VectorCode 0.7.2 Update: Summarisation on retrieval results

[Davidyz]

VectorCode 0.7.2 release contains a new feature in the @vectorcode_query tool: result summarysation.

This opt-in feature calls an LLM (defaults to the chat LLM, but you can configure it to use a different model/provider) to summarise the retrieval results returned by VectorCode. This makes the tool output more concise.

This is a useful feature for you if:

1. Your main chat LLM is expensive, and you want to save some input tokens, then you can configure the summariser to use a cheap/free model so that your main model sees a good outline of the repository structure but consumes less tokens;
2. You often run out of the context window, then you can enable the summarisation so that the tool's output in the chat will be shorter than the plain source code, allowing you to maintain longer conversations with the model;
3. You don't want to send your code to the cloud service providers, but your locally hosted models aren't good enough for complex SWE tasks, then you can use the local model as a summariser, and the cloud providers will only see a summary, not the actual code.
4. The possibility goes on.

If you find this interesting, check out the wiki to see how you can use it. I'd love to hear your feedback!

Regards,
David

[olimorris]

The VectorCode rocket ship continues to reach new heights 👏🏼

[Davidyz]

Wouldn't have happened without codecompanion.nvim as such a great foundation!