You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Part of #1544
Part of #1464
## Description
"Ask Cody to Fix" is one of the most common command actions we currently
have.
It makes sense to support this better instead of treating this like a
normal "edit"
- Skip fetching context, from my personal testing it very rarely helps.
Most information is contained within the error, or is part of
`relatedInformation` attached to the diagnostic
- We can start fetching more relevant context from sources directly
related to the fix: #1723
- Use a more targeted prompt preamble to address the task
## Test plan
- Fixups without code actions (edit/document/edit with no selection)
- Fixups through code action (Ask Cody to Fix)
<!-- Required. See
https://docs.sourcegraph.com/dev/background-information/testing_principles.
-->
Part of #1544
Part of #1464
## Description
The LLM reranker adds a significant amount of latency to reranking any
fetched context. This applies to all commands, chat and edits. The
latency is typically between 2-5 seconds.
This PR just disables the reranker for now.
I think this is worth doing as:
- The added latency from the reranker is not worth the possible benefit
right now. It still can sometimes rerank poorly and doesn't provide
enough value to justify waiting that long.
- Context doesn't have a constant value, in many cases context will
provide little value (e.g. simple fixups) in some cases, like chat,
context would provide a lot more value. We don't have a way of
determining that right now, so we're often adding lots of latency for no
good reason.
I created an issue here: #1721
to evaluate a different or improved approach. Possible details there on
how we could improve this.
## Test plan
On a repository without embeddings, run:
- Chat
- Commands
- Edits
Review results and context included.
closes#1464
## Description
This PR:
- Streamlines the `doc` intent flow for fixup tasks. Previously we would
treat these as "edit", which means we would try to fetch lots of context
from the selected code. That is very rarely helpful, for documenting we
really just want to take the surrounding code and the current selection.
I also investigating going the other way, and fetching relevant
reference code for a documentable symbol. This confirmed to me that it's
not necessarily helpful. It can lead to the LLM documenting code based
on how it is used elsewhere rather than how it should be used. That
might help for "explaining" code but not for documenting. (Related PR:
#1904)
## Test plan
1. Run doc command via right click menu
2. Run doc command via sidebar
3. Run doc command via code action
<!-- Required. See
https://docs.sourcegraph.com/dev/background-information/testing_principles.
-->
Description
Fixups are still quite slow. We should be improving the response speed
Areas identified already:
Reranker:rerank: {"duration":4398.333707809448}
The text was updated successfully, but these errors were encountered: