Package com.composum.ai.backend.slingbase.impl

Class HtmlToApproximateMarkdownServicePlugin

java.lang.Object

com.composum.ai.backend.slingbase.impl.HtmlToApproximateMarkdownServicePlugin

All Implemented Interfaces:

ApproximateMarkdownServicePlugin

public class HtmlToApproximateMarkdownServicePlugin
extends Object
implements ApproximateMarkdownServicePlugin

A plugin for the ApproximateMarkdownService that transforms the rendered HTML to markdown. That doesn't work for all components, but might more easily capture the text content of certain components than trying to guess it from the JCR representation, as is the default.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
protected static class	HtmlToApproximateMarkdownServicePlugin.CapturingResponse	We wrap a response to capture the content, forwarding all but modifying methods to the original response.
protected static interface	HtmlToApproximateMarkdownServicePlugin.Config
protected static class	HtmlToApproximateMarkdownServicePlugin.EmptyRequestParameterMap
protected class	HtmlToApproximateMarkdownServicePlugin.NonModifyingRequestWrapper	Wraps the request to make sure nothing is modified.
protected static class	HtmlToApproximateMarkdownServicePlugin.UnsupportedOperationCalled	Thrown when unsupported operation was called that requires blacklisting.

Nested classes/interfaces inherited from interface com.composum.ai.backend.slingbase.ApproximateMarkdownServicePlugin

ApproximateMarkdownServicePlugin.PluginResult

Field Summary

Fields

Modifier and Type	Field	Description
protected Pattern	allowedResourceTypePattern
protected Map<String,Long>	blacklistedResourceType	ResourceTypes we ignore since their rendering uses unsupported methods.
protected Long	blacklistedResourceTypeCleanupTime
protected Pattern	deniedResourceTypePattern

Constructor Summary

Constructors

Constructor	Description
HtmlToApproximateMarkdownServicePlugin()

Method Summary

Modifier and Type	Method	Description
protected void	activate(HtmlToApproximateMarkdownServicePlugin.Config config)
protected void	cleanupBlacklist()
protected void	deactivate()
@Nullable String	getImageUrl(@Nullable org.apache.sling.api.resource.Resource imageResource)	Retrieves the imageURL in a way useable for ChatGPT - usually data:image/jpeg;base64,{base64_image} If the plugin cannot handle this resource, it should return null.
protected boolean	isBecauseOfUnsupportedOperation(Throwable e)
protected boolean	isIgnoredNode(org.apache.sling.api.resource.Resource resource)	We start with depth 3 since the higher nodes often contain headers, navigation and such that don't help for ChatGPT.
@NotNull ApproximateMarkdownServicePlugin.PluginResult	maybeHandle(@NotNull org.apache.sling.api.resource.Resource resource, @NotNull PrintWriter out, @NotNull ApproximateMarkdownService service, org.apache.sling.api.SlingHttpServletRequest request, org.apache.sling.api.SlingHttpServletResponse response)	Checks whether the resource should be handled by this plugin and if so, handles it by printing an appropriate markdown representation to the PrintWriter.
protected String	renderedAsHTML(org.apache.sling.api.resource.Resource resource, org.apache.sling.api.SlingHttpServletRequest request, org.apache.sling.api.SlingHttpServletResponse response)	We render the resource into a mock response and capture and return the generated HTML.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface com.composum.ai.backend.slingbase.ApproximateMarkdownServicePlugin

cacheMarkdown, getMasterLinks, resourceRendersAsComponentMatching

Field Detail

allowedResourceTypePattern

protected Pattern allowedResourceTypePattern

deniedResourceTypePattern

protected Pattern deniedResourceTypePattern

blacklistedResourceType

protected Map<String,Long> blacklistedResourceType

ResourceTypes we ignore since their rendering uses unsupported methods. Blacklisting for only 1h since there might be a deployment in the meantime. Maps the resource type to the time (ms) until it is blacklisted.

blacklistedResourceTypeCleanupTime

protected volatile Long blacklistedResourceTypeCleanupTime

Constructor Detail

HtmlToApproximateMarkdownServicePlugin

public HtmlToApproximateMarkdownServicePlugin()

Method Detail

maybeHandle

@NotNull
public @NotNull ApproximateMarkdownServicePlugin.PluginResult maybeHandle(@NotNull
                                                                          @NotNull org.apache.sling.api.resource.Resource resource,
                                                                          @NotNull
                                                                          @NotNull PrintWriter out,
                                                                          @NotNull
                                                                          @NotNull ApproximateMarkdownService service,
                                                                          @Nonnull
                                                                          org.apache.sling.api.SlingHttpServletRequest request,
                                                                          @Nonnull
                                                                          org.apache.sling.api.SlingHttpServletResponse response)

Description copied from interface: ApproximateMarkdownServicePlugin

Checks whether the resource should be handled by this plugin and if so, handles it by printing an appropriate markdown representation to the PrintWriter.

Specified by:

maybeHandle in interface ApproximateMarkdownServicePlugin

Returns:

what is already handled by this plugin. It is possible to write to the PrintWriter in any case.

cleanupBlacklist

protected void cleanupBlacklist()

getImageUrl

@Nullable
public @Nullable String getImageUrl(@Nullable
                                    @Nullable org.apache.sling.api.resource.Resource imageResource)

Description copied from interface: ApproximateMarkdownServicePlugin

Retrieves the imageURL in a way useable for ChatGPT - usually data:image/jpeg;base64,{base64_image} If the plugin cannot handle this resource, it should return null.

Specified by:

getImageUrl in interface ApproximateMarkdownServicePlugin

isBecauseOfUnsupportedOperation

protected boolean isBecauseOfUnsupportedOperation(Throwable e)

isIgnoredNode

protected boolean isIgnoredNode(@Nonnull
                                org.apache.sling.api.resource.Resource resource)

We start with depth 3 since the higher nodes often contain headers, navigation and such that don't help for ChatGPT.

renderedAsHTML

protected String renderedAsHTML(org.apache.sling.api.resource.Resource resource,
                                org.apache.sling.api.SlingHttpServletRequest request,
                                org.apache.sling.api.SlingHttpServletResponse response)
                         throws javax.servlet.ServletException,
                                IOException

We render the resource into a mock response and capture and return the generated HTML. The response is wrapped so that the real response cannot be modified. We don't do that for the request, because that would be more complicated and probably not needed.

Throws:

javax.servlet.ServletException

IOException

activate

protected void activate(HtmlToApproximateMarkdownServicePlugin.Config config)

deactivate

protected void deactivate()