meta/METADATAID

Type:

member of a :ref:`public API <raw>`

Returns the metadata item METADATAID from the ZIM file with :term:`name <ZIM name>` ZIMNAME.

/search

Type:

:ref:`public <public-api-endpoint>`

Performs a full text search on one or more ZIM files and returns an HTML page with a list of links to matching pages along with snippets of the matching portions of those pages.

A multi-ZIM search request must comply with the following constraints:

• the number of ZIM files participating in the search operation must not exceed the limit imposed by the :option:`--searchLimit` option of kiwix-serve.
• all of the ZIM files participating in the same search operation must be in the same language.

Parameters:

ZIM file selection parameters:

At least one of the following parameters must be provided in order to specify in which ZIM file(s) to search. Parameters appearing earlier in below list take precedence over subsequent ones (the later ones, even if present in the request, are simply ignored).

content: :term:`name of the ZIM file <ZIM name>` (for a single-ZIM search). This is a :ref:`legacy parameter <deprecation>`. books.name should be used instead.

books.id: :term:`UUID <ZIM UUID>` of the ZIM file. Can be repeated for a multi-ZIM search, however must respect the :ref:`multi-ZIM search constraints <multi-zim-search-constraints>`.

Note

If any of the provided books.id values refers to a book missing from the library then an error is returned instead of running the search on the remaining (valid) entries.

books.name: :term:`name of the ZIM file <ZIM name>` (not to be confused with books.filter.name which selects/filters based on the :term:`book name <Book name>`). Can be repeated for a multi-ZIM search, however must respect the :ref:`multi-ZIM search constraints <multi-zim-search-constraints>`.

Note

If any of the provided books.name values refers to a book missing from the library then an error is returned instead of running the search on the remaining (valid) entries.

books.filter.{criteria}: allows to take full advantage of :ref:`library filtering <library-filtering>` functionality of the /catalog/v2/entries endpoint ({criteria} must be replaced with an attribute/filtering criteria name supported by :ref:`library filtering <library-filtering>`).

Query parameters:

pattern (optional; defaults to an empty string): text to search for.

latitude, longitude & distance (optional): geospatial query parameters. If all of these are provided, then the results will be restricted to geotagged pages that are within distance metres from the location on Earth with coordinates latitude and longitude.

Pagination parameters:

pageLength (optional, default: 25): maximum number of search results in the response. Capped at 140.

start (optional, default: 1): this parameter enables pagination of results. The response will include up to pageLength results starting with entry # start from the full list of search results (the first result is assumed to have index 1).

Other parameters:

format (optional, default: html): format of the search results. Allowed values are: html, xml.

Examples:

# Search for 'android' in the book with name 'scifi-library'
# Return results ## 51-60.
$ curl 'http://localhost:8080/search?pattern=android&books.name=scifi-library&start=51&pageLength=10'

# Search for 'napoli' in books in Italian
$ curl 'http://localhost:8080/search?books.filter.lang=ita&pattern=napoli'

# Search for 'chateau' in books in French that have a category of 'wikipedia'.
# Return the results as XML.
$ curl 'http://localhost:8080/search?pattern=chateau&books.filter.lang=fra&books.filter.category=wikipedia&format=xml'

/search/searchdescription.xml

Type:

:ref:`public <public-api-endpoint>`

Describes the /search endpoint in OpenSearch description format.

/skin

Type:

:ref:`private <private-api-endpoint>`

Static front-end resources (such as CSS, javascript and images) are all grouped under /skin.

Usage:

/skin/PATH/TO/RESOURCE[?cacheid=CACHEID]

Cache busting of static resources is supported via the optional param cacheid. By default, i.e. when the cacheid parameter is not specified while accessing the /skin endpoint, static resources are served as if they were dynamic (i.e. could be different for an immediately repeated request). Specifying the cacheid parameter with a correct value (matching the value embedded in the kiwix-serve instance), makes the returned resource to be presented as immutable. However, if the value of the cacheid parameter mismatches then kiwix-serve responds with a 404 HTTP error.

kiwix-serve's default front-end (the :ref:`welcome page <welcome-page>` and the :ref:`ZIM file viewer <zim-file-viewer>`) access all underlying static resources by using explicit cacheid s.

