AI Assistant is now even more deeply integrated into Jupyter notebooks in PyCharm. You now have the option to add a prompt directly inside your notebook and work with AI Assistant right from there. Just choose the AI cell option that appears above or below each cell, and a cell with a field that’s ready for your prompt will appear.

If you get stuck while performing data analysis and would like a hint about the most common steps that other data analysts usually perform at that stage, click on the bulb on the right side of the AI cell. AI Assistant will provide the most relevant suggestions based on the notebook’s current context so that you can clean your data faster and more efficiently.

If you would like AI Assistant to help you iterate on the code of a particular cell, use the AI Edit Cell option from the context menu. This will open the field for the prompt directly in the selected cell. Changes suggested by AI Assistant will be shown in a diff view so that you can see how your code will be changed.

Visualize your dataframes with the help of AI Assistant, which can now provide suggestions about the graphs and plots most suitable to your context.

Use the AI icon in the upper right-hand corner of the dataframe and select Suggest Visualization. PyCharm will then open an AI chat where you’ll receive suggestions. Choose the one you like best, and AI Assistant will generate a new cell with the appropriate code snippet in the notebook. Run it, and voilà! The graph will appear in your notebook.

PyCharm now provides direct integration with Databricks. You can connect to a Databricks cluster, execute scripts and notebooks as workflows, execute files directly in a Spark shell on a cluster, and monitor the progress – all from the comfort of your IDE.

PyCharm can detect Databricks configuration files on your disk and use them to connect to Databricks clusters. You can run code in Databricks in two modes: Run on Cluster or Run as a Workflow.

To use the Databricks integration, you must install this dedicated plugin.

PyCharm can now suggest a list of relevant Hugging Face models that best suit your needs. This allows you to decide which model to use and install it without ever having to leave the IDE. Right-click in the editor tab to call the context menu, choose Insert HF Model, and identify what you would like the model to do. PyCharm will provide a list of suggestions! After you select a model, PyCharm will suggest inserting a code snippet that allows you to use this model directly in the open file, and it will download and install all the missing packages on which this model depends.

You can now inspect your Hugging Face Datasets library data as an interactive dataframe. This allows you to take advantage of the features you already use for pandas, Polars, PyTorch, and TensorFlow dataframes, including the chart view, pagination, and the ability to sort and export tables.

A new dedicated Hugging Face tool window allows you to monitor and manage Hugging Face models stored on your machine. This will help you identify unused models that can be deleted to free up disk space, in addition to making it easy to copy the paths to models or open the folders where they are stored.

You no longer need to start the debugger or use a print statement if you just want to check variable values. You can now get an instant preview of the values of a chosen variable simply by putting the cursor on the variable’s line.

The Endpoints tool window recognizes routes specified via the @action decorator for viewsets. You can see the routes and HTTP methods specified in the methods parameter. Further, PyCharm now generates routes based on the detail and url_path parameters in addition to the initial route assigned to the viewset. The HTTP Client will suggest a request with the relevant methods for each route.

The Search Everywhere dialog now includes an option to preview the codebase elements you’re searching for. Enabling this feature through the Preview icon on the dialog’s toolbar will cause a preview pane to appear under the search results, offering additional context and making it easier to navigate through your project.

In version 2024.2, we’ve made the process of accepting full line code completion suggestions more intuitive and precise. Inline code completion suggestions now include code highlighting, and new shortcuts allow you to accept individual words or entire lines from longer suggestions. We’ve also refined how accepted changes are integrated into your code, eliminating formatting and related issues.

PyCharm can now natively render mathematical expressions in Markdown files. When working with Markdown, you can use $ to insert inline math expressions and $$ for code blocks with math content.

You can now choose the languages you want sticky lines to appear for. You can tailor the feature to your preferences either in Settings/Preferences | Editor | General | Sticky Lines or by calling the context menu with a right-click on a sticky line in the editor.

Debugging and browsing long string variables with complex data formats is now much easier. The updated debugger offers properly formatted visualizations for string variables with strings encoded in JSON, XML, HTML, JWT, and URL. Simply click the View hyperlink next to a variable to quickly identify the root cause of your bug. The relevant visualizer is automatically chosen based on the contents of your variable.

We’ve given tags a dedicated node in the Version Control widget, making it easier to check them out. The node is displayed by default, but if you prefer to hide it, simply click the gear icon and uncheck the Show Tags option.

We’ve introduced valuable updates to the HTTP Client. The new XPath function allows you to query elements directly from XML and HTML documents. Additionally, we’ve added numerous methods for manipulating XML and HTML documents, eliminating the need to parse XML manually.

The HTTP Client now also supports iterating through variable collections, allowing you to compose a single HTTP request with a JSONPath collection and automatically send requests for each element in the collection. Furthermore, you can access specific collection items in post-processing scripts using JSONPath syntax.

Additionally, we’ve implemented the ability to create custom methods for working with APIs that provide specific methods. You can now add custom methods on the fly with a quick-fix or enter them with a comma-separated list in Settings | Tools | HTTP Client.

We’ve significantly extended PyCharm’s coding assistance for Terraform by making full line code completion available for it. Additionally, essential code insight features such as context-aware code completion, refined syntax highlighting, and the error detection system with quick-fix suggestions have been improved for Terraform. We've also added a quick documentation feature that displays tooltips when you hover over elements, providing instant information to help you use Terraform resources more effectively without interrupting your workflow.

In this release, we’re introducing new AI Assistant features for databases. With the text-to-SQL feature, you can generate SQL code directly in the editor by clicking Generate Code with AI and entering your prompt. You can accept, regenerate, or refine the code, and take it to the AI chat if you have further questions.

Additionally, AI Assistant can help with modifying tables, allowing you to request changes like switching all VARCHAR data types to CHAR. It can also help you understand and fix SQL problems, suggesting explanations and fixes. In addition, you can attach your database schema to the AI chat by mentioning it in your prompt.

You can now import data from tabular files using customizable scripted loaders for the Excel, JSON, Parquet, and shapefile formats. Drag your file from the Project tool window to the schema in the database explorer and configure the settings in the Import dialog.

In addition, you can view these files in the data editor with a single click in the Project tool window. To disable this behavior, go to Settings/Preferences | Advanced Settings | Open file as table if detected by scripted loader.

You can also create custom loaders by placing them in Scratches and Consoles | Extensions | Database Tools and SQL | data | loaders.

You can now run and debug TypeScript files without any additional setup. You can execute them from different entry points, including the file context menu, the Run widget, and the Current File configuration.

PyCharm 2024.2 is able to resolve paths for frameworks that use file-system-based routing. It can resolve link paths based on your project's file system, providing autocompletion and navigation for Next.js, Nuxt, SvelteKit, and Astro. There is also support for new Svelte 5 snippets and render tags. Additionally, we’ve implemented language server protocol (LSP) support for Astro and Vue Language Service v2, improving code completion and the overall developer experience.