Hello, I am a licensed user, but despite this, I am using copilot, and it is driving me crazy. If you do not resolve this slowdown, I am sorry, but I will not pay for an application that I can no longer use.
The time for the answer depends on the analysis, LLM load (cause we have multiple models for selection) and the network situation, based on the usage, thank you for your understanding.
I have used it and had lots of help, but since latest update it is so slow to give an answer on anything I might as well copy/paste/describe/prompt in another client, Gemini, CoPilot or ChatGPT. It must be possible to solve. It costs a lot of money.
I've too noticed a massive slowdown - can take anywhere from minutes for 10 minutes or so - for an answer not even dependent on context.
It's the Codebase button doing it in my instance. Turn it OFF and all is quick again. Maybe it's the massive Unreal Engine codebase reference in the project, maybe it's because I might have 30 files open, not sure but it stays off now and I add files manually.
Hello, thank you. "I’ll test Andrewkehrig—that makes sense.
However, in other AI assistant apps, ‘full project indexing’ (embeddings) is typically done once and cached if Semantic Indexing is enabled. But JetBrains runs context gathering (relevant file selection) every single time a new chat starts. That’s why the indexing/progress bar keeps popping up, wasting tokens, and slowing things down. Why isn’t there a permanent “index once, use across all chats” option? This is ridiculous—JetBrains must want their customers to waste more tokens.
That’s exactly why the tokens run out so quickly. But this goes against business ethics and professionalism.
When using AI-powered features, the IDE sends your prompt to the selected LLM provider. Along with the prompt, additional context may be included, such as relevant code snippets, file types, or detected frameworks.
This data is used only to generate the response for that specific request and is not stored or reused afterward. Because of this, the full context needs to be sent with each request.
JetBrains collects only limited and controlled data to improve its products. Users can control the sharing of detailed AI-related data by navigating to the “Allow detailed data collection by JetBrains AI” setting: Settings | Appearance & Behavior | System Settings | Data Sharing
Here are a few best-practice recommendations that often help decrease AI Quota usage:
Start new chats for new topics: Long conversations increase token usage since more context is sent with each request.
Yes indeed.
It has never performed well, buy now it became unusable.
For the price I really expected a Cursor like experience. This is really disappointing.
Hello, I am a licensed user, but despite this, I am using copilot, and it is driving me crazy. If you do not resolve this slowdown, I am sorry, but I will not pay for an application that I can no longer use.
I have used it and had lots of help, but since latest update it is so slow to give an answer on anything I might as well copy/paste/describe/prompt in another client, Gemini, CoPilot or ChatGPT. It must be possible to solve. It costs a lot of money.
I've too noticed a massive slowdown - can take anywhere from minutes for 10 minutes or so - for an answer not even dependent on context.
It's the Codebase button doing it in my instance. Turn it OFF and all is quick again. Maybe it's the massive Unreal Engine codebase reference in the project, maybe it's because I might have 30 files open, not sure but it stays off now and I add files manually.
Hello,
To help us troubleshoot the issue, please reproduce it and provide the following information in the ticket LLM-3570:
Turning off ‘Codebase mode’ made AI Assistant chat fast again for me.
Hello, thank you. "I’ll test Andrewkehrig—that makes sense.
However, in other AI assistant apps, ‘full project indexing’ (embeddings) is typically done once and cached if Semantic Indexing is enabled. But JetBrains runs context gathering (relevant file selection) every single time a new chat starts. That’s why the indexing/progress bar keeps popping up, wasting tokens, and slowing things down. Why isn’t there a permanent “index once, use across all chats” option? This is ridiculous—JetBrains must want their customers to waste more tokens.
That’s exactly why the tokens run out so quickly. But this goes against business ethics and professionalism.
When using AI-powered features, the IDE sends your prompt to the selected LLM provider. Along with the prompt, additional context may be included, such as relevant code snippets, file types, or detected frameworks.
This data is used only to generate the response for that specific request and is not stored or reused afterward. Because of this, the full context needs to be sent with each request.
JetBrains collects only limited and controlled data to improve its products. Users can control the sharing of detailed AI-related data by navigating to the “Allow detailed data collection by JetBrains AI” setting:
Settings | Appearance & Behavior | System Settings | Data Sharing
Here are a few best-practice recommendations that often help decrease AI Quota usage:
Long conversations increase token usage since more context is sent with each request.
To ensure the AI Assistant follows your project’s coding standards and technology stack.
For more details, see:
Product Data Collection and Usage Notice