/suggest

Type:

:ref:`private <private-api-endpoint>`

Usage:

/suggest?content=ZIMNAME[&term=QUERY][&count=N][&start=S]

Returns suggestions (in JSON format) for a text string that is assumed to be a partially typed search for a page inside a particular ZIM file.

Suggestions are obtained as matches of the query text against the page titles in the ZIM file using the title index database generated during the creation of the ZIM file.

In case of a multi-word query the order of the words matters in two ways:

1. the last word is considered as partially typed, unless followed by a space;
2. ranking of the matches.

If the ZIM file doesn't contain a title index then suggestions are generated by listing page titles starting exactly (in a case sensitive manner) with the query text. Otherwise, suggestions are case-insensitive.

If the ZIM file contains a full text search index, then an extra suggestion is added as an option to perform a full text search in the said ZIM file.

Parameters:

content (mandatory): :term:`name of the ZIM file <ZIM name>`.

term (optional; defaults to an empty string): query text.

count (optional, default: 10): maximum number of page suggestions in the response (i.e. the extra option to perform a full text search is not included in this count).

start (optional, default: 0): this parameter enables pagination of results. The response will include up to count entries starting with entry # start from the full list of page suggestions (the first result is assumed to have index 0).

Example:

$ curl 'http://localhost/suggest?content=stackoverflow_en&term=pyth&count=50'

/viewer

Type:

:ref:`private <private-api-endpoint>`

ZIM file viewer. The ZIM file and entry therein must be specified via the hash component of the URL as /viewer#ZIMNAME/PATH/IN/ZIMFILE.

/viewer_settings.js

Type:

:ref:`private <private-api-endpoint>`

Settings of the ZIM file viewer that are configurable via certain command line options of kiwix-serve (e.g. --nolibrarybutton).

/ANYTHING/ELSE

Type:

:ref:`private <private-api-endpoint>`

Any other URL is considered as an attempt to access ZIM file content using the legacy URL scheme and is redirected to /content/ANYTHING/ELSE.

Glossary

.. glossary::

  Book name

    Name of the book as specified in the ZIM file metadata (for a
    ``kiwix-serve`` started *WITHOUT* the :option:`--library` option) or the
    library XML file (for a ``kiwix-serve`` started with the
    :option:`--library` option).

    .. note::

      Two or more books may have the same name in the library. That's not
      considered a conflict, because there may be multiple versions of the
      "same" book (differing by the settings of the scraper, date, etc).
      :ref:`Library filtering <library-filtering>` by name will return all
      matching books.

  ZIM filename

    Name of a ZIM file on the server filesystem.

  ZIM name

    Identifier of a ZIM file in the server's library (used for referring to a
    particular ZIM file in requests).

    ZIM names are derived from the filenames as follows:

    - file extension is removed,
    - all characters are converted to lowercase,
    - diacritics are removed,
    - spaces are replaced with underscores,
    - ``+`` symbols are replaced with the text ``plus``.

    Presence of the :option:`-z`/:option:`--nodatealiases` option will create
    additional names (aliases) for filenames with dates.

    ZIM names are expected to be unique across the library. Any name conflicts
    (including those caused by the usage of the
    :option:`-z`/:option:`--nodatealiases` option) are reported on STDERR but,
    otherwise, are ignored (i.e. only one of the entries can be accessed via
    the conflicting name).

  ZIM title

    Title of a ZIM file. This can be any text (with whitespace). It is never
    used as a way of referring to a ZIM file.

  ZIM UUID

    This is a unique identifier of a ZIM file designated at its creation time
    and embedded in the ZIM file. Certain ``kiwix-serve`` operations may
    require that a ZIM file be referenced through its UUID rather than name.