Full Line Code Completion

Compatible with IntelliJ IDEA Ultimate, DataSpell and 5 more
Screenshot 2

Designed to supercharge your code completion with multi-token proposals utilizing a deep learning model. Enjoy the benefits of advanced code completion offline, as it runs on your local device without sending anything over the internet.


Language support

The plugin integrates C#, Python, Kotlin, Java, JavaScript/TypeScript, their frameworks (React, Angular, Vue), and CSS/LESS/SCSS/SASS, Go, PHP, Ruby.

Note that from the 2024.1 release, Full Line is bundled to most of the IDEs and there is no need to download it from the marketplace.

For Rider and CLion - Full Line is bundled to 2024.2 release


We value your feedback

You can upvote existing issues in our YouTrack or log in and create a new one.


Get started in moments

After installation, the plugin immediately enables you to enjoy enriched inline suggestions for code completion:

Accept suggestions with the Tab key. Customize the shortcut to your preferences via the hover menu over the gray text.

Customize your settings

Navigate to Preferences | Editor | General | Code Completion | Full Line Code Completion to adjust your settings.

FAQ

- Do you send my private data or code over the internet?
- We prioritize the privacy of your data. The Full Line Code Completion plugin works on your local device, ensuring no data is sent over the internet.
- What are the plans for supporting additional languages?
- The 2024.2 plan is to add support for C++, C#, and Rust.
- What are the hardware requirements for the plugin?
- It is fully compatible with Linux, Mac, and Windows operating systems and takes less than 200 MB of disk space. Our model is executed entirely on the CPU with the ability to run on GPU on Macs, automatically adapting to your device's capabilities to ensure seamless operation.

What’s New

Full Line

241.18034.76

Special Rider 2024.1 release!

Added

  • New 100M code completion model for C# with rewritten code correctness checks and improved UX!
  • Improved quality for JS/TS and CSS languages
  • Improved suffix handling quality

233.14732

Added

  • New 100M (previously it was 20M) model for Python with better quality and performance
  • Add new inline completion mode (experimental)

Changed

  • Tuned generation for a bigger python model
  • Improve local generation for all models
  • Simplify settings panel and make it clearer
  • Changed maximum shown proposals to 2 (by default, was 3)
  • Feedback form is now know about most popular language
  • Replaced icon in Full Line diagnostics tab
  • Enhance insertion handler by matching brackets and quotes after caret
  • Updated plugin description

Removed

  • JS/TS are temporary hidden to improve speed, quality and support more dialects

Fixed

  • Fix cases with incorrect brackets and quotes balance in string with curly brackets
  • Fix showing Full Line in string interpolation
  • Fix issue with downloading all models, even for disabled languages
  • Fix updating language settings in preferences (case when settings keep rollback to default one)

233.5200

Added

  • New cosmetics (logo and completion lookup icons)
  • Smart shutdown based on user machine
  • Optional feedback form
  • New smart insertion matching
  • Model now working via native server, not jvm
  • Highly accelerated generation of variants
  • Updated local models for all languages
  • Improved lookup showtime and cases with long suggestion generation

Changed

  • Suggestions now will show only in specific places
  • Disabled first token insertion (tab)
  • Turned off string walking by default

233.2912

Added

  • Use minimal formatting for context so proposals are formatted appropriately

Fixed

  • Fix prefix calculation for cases where the last token in context exists in the vocabulary.
  • Fix cancellation when typing a new code

232.469

Added

  • Introduce dynamic Beam Search iterations based on the probability of terminated sequences
  • Increase the number of Beam Search iterations in Python to 10

Fixed

  • Fix broken completion setting panel

232.3487

Added

  • [IDEA-309987] Filter out profanity words and phrases
  • New UI for incorrect code on new IDEs

Fixed

  • [IDEA-311177] Do not suggest calling Python objects that cannot be called
  • [IDEA-312668] Do not suggest nonexistent arguments when calling a Python function

232.2374

Fixed

  • Fix asking for anonymously logs in old IDE versions
  • Fix loading models only when specific tool window open

232.2142

Fixed

  • Put the file path in context instead of the file name
  • Fix broken completion in tsx files
  • Fix completion working in one-mode in all languages except python
  • Fixed auto-import bug in JavaScript and TypeScript

Added

  • Filter out sensitive data, e.g. secret tokens, passwords, etc., from the proposals

232.1515

Fixed

  • Fix Javascript and Typescript auto-import exception because of not running under the write action

Changed

  • Update python model with better quality, but more weight
  • [IDEA-311791] Do not show FLCC completions in Javascript and Typescript import statements

