feat(autocomplete): add single-line truncation for improved acceptance rate #4321
Conversation
This file contains hidden or 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
…e rate - Add multiline parameter to postprocessGhostSuggestion to control single-line vs multi-line completions - Integrate shouldCompleteMultiline from continuedev to determine when to use single-line mode - Update FimPromptBuilder and HoleFiller to include multiline flag in prompts - Pass multiline flag through GhostInlineCompletionProvider to postprocessing - Add comprehensive tests for single-line truncation behavior When multiline=false, completions are truncated at the first newline character, which typically results in higher acceptance rates for inline completions.
|
Add rate limiter configuration (Map, RATE_LIMIT, WINDOW_MS) and clientIP variable to provide better context for autocomplete. Move autocomplete marker to new line after TODO comment to encourage actual implementation rather than more TODO comments. Remove rejected approval files as they are no longer valid for the updated test case.
Add cache Map and cacheTTL in constructor to provide context for caching implementation. Move autocomplete marker inside fetchData method to encourage actual cache lookup implementation rather than more TODO comments. Remove rejected approval files as they are no longer valid for the updated test case.
Add clear purpose (sum calculation) with total variable and comment to guide the completion. Place autocomplete marker after 'total' to encourage += num completion. Add end block and puts statement to constrain the expected output. Remove rejected approval files as they are no longer valid for the updated test case.
Wrap method in Calculator class with end statement and add sibling calculate_product method to provide pattern context. This constrains the completion to just the method body rather than generating a whole program. Remove rejected approval files as they are no longer valid for the updated test case.
Add sibling functions (multiply, divide) and module.exports to provide pattern context and constrain the completion. This prevents the LLM from generating additional functions after completing the add function. Remove all approval files (both approved and rejected) as they are no longer valid for the updated test case.
Consolidate multi-file test into single file with class definition directly visible above the autocomplete point. Add userId variable to provide context for findById call. This ensures the LLM has direct access to the actual method names (findById, findByEmail, updateUser). Remove rejected approval files as they are no longer valid for the updated test case.
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.
Summary
This PR adds context-aware single-