IntelliCode team completions: AI-assisted IntelliSense based on your code

  • Article

Use IntelliCode team models for completions to get AI-assisted IntelliSense recommendations based on your C# and C++ code bases. Team completions are useful when working with your own types or domain-specific libraries that aren’t commonly used in open-source code, because IntelliCode’s base model recommendations are based solely on patterns learned from open-source GitHub repos. If you're working on code that isn’t in that set of open-source repos, the base recommendations won't be useful to you. If you're writing C# and C++ code in Visual Studio, use IntelliCode to learn patterns from their code to make recommendations tailored to your code.

IntelliCode models are an encapsulation of a set of rules that let you predict some useful information (for example, recommendations in the IntelliSense list) based on an analysis of code. IntelliCode creates team models using the same learning process as for the IntelliCode base models, except they're trained on your own code. The more code you provide to illustrate your patterns of usage, the more capable your team model will be at offering useful recommendations.

To build your team model, we extract a summary file with metadata on your types and their usages and securely upload it to our service.

How models get applied

IntelliCode generates its recommendations from multiple models by merging together:

  • The base model for the language you're using (which is trained on thousands of public GitHub repos).
  • Any team models you've trained.
  • Any team models that are associated with the Git repository you’re working in.

You don't need to manage which models apply to which solution or codebase because IntelliCode takes care of this.

Types of team completions models

There are two ways you can obtain team completions models:

  1. Repository-associated: Models are tied to the repository. All users who can clone and edit the repository are granted automatic access to the model. Your codebase must be under Git source control and pushed to a remote using Azure Pipelines Task or GitHub Action from CI build to create a repository-associated model.

  2. Machine-associated: Models are available only on the machine they're trained upon.

Repository-associated team models

Repository-associated team models are available to users who train them using either Azure Pipelines or GitHub Actions.

Sharing your repository-associated models

Repository-associated models are automatically shared with other users working in the same codebase and have enabled automatic acquisition of team models in Visual Studio. Enable automatic acquisition by selecting Tools > Options > IntelliCode > Acquire team models for completion.

When you clone and open codebase model, which was trained on, any models associated with the configured Git remote repositories are downloaded and activated. If you're working on a fork of the codebase, add the upstream codebase as a remote repository to get the model.

If you have access to the repository, you'll also have access to the model. When training, we collect some information about the checked-out commit. If you request the model, then you must have the same commit in the repository and produce the same information, which was collected during training to receive the team model.

Note

The ability to share your user-associated team completions models via a sharing link, available in some preview releases of Team completions, is now deprecated.

Delete your model

You can remove models from your account so they can no longer be used.

Delete a user-associated model created within Visual Studio

While your solution is open in Visual Studio, uncheck the checkbox accepting model training, on the View > Other Windows > IntelliCode page. The model is deleted.

Delete a repository-attached model created using Azure DevOps or a GitHub CI workflow

Remove the training task from your pipeline, the associated model is removed within 30 days if not trained.

Machine-associated models

Create and retrain your model

To train a machine-associated model:

  1. Open the project or solution in Visual Studio.
  2. Open the IntelliCode page by selecting View > Other Windows > IntelliCode.
  3. Review and accept the license terms checkbox at the bottom of the page. A machine-associated model is trained automatically.

Note

You must repeat the above steps for each solution you want to train on.

Visual Studio will automatically retrain your machine-associated team completions models, periodically.

Train on a public codebase

Before you train on your own code, you might want to create a completions model on a public codebase. You can see how the completions model affects IntelliSense, or if you're concerned about the kind of data that IntelliSense collects, you can inspect the extracted data. Some interesting samples to train on are:

  • Azure ConferenceBuddy

    Fork the repo to your personal account, clone the repo, open the ConferenceBuddy.sln solution, build to check that it's working, and then train the model. You'll find some good completions on instances of the AskWhoTask class.

  • Windows RSS reader

    Fork the repo to your personal account, clone the repo, open the RssReader.sln solution, build to check that it's working, and then train the model. You'll find some good completions on instances of the MainViewModel class.

Autotraining models for your codebase

After successfully building a solution, you might be prompted to enable IntelliCode to autotrain a model for IntelliCode completions for that solution.

By enabling autotraining models for IntelliCode completions, IntelliCode will train a machine-learning model for completions for the active solution and only the user who has access to the solution on the machine where the autotraining was enabled will have access to the respective model. If you'd like to share your custom code completions with anyone who can access your repository, you should set up automatic training Team completions as part of your CI workflow.

Note

For autotraining a model for IntelliCode custom code completions for your solution in Visual Studio, there are no source control requirements. However, if you'd like to share your custom completions with your team, your codebase must be under Git source control and pushed to a remote to create a repository-associated model.

Enabling autotraining models for custom code completions in Visual Studio

To enable automatic model training for IntelliCode completions for your code in Visual Studio:

  1. Open the solution or repository folder in Visual Studio.

  2. By enabling autotraining a model for custom code Completions via the infobar prompt after a successful build, through the IntelliCode UI, by selecting View > Other windows > IntelliCode, or through Tools > Options > IntelliCode setting "Autotraining team models for IntelliSense completions" or by searching for "IntelliCode autotrain" in Visual Studio Search (Ctrl + Q).

  3. Upon successful creation of the model, the model is automatically downloaded to Visual Studio. You can track the model’s progress by opening the Output Window and switching to IntelliCode in the dropdown OR in View > Other windows > IntelliCode.

Note

Be sure that you've installed Visual Studio version 16.7 Preview 3 or above. Once the preview has been installed, you can enable automatic models for custom code completions through the infobar after a successful solution build OR via Tools > Options > IntelliCode.

Once the training is complete, try writing some code using the classes/types that are particular to your repo - you should see starred suggestions for the most common cases.

Once you're done with the custom code completions on your solution, you can automatically create, retrain, and share IntelliCode custom code completions with your entire dev team as part of your continuous integration pipeline in Azure Pipelines or GitHub Action for Team Completions.

Data and privacy

To build your team model, we extract a summary file with metadata on your types and their usages. For example, the summary file contains the names of classes and methods and how often they're called in different circumstances. IntelliCode doesn't track your keystrokes or extract whole expressions, statements, or literal values (such as strings) from your code.

The extracted data is transmitted over HTTPS to the IntelliCode service. The service then uses machine-learning algorithms to train a model for your code. It returns the model to your computer where it's merged with the base model.

When you enable IntelliCode to kickoff training and autotraining your model for custom code completions:

  • We analyze your code locally.
  • We extract a summary file with metadata on your types and their usages.
  • We securely upload it to the IntelliCode service and train a completions model tailored to your code.
  • Your completions model is never shared with the users who have access to your repo (if a cloned git repository).
  • You can delete your model at any time and cancel the model training directly in the IntelliCode UI at View > Other windows > IntelliCode. Uncheck the box in the UI to delete the model for the custom code completions.
  • You can refresh your learned patterns directly in the IntelliCode UI found at View > Other windows > IntelliCode.

You'll see the training progress in your Visual Studio output window's IntelliCode section or IntelliCode UI at View > Other windows > IntelliCode. Once the training is complete, you can see your summary and the new model tailored to your code. Then, try writing some code using the classes/types that are particular to your repo. You should see starred suggestions for the most common cases.

View extracted data

To inspect the contents of the extracted data:

  1. Open the extracted data directory:

    • For repository-associated models: %temp%\Intellicode_Extraction_2019-10-23—234524
    • For machine-associated models: %TEMP%\Visual Studio IntelliCode

  2. To find and open the training for your most recent training session, sort the folder view by date (descending). The folder for your most recent training session is now at the top.

    Tip

    There's one folder per training session in the %TEMP%\Visual Studio IntelliCode directory, each with a randomized name.

The folder contains the entire set of files that are sent to Microsoft when extraction is complete. The UsageOutput subfolder contains a JSON file that has the information, which IntelliCode extracted from your code to train the model. The UsageOutput_ErrorStats file contains any errors found when trying to build the extracted file and can help if Microsoft needs to debug issues.

IntelliCode model-training directory

If you want to inspect the extracted data for a different codebase before trying it on your own code, train a model on a public codebase.

How we secure your data

All data you send to and receive from the IntelliCode service is transmitted over HTTPS. You must sign in to Visual Studio to communicate with the service.

Models can be retrieved by:

  • Machine-associated models: the machine, which submits the extracted data for training.
  • Repo-associated models: users who can prove they have access to the repository for repository-associated models.

Your model and what's learned about your code stays private to you and your intended collaborators.

If Microsoft needs to troubleshoot, authorized Microsoft service personnel might be granted access to your models and extracted data for diagnostic purposes only.

See also