The site that you are currently viewing is a static version of the Software Factory documentation delivered with the

version v7.2.
For up-to-date documentation, see the latest version.

FAQ

Various questions (and answers) about the Code Companion component.

Table of Contents

General Information

What is the price?
  • The Code Companion costs only 30€ per user per month with unlimited requests and tokens. Only the systems limits apply. Once your TDF Account is created, the cost will be charged to your entity, like any other license or platform cost.
    You can unsubscribe whenever you want on PostIT.
Is there a trial period?
Why OpenAI? Why not Mistral or others?
  • OpenAI has the advantage to be available on Azure, meaning it is served as a service and scale. The IS/IT is working on the deployment of Mistral models on premise.
What are the supported languages?
  • Code Companion is efficient in mainstream languages, particularly in Python, but also in JavaScript, Java, Ada, C, C++, C#, Go, Perl, PHP, Ruby, Swift, TypeScript, Shell, and more. And you can talk to Code Companion in your native language: French, English, German, Spanish or any other.
Is Code Companion searching on the internet?
  • No, the Code Companion does not search and give result based on live internet information.
What is the difference between Hermes, Code Companion and GitHub Copilot?
  • Hermes is the previous name for Code Companion, so Hermes = Code Companion.
    GitHub Copilot or any other solution using public models can’t be used with C2 data. Only open source C1 data.

Setup and Installation

What IDE can I use?
  • Only VSCode and JetBrains IDE (IntelliJ, PyCharm) are supported for now. For others (Visual Studio, Eclipse), it’s in the backlog but we’re not working on it for the moment. And nothing is planned for vi, emacs, NetBeans and notepad++.
Is it possible to run it locally (no internet connection)?
  • It is not yet supported outside of TDP.
Are the updates automatically done?
  • No, you will need to get the latest manually through our communications since we’re not in the marketplace

Security and Data Sensitivity

What data sensitivity (C2, C3-CA, C3…) can be used on code companion? Is export controlled allowed?
  • C2 data is supported.
  • C3-CA (Cloud Authorized - in Azure) is supported.
  • C3-BS (Business Sensitive - on premise) is entering the Early Adoption phase.
  • Export controlled data are not allowed.

Code Companion is the only solution validated for Thales Data.

Are the prompts sent to OpenAI?
  • The models we use are not trained on your data nor sent back to OpenAI. Additionally, our team implements additional security measures.
Who owns the generated code intellectual property? Is it plagiarism?
  • Generated code Intellectual Property rights are protected by our contract with Microsoft.
    Additionally, it has a generated code verification filter to ensure that the code is not proprietary. In any case, any source code must be scanned by Software Composition Analysis (SCA). The Software Factory provides GitLab SCA.
May we know the list of forbidden words?
  • Due to security rules we’re not allowed to share it.
Can I safely open a C3 data/source while Code Companion is installed but I don’t interact with the extension in any way?
  • Code Companion v2 does codebase indexing and autocomplete which will be active whenever a project is opened in VSCode, regardless of whether there’s interaction with the extension or not. The codebase indexing generates embeddings for codebase retrieval and stores the index locally, while autocomplete automatically determines context based on the current cursor position. However, for C3 files (a data level not yet supported by our product), it is highly recommended to disable Code Companion’s autocomplete or to split your C3 projects into a separate workspace, if feasible.
Can I still use the chat as a general chatbot (without providing code context or using quick actions) while in a C3 workspace?
  • You can use the chat feature as a general chatbot since it does not interact with your files unless explicitly specified through actions, commands, context, etc. Nevertheless, it is advisable to turn off autocomplete when dealing with any data level beyond our supported recommendations or to separate your C3 projects into a distinct workspace for better management and compatibility.

Model Information

What is the model? When is it updated? Can I choose the model or use my own?
  • Code Companion V2.x uses gpt-4o.
    For model updates, we rely on the Microsoft’s roadmap. New models usually arrive a few months after their public release. Selecting the model is not yet possible.
Is it trained to apply Thales guidelines or industry standards?
  • No, it is currently a standard Azure OpenAI model, except that it has been configured and secured to handle C3-CA sensitivity data.
    It may know about some industry standards, but you cannot expect the answer to be 100% accurate, always verify.