sweepai/sweep

Sweep: open-source AI-powered Software Developer for small features and bug fixes.

sweep's Issues

Sweep: allow vector db to take multiple queries and then perform majority voting of the result files

Sweep: add suffix to PR creation message with a zsh script to checkout to that branch

Sweep: Remove the represent this file as ... prefix from vector_db.py and from querying

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep shouldn't retry if issue is closed

Sweep: In on_comment.py it's possible that the comment is on a closed PR. If this is the case then we should not fire an event. Add logic to on_comment to address this. Use pr state

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep: Reverting a file should not toggle it. instead if a file has been reverted two times in a row, go two commits back

Sweep: In on_comment.py it's possible that the comment is on a closed PR. If this is the case then we should not fire an event. Add logic to on_comment to address this. Use pr state

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit

Sweep: prefix/suffix snippet with "..." in the prompt if a snippet is not at the beginning or end of file

Go to src/core/models.py and references to Snippet
and add fields like:
is_snippet_file_start = False
is_snippet_file_end = False
also set these fields correctly

Sweep: Make install script for setting up repo.

Install poetry packages. deeplake package is not supported by poetry, so use pip to install this package.

Sweep: Remove "sweep" or "sweep: " from the issue title if it exists.

do this in "issues labeled" for the input to sweepbot

Sweep: We want a github action which runs benchmarks on merge to main. This should use logic from tests/create_sample_issue.py but create more sample issues and somehow measure their quality

Sweep: Pass in comment.id to on_comment.py. Then fetch this comment object to add eyes emoji, delete the eyes emoji, and add a rocket

Sweep: add a config file called sweep.yaml where you can specify a git commit hash to index the repository at that state

Current behaviour of Sweep AI's indexing always picks up the latest master at the time of request.
The task if to add a config called sweep.yaml where you can specify a parameter called commit-hash which can then be used to index the repository's content at that specific commit.

Repo Source for a commit hash xyz would be
https://github.com/user/project_name/tree/xyz
where user and project_name are used to the specify the owner of the repository and the repository on which the issue is created.

Sweep: add a config file called sweep.yaml where you can specify a git commit hash to index the repository at that state

Current behaviour of Sweep AI's indexing always picks up the latest master at the time of request.
The task if to add a config called sweep.yaml where you can specify a parameter called commit-hash which can then be used to index the repository's content at that specific commit.

Repo Source for a commit hash xyz would be
https://github.com/user/project_name/tree/xyz
where user and project_name are used to the specify the owner of the repository and the repository on which the issue is created.

Sweep: prefix/suffix snippet with "..." in the prompt if a snippet is not at the beginning or end of file

Go to src/core/models.py and references to Snippet
and add fields like:
is_snippet_file_start = False
is_snippet_file_end = False
also set these fields correctly

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep: We want to add a feature where the user may specify certain files to exclude from Sweep's vector retrieval.

This can be found in src/core/vector_db.py. We want to use SweepConfig to do this. The file specifying the excluded files will be in the root of the user's repo (get_repo).

Sweep: Add an indentation parameter in src/utils/file_change_functions.py which will allow Sweep to specify how much to indent (number of tabs)

Sweep: Set up Github actions for this repo

Just look at README.md

sweep: add a feature that allows repositories to specify sweep.toml, which configures included and excluded file paths and extensions. Also add an example in sweepai/sweep

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit

Sweep: In on_comment.py it's possible that the comment is on a closed PR. If this is the case then we should not fire an event. Add logic to src/api.py to address this.

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit

Sweep: Add an indentation parameter in src/utils/file_change_functions.py which will allow Sweep to specify how much to indent (number of tabs)

Sweep: Remove the "sweep" label on failure. This should be in src/handlers/on_ticket.py

Sweep: Rerank the search results using fuzzywuzzy. Use src/core/vector_db.py

Sweep: Randomly use GPT4 or GPT4-0613 for both the normal and 32k versions. This will let us load balance and avoid rate limits

Sweep: Rerank the search results using fuzzywuzzy. Use src/core/vector_db.py

Sweep: Add a prefix to the embedded code

Ex: if the snippet comes from filename: say "Represent this code snippet from {filename} for retrieval: {snippet}"
Also do it for query, say "Represent this natural language query for code retrieval: {query}"

Sweep: Add eyes reaction to replies in on_comment when they have been addressed using comment id

Sweep: allow vector db to take multiple queries

Sweep: Add an indentation parameter in src/utils/file_change_functions.py which will allow Sweep to specify how much to indent (number of tabs)

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit

Sweep: Reduce the number of retries/multiple tickets from sweep

Sweep: Rerank the search results using fuzzywuzzy. Use src/core/vector_db.py

Sweep: Modify posthog to log on_ticket successes separately from failures src/handlers/on_ticket.py

Sweep: If multiple file_change_requests have the same file_name, fuse them and concatenate their instructions. Do this with as few changes as possible.

Sweep: parallelize vector_db embedding generation

Sweep shouldn't retry if issue is closed

It seems that sweep continues to retry even if an issue is closed. See: sustainable-processes/ORDerly#148

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit

Sweep: Add an option for the user to comment REVERT In src/handlers/on_comment.py which will rollback that file one commit