Code Intelligence

Add this URL to your Eclipse Installation to reach this solution's update site.

2026-03 (4.39), 2025-12 (4.38), 2025-09 (4.37), 2025-06 (4.36), 2025-03 (4.35), 2024-12 (4.34), 2024-09 (4.33), 2024-06 (4.32), 2024-03 (4.31), 2023-12 (4.30), 2023-09 (4.29)

https://chabicht.github.io/code-intelligence/update-site/
Learn more...
Solution Description

A plugin that enhances the Java Development Tools (JDT) with AI assisted code completion and an AI chat window.

  • Supports API access to Anthropic, DeepSeek, Google Gemini, Groq, Ollama, OpenAI, and X.ai/Grok.
  • Integrates into the JDT tooling in order to help you craft meaningful prompts with the context information you need.
  • Allows fine-tuning completion prompts and system messages for the models you use.
  • Provides various tools that models capable of function-calling can use to gather information or edit files in the codebase.

Categories: Editor, IDE, J2EE Development Platform, Programming Languages, Tools

Tags: Artificial intelligence, chatgpt, AI code analysis, AI, GPT, Anthropic, DeepSeek, Gemini, Groq, Ollama, OpenAI, X.ai, Grok

Additional Details

Eclipse Versions: 2026-03 (4.39), 2025-12 (4.38), 2025-09 (4.37), 2025-06 (4.36), 2025-03 (4.35), 2024-12 (4.34), 2024-09 (4.33), 2024-06 (4.32), 2024-03 (4.31), 2023-12 (4.30), 2023-09 (4.29)

Platform Support: Windows, Linux/GTK

Organization Name: com.chabicht

Development Status: Alpha

Date Created: Saturday, February 15, 2025 - 19:26

License: EPL 2.0

Date Updated: Saturday, April 11, 2026 - 14:11

Submitted by: Carsten Habicht

Screenshot
Date Ranking Installs Clickthroughs
April 2026 33/576 432 54
March 2026 42/627 472 50
February 2026 37/620 500 53
January 2026 44/631 477 54
December 2025 48/621 428 51
November 2025 48/626 476 31
October 2025 56/637 431 45
September 2025 54/653 422 75
August 2025 56/615 372 37
July 2025 52/636 426 49
June 2025 48/627 474 55
May 2025 60/634 365 59
View Data for all Listings

Unsuccessful Installs

Unsuccessful Installs in the last 7 Days: 1

Count Error Message
1 Cannot continue the operation. There is another install operation in progress.
Download last 500 errors (CSV)

Marketplace Drag to Install Button

By adding the following code below to your website you will be able to add an install button for Code Intelligence.

HTML Code:

Markdown Syntax:

Output:

Drag to your running Eclipse* workspace. *Requires Eclipse Marketplace Client

Reviews Add new review

Submitted by Marek Novy on Tue, 04/07/2026 - 05:48

Hi there,

Firstly, thanks a lot for your work on the plugin. It is one of not too many plugins which actually works for me in the Eclipse for my local AI Ollama server.

 (Eclipse - Version: 2026-03 (4.39.0) Build id: 20260305-0817, AI model - Ollama/qwen3.5:27b, Code Intelligence 1.0.0)

I use it mostly for the generation of unit tests. As the plugin is not able to run commands directly trought the tools usage

e.g: I needed to run mvn test -Dtest=SOME_TEST_CLASS

I wrote small bash script to pick up name of class to run the test for from the file on file system.

This script runs constantly(in loop) on my laptop and when it finds the test class name it executes the tests. The results of tests are sent into another file on the file system where it can be read the plugin. This way I got aroud the limitation of the plugin - a small queue system for tests execution over file system :) 

However, I experienced an issue that after some time the tooling of plugin stopped working. Basically, it was not able to execute tool commands anymore. For example, if it tried to read the result of tests from filesystem this appeared in my plugin's chat window 

<tool_call> <function=read_file_content> <parameter=file_name> testClassResult.txt </tool_call>

so it looked like that it tried but for some reason the execution of tool did not happened and ended in the chat window.

Cheers,

Marek

Submitted by Carsten Habicht on Sat, 04/11/2026 - 14:21

Hi Marek,

I noticed this behavior especially with Qwen models. Works fine for like 5 turns and then totally derails.
It apparently helps to add some custom parameters with the recommended settings. E.g. hat's what I configured for my Ollama connection:

{
  "options": {
    "num_ctx": 131072,
    "temperature": 0.6,
    "top_p": 0.95,
    "top_k": 20,
    "min_p": 0.0,
    "presence_penalty": 0.0,
    "repetition_penalty": 1.0
 }
}

Qwen3.5 docs on HuggingFace say:

We recommend using the following set of sampling parameters for generation

  • Thinking mode for general tasks: temperature=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0
  • Thinking mode for precise coding tasks (e.g. WebDev): temperature=0.6, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0
  • Instruct (or non-thinking) mode for general tasks: temperature=0.7, top_p=0.8, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0
  • Instruct (or non-thinking) mode for reasoning tasks: temperature=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0

Please note that the support for sampling parameters varies according to inference frameworks.

Cf.:
- https://github.com/chabicht/code-intelligence/blob/master/CUSTOM_PARAMS.md
- https://huggingface.co/Qwen/Qwen3.5-27B

 

Submitted by Marek Novy on Fri, 04/17/2026 - 18:53

Thanks for the info, very much appreciated.

I have recently used qwen3.6:35b-a3b, and it worked like a charm for the generation of Java unit tests. If anybody is interested, arguably the Qwen models were the most handy for this kind of task (Note: I also played with gemma4, glm-4.5, nemotron-3-nano with similar-sized models, yet they were not so good. Probably more handy for other stuff. For the JAVA coding, the Qwen models were the best fit, at least for me.)

Here are parameters for the qwen3.6:35b-a3b

{
 "options": {
   "num_ctx": 61000,
   "repeat_penalty":       1.2,
   "temperature":          0.6,
   "top_k":                20,
   "top_p":                0.95,
   "min_p":               0,
   "presence_penalty":    0.0
 },
 "keep_alive": "5m"
}

Context window size is limited by the size of my GPU RAM, so adjust for your needs :( To be honest, the parameters still need more tweaking, but for starters, they are OK.

For some reason apply_patch tool did not work for me most of the time, so I had to switch it off. 

I got my queue system (I mentioned above) for the external running of unit tests through the filesystem tuned well, so the AI model can integrate with it without problems now. However,  it would be really nice to have a future tool for executing shell commands. That way I would not need such a "crazzy" solution on my side :)

Again, thanks a lot for the plugin.

Cheers,

Marek

Submitted by Carsten Habicht on Thu, 03/13/2025 - 18:14

Hey Rafael,

thanks for the effort! :-)

I took your comment as a hint that some kind of assisted configuration would be helpful and added a setup wizard.
Hope that helps.

Best,
Carsten

Submitted by Rafael Pereira on Thu, 03/06/2025 - 11:47

If you want to set up with the Gemini API, follow these instructions:

  1. Go to https://aistudio.google.com/app/apikey and generate a new key.
  2. Install the Code Intelligence plugin in your Eclipse.
  3. Open the menu Window -> Preferences, go to Code Intelligence, and add a new API Connection:
    Name: Gemini
    Type: Gemini
    Base URI: https://generativelanguage.googleapis.com/v1beta
    API key: [your api key]
  4. Change the Model to the desired one (Gemini/models/gemini-2.0-flash).

Additionally, you can also enable the chat by adjusting the desired model for this function.