forked from princeton-nlp/SWE-agent
-
Notifications
You must be signed in to change notification settings - Fork 0
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[PRO-812] manual input #18
Merged
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ry + patch to a directory. add flags to run.py to continue a conversation with manual input
toshok
force-pushed
the
toshok/pro-813-manual-input
branch
2 times, most recently
from
September 26, 2024 19:49
ad61682
to
11c2881
Compare
…ens within the container now, so it shows up properly in the tool result content
bhackett1024
approved these changes
Sep 27, 2024
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nice!
Domiii
approved these changes
Sep 30, 2024
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Cool!
- Can you update the runbook here?
- Looking forward to seeing a demo of this in action 🥳
…o), remove the line about the last reproduction (which the model didn't seem to have a problem with) and remove the tdd_results code since it's not used anymore
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
oops, ignore the branch name.
enabled by setting
MANUAL_TDD_INPUT_DIRECTORY
to the base storage dir.Removes the tdd result from the initial instructions and replaces it with:
That way the flow is substantially easier to reason about (and the directory structure makes quite a bit more sense), as only
tdd_repro
usage needs to be considered.when you run the
run-instance.sh
script you can now run it in two ways:Starting a new conversation tree
./run_instance.sh django__django-12419
This will run the instance with the id
django__django-12419
and will create the directory$BASE/django__django-12419/root-$conversation_hash
where$conversation_hash
is a hash of the conversation.json file contained within it.This will result in the following directory structure:
Testing new manual input (adding a new branch to the conversation)
First, create a
.md
file inside a conversation dir (sibling to theconversation.json
file)Then
./run_instance.sh django__django-12419 $CONVERSATION_PATH $CONTINUATION_LABEL
This will compute the directory
$BASE/django__django-12419/$CONVERSATION_PATH
, then load the following files from that directory:conversation.json
patch.diff
$CONTINUATION_LABEL.md
The environment will be initialized with the normal repo contents + the patch applied, and the model will be initialized with the contents of conversation.json. This conversation always ends with a tdd_repro tool_use. The agent then acts as if the model has responded with this tdd_repro tool_use, runs the tool, and augments the tool response with the contents of the file
$CONTINUATION_LABEL.md
. Execution proceeds normally until the nexttdd_repro
tool use, at which point the conversation and new patch will be saved to the subdir$CONTINUATION_LABEL-$conversation_hash/
(again where conversation_hash is a hash of the conversation.json file contained within that subdir).As an example, the command line:
When run in a directory structure like this:
will result in a directory structure like this: