The Prgatix code assistant allows autocomplete code and chat with your code base using AI based on a private/on-prem model.

Code Assistant components

Cloud-Based Models

• GPT-40 for chat

• Codestral (by Mistral) for autocomplete

On-prem models

• qwen2.5-coder:1.5b-base for autocomplete

• openai/gpt-oss-20b for chat

Transformers.js for the local embeddings of the code.

Here you can find some examples of questions you can ask:

https://docs.continue.dev/customize/deep-dives/codebase

The continue is under Apache 2.0 license
​https://github.com/continuedev/continue/blob/main/LICENSE

Installations guide:

Extension installation:

Download the extension VSIX file from here

Go to VS Code and click on the extension in the left sidebar.

Click the three dots (where the arrow points).

Select "Install from VSIX" and choose the appropriate continue file.

Chose the file you have downloaded

You should see the Prgatix add-on in the list after you finish as here:

Code Assistant Configuration:

For Cloud-Based models configuration is almost ready to use.

• Add API keys to the relevant placeholders in the config file (two places one for OpenAI and one for Mistral).

• Please contact AGAT for the demo API Keys

For on-premises models configuration, download the following config file:

• Click the Pragatix extension icon on the left sidebar.

• In the main view, press the Edit models configuration button (you’ll see this only the first time).

  

• Replace the config.yaml content with the downloaded on-prem_config.yaml file content.

Updating an Existing Configuration

• Click the Pragatix icon in the left sidebar.

• In the top chat area, click the dropdown menu.

• Click Settings (gear icon) to edit the current model

  

Examples

After installing the extension, you’ll see a demo file that walks you through how to use it.

Here are some other example prompts and tasks to help you try Pragatix Code Assistant.

Autocomplete Example (Code Generation)

Start writing a Python function, and the model will complete it.

1. Create a new file (e.g., prime.py).
   You can begin typing your instruction, starting with the # sign.

   1. Example 1: #create a function to check if a given number is prime

   2. Example 2:
      #create a tax calculator program in Python that calculates monthly tax based on annual income. If the annual income is less than $50,000,apply 25% tax. If the annual income is above $50,000 apply 35%

3. Wait a couple of seconds — the model will automatically suggest completions.

   

File Edit via Chat

Use chat to modify an existing code.

1. Select the code from the Tax example.

2. Press Ctrl+I to open the Prgatix chat box.

3. 
   

   Update the code, add an argument number_of_childrens and change the tax rate based on it
   

Prgatix will modify the open file according to your request.

Debugging with Chat

Identify and fix an error in a file.

1. Add a piece of faulty code, for example:

   if x = 5:
       print("x is 5")
   

   (Here, = should be ==.)

2. Open the chat box and type @, then select the relevant file from the list.

3. Ask:

   What's wrong in this file?
   

Prgatix will analyze the file and point out the error.