Docs / FAQ

FAQ

Question: Is the plugin free to use?

Answer: ChatGPT EasyCode is completely free to use. However, to access GPT-4, you need to buy tokens. All the other features including unlimited access to gpt-3.5-turbo are free. Pricing is subject to change.

Question: ConnectException: Connection timed out: connect?

Answer: This situation indicates that you cannot access the official server of Open AI and may need a proxy. You can use the following method to test. Use the curl tool to access directly, please remember to replace the following API Key with your own:

`curl https://api.openai.com/v1/chat/completions -H "Authorization: Bearer <Your API Key>" -H "Content-Type: application/json" -d '{ "model": "gpt-3.5-turbo", "messages": [{"role": "system", "content": "Hello!"}] }'`

Question: Cannot find the plugin in the marketplace?

Answer: Currently the plugin supports versions from 2020.1 to 2023.1.(Support EAP platform). If your IDE version is less than 2020.1, it cannot be searched. There is no way to use either. Please consider updating your IDE version to 2020.1 or later.

Question: What happens to my data? Do you store my data?

Answer: We never store your code. Your data will not be used for training other AI models if you opt out of data collection. We use OpenAI to process the data. Your data does leave the machine for embedding & querying. It is retained for 30 days by OpenAI for abuse & misuse monitoring, after which it is automatically deleted. See OpenAI’s Privacy Policy.

Question: How does codebase indexing work?

Answer: At a high level, we use embeddings to create a vectorization of your codebase and use it to intelligently query GPT.

Question: Can I use my own OpenAI key?

Answer: Yes, add it in extension settings.

Question: I asked “Which GPT version are you” and it tells me it’s GPT-3, but when I ask the same question to chatGPT, it tells me it’s GPT-4. How do I know if this is really GPT-4?

Answer: GPT-4 is trained on pre-2021 data and doesn't know which model it is. The latest version of ChatGPT Plus may have addressed this using prompt engineering, but earlier versions had the same issue.

You can see it being discussed on OpenAI Community forum (here and here), and on Reddit (here and here). To truly test which GPT model you are using, you should give it a fairly complex task and see how it performs. Here is a good example (source):

Prompt Bob is in the living room.He walks to the kitchen, carrying a cup.He puts a ball in the cup and carries the cup to the bedroom.He turns the cup upside down, then walks to the garden.He puts the cup down in the garden, then walks to the garage.Where is the ball?

GPT-3 will tell you the ball is in the garden (incorrect)

GPT-4 will tell you the ball is in the bedroom (correct)

Question: Do you accept other forms of payment such as crypto or PayPal?

Answer: Not at the moment, but we may add these in the future.

Question: Why are GPT-4 tokens used up so quickly?

Answer: First, let’s make sure you understand how tokens work:

Each word is roughly 1.33 tokens.

Completion (GPT output) costs twice as much tokens as prompt (your question).

Follow up questions automatically include history as context, so it consumes more tokens.

“Ask codebase” injects relevant code from your codebase as context, so it’s the most costly.

In general, GPT-4 is not cheap. We have some recommendations for saving cost:

Only use GPT-4 for questions that GPT-3.5 can’t handle.

For questions that only requires local context, select code and ask GPT instead of “asking codebase”.

“Asking codebase” should be used when larger codebase context is required, and with careful prompting.

FAQ

Getting Started

FAQ