232.1058

Fixed

  • Fix showing extended settings panel, which might fail in some cases
  • Fix model not downloading on the project open
  • Fix OOM issue while storing huge diagnostics content

Changed

  • Do not hide duplicate suggestions from basic completion

Added

  • Filter out dangerous, short and single token suggestions
  • Filter out suggestions with $x=$x pattern
  • Add plugin's changelog

231.6375

Added

  • Make different levels for red code errors (hide only critical red code)
  • Add semantic correctness check
  • Check access fields in current fields and mark them as red code
  • Add prefix matching for corner cases and fix inconvenient cases
  • [IDEA-311473], [IDEA-310979] Trimming last commas, dots, and other symbols
  • [IDEA-304742] Add smarter deduplication between full line suggestions and with Basic completion
  • Improve semantic red code analysis
  • Add smart checks for Python syntax correctness
  • [IDEA-304666] Add a better UI settings panel

Changed

  • [IDEA-312239] Remove unused apache library and reduce plugin weight
  • Remove auto-imports without popups
  • Restrict auto-import to library modules
  • Do not trim underscore for double underscores
  • Do not import too short references

Fixed

  • Fix completion after typing Cyrillic letters
  • Fix a bug when all completions are red because there is an identifier right after the caret
  • [IDEA-310853], [IDEA-310802], [IDEA-310775], [IDEA-310462] UI and plugin description fixes
  • [IDEA-304510] Fix invalid syntax suggestions in type annotations
  • [IDEA-304747] Fix freeze while inserting suggestions if applying completion in a bigger Python file
  • [IDEA-304658] Fix completing the code with quotation marks
  • [IDEA-304773] Fix some cases with only partial Python completion
  • [IDEA-304746], [IDEA-304512] Fix cases with incorrect red code detection

231.3775

Added

  • Add support for JavaScript and TypeScript
  • Add auto import for JavaScript and TypeScript
  • Support v2 models and Python tokenizer
  • [IDEA-303336] Add fuzzy search in completions
  • Make settings and actions searchable
  • Bundle plugin into PyCharm EAP

Changed

  • Improv the quality of generated beam search suggestions
  • Evaluate only in supported js dialects
  • Improve red code verification for Python
  • Disable the plugin by default when bundled, enable it otherwise
  • Make plugin fully dynamic
  • Make topk1d much faster
  • Freeze model versions for all languages
  • Update K-Inference version up to 0.2.2
  • Reduce plugin's jar file size
  • Sort non-full line items with ML and keep full line items on the top

Fixed

  • Don't load model again if it's already there
  • Fix in-place prefix masking and logSoftmax in cached output
  • Fix bug with Unicode symbols in the tokenizer
  • Fix cases with spawning new threads on each tokenization
  • Fix cases with a reading model twice

0.2.69

  • [IDEA-292412] Fix conflicts with entering space and other special characters

0.2.69

  • [IDEA-292412] Fix conflicts with entering space and other special characters
  • [ML-50] Improve char filter impression
  • [ML-337] Log if local models weren't loaded

0.2.67

  • [ML-294] Prepare plugin for EAP
  • [ML-294] Improve full line settings panel
  • [ML-294] Made multiple renames (most of them ML Completion Server -> Full Line)
  • [ML-296] Add more local model management
Jul 10, 2024
Version 241.18034.76

Getting Started

After installing Full Line, a model will be downloaded automatically, and you will see new suggestions in the popup while calling completion. Plugin’s suggestions are marked with Python icon. If the model failed to download, you can re-download it through plugin settings or by restarting the IDE.

Rating & Reviews

4.0
59 Ratings (103,328 Downloads)
5
4
3
2
1

조인혁

7 days ago

Today, I installed CodeGPT, and now the Shift + → shortcut no longer works.

I don't remember if I had customized it before or if it was there by default.

Maybe the keybindings are conflicting?

I clicked on the Custom button, but I'm not sure what to do next.

0

brianaderer

1 week ago

This 'plugin' is an absolute dumpster fire. For some reason, a person at JetBrains, an otherwise great company, has deemed it a good idea to automatically bundle software which automatically inserts irrelevant code every fucking time I start typing. This plugin is on by default, so any time I install one of the 5 Jetbrains editors I use on the 3 machines I maintain, I need to go find it and turn it off before I can code, and then restart the IDE.

Dear Jetbrains: Getting rid of this piece of shit is not enough. Please get rid also of the person or people who thought this was a good idea. It fucking sucks

-2

Antonio Norman

29.01.2025

More likely to introduce bugs in your code by suggesting variables that don't exist or switching the order of the parameters to not match the function call ( even when the funtion definition and the variables have the same name )

0

Additional Information

Vendor:
Plugin ID:
org.jetbrains.completion.full.line