Intelligent services

It's more the other way around than what I believe you are thinking: the intelligence services primarily provide suggestions for users in the client and only with the recent introduction of M-Files Discovery can the metadata tagging be automated. There is general information on the intelligence services in the Intelligent Metadata Layer Overview.

Thanks Joonas!

Intelligence services are at work in the background, for example, when the user drags and drops a new document to M-Files
or modifies a specific property value on the metadata card. In such instances, intelligence services analyze the contents of
new documents and metadata modifications and offer metadata suggestions based on the analyses that they conduct. See
the subsections below for a more specific description of how intelligence services produce metadata suggestions.

I suppose that answers my question if users get a benefit by bringing new documents manually.
Page 29 kind of scratches my other question if intelligent services can be used to automatically promote documents or if promotion needs always to be done manually. I suppose that selection of document class could be a problem. Can you confirm how promotion process would work with intelligent services or if it always need to be done manually by user?

Thanks.

Dejan

Intelligence services aren't directly related to the promotion process. These are two distinct things, although they can work together.

Promotion is the act of finding unmanaged content (content in other repositories) and augmenting it with metadata. This is typically a manual process. This metadata includes the class and any other properties that are either mandated by the class or that the user wishes to add. Once promoted, the file content continues to (primarily) live within the remote system but the object itself can be involved in M-Files processes such as workflows, notifications and assignments, have permissions applied to it, appear in views, etc.

An intelligence service can suggest metadata for a given file. When the file is added to the vault, the intelligence services run to analyse the file content and provide suggestions on the property values that should be on the object. The user can choose any set of these to apply to the new object. The intelligence service runs if the document is dragged into the vault via the desktop client, uploaded via the web or mobile interfaces, or whether it's promoted from an external repository. Intelligence services can also be run via the M-Files APIs so you can retrieve the suggestions for a given file via script, for example.

So, whilst intelligence services may be (and often are) kicked off by the promotion process, they are not directly related.

The concept with Discovery is that it can automatically trawl external repositories, locate content that it deems should be promoted, and kick off the process. Where this blurs the line between promotion and intelligence services is that this can be used to promote content automatically without any user interaction, working with Smart Classifier or Information Extractor to identify content with PII, for example.

Regards,

Craig.

Hi Craig,

Thank you for this detailed explanation.

Let me try to summarize shortly:
Promotion is mostly manual process (user needs to select the class). But when document is manually made managed, intelligent services would kick in and basically provide metadata suggestions. Specifically Information Extractors (Matcher & Text Analytics) and would suggest metadata for selected class. User would accept suggestions or slightly change it. That would help users with promotion process.
Discovery is more aggressive approach and would promote documents directly to the vault without really making suggestions and waiting that users accept them.

I suppose a key point is that promotion is a manual process if we don't want to use discovery which automatically classify and promote objects.

Let me know if I am correct with my understanding.

one question though: is it possible to assign automatically non-managed documents to certain users? The idea is to create assignment for specific users to promote documents to the vault. We were searching in the API if we can differentiate between managed and non-managed documents but this information seems not be correctly filled.

Thanks.

Dejan

Promotion is primary a manual process, yes.

One clarification I'd make is that, when working with documents in general, intelligence services could suggest the class as well. Think of dragging a file into M-Files; it could suggest it here. But for unmanaged documents they're typically invoked during the promotion process, so you're correct in that regard.

is it possible to assign automatically non-managed documents to certain users?

Kind of. Perhaps. You can create an assignment object for anything, remember, not just assigning documents or other content. So, whilst you could not assign an unmanaged object to anyone (it's unmanaged; M-Files doesn't know anything about it), you could perhaps create an assignment and include in it a link to the unmanaged object for them to follow.

Regards,

Craig.

Thanks Craig.

The only question is how to find out that document is un-managed in M-Files. For users would be important to have a link to that "un-managed" object in M-files where user can select the class manually. As I mentioned we have tried the API but could not find out "un-managed" documents. But perhaps, this is by design because those documents are not seen in the vault yet. Still wondering why user could see un-managed document in the client but this information is not accessible over API. If this would be seen through API, this part could be automated and we could create assignments for those specific un-managed documents.

But definitely quite clear now.

Best regards,
Dejan

The difference between managed an unmanaged documents lies in the fact that managed documents have been registered as objects in the database whereas unmanaged objects are only indexed. They exist in the index but not in the database. Users can see them in their user interface because Views are configured to show indexed documents even though they are not objects in the database. If the API only works on the database it will not be able to show documents that only exist in the index.

Hi Dejan,

There are only two ways to "find" unmanaged content via the M-Files Desktop client, currently:

• Via a full-text-index search (returning both managed and unmanaged content).

• Navigating via the "external views" (returning managed and unmanaged content that's exposed by the connector in each view).

The same is true of the content via the API: it is currently possible to execute a full-text search that will return both managed and unmanaged content, but it's not possible to limit the search to only show unmanaged content. There is an improvement request open inside our tracker asking that this is added but, as the use-cases are fairly edge, it's not something that's on our current development backlog.

Regards,

Craig.

Thank you both for answers. It would be really usuful to differentiate between objects or at least to group them separately. One view containing unmanaged objects would potentially give enough options.

I understand the request. However the fundamental difference between an unmanaged and a managed object is, as Bright Ideas said, unmanaged objects have no data aside from entries in the full-text index. Because of this they can never appear in M-Files internal views. If you want them to appear in M-Files views then you must promote them to managed objects.

Unmanaged objects appear in each external repository's external views (i.e. the pseudo-views exposed by the Network Folder Connector that correspond to folders in the network share).

Regards,

Craig.