search

TabbyML / tabby

Self-hosted AI coding assistant
https://tabbyml.com
Other
21.88k stars 997 forks source link

Enable repository context in Answer Engine #2561

Open coffeebe4code opened 4 months ago

coffeebe4code commented 4 months ago

Downloaded 0.13 tabby, installed and ran successfully on port 8080 with DeepseekCoder-6.7B and Mistral-7B

Using Cuda.

When asking the tabby assistant on the webpage, or after clicking ask tabby in the repository with the file in question. It seemingly has no access to the context.

After asking it to list the first 5 lines of the file, it fails miserably.

How do I configure a chat instance or something to actually use the indexed repository

The index does show success.

wsxiaoys commented 4 months ago

Hello, source code context has not yet been enabled in the Answer Engine (the homepage search feature), but we expect to make it available very soon (e.g 0.14)

coffeebe4code commented 4 months ago

While my question was directed to be the Answer Engine, and it's good to hear that it will be added soon! Thank you very much for your quick reply. My intent was not exclusive to the Answer Engine.

I was unclear when I said "How do I configure a chat instance or something". So I am sorry about that!

I am interested in how I can get it working right now. So if there is another way to run or provide the vectorizations to a query, whether through vim extension or just the server api. Please let me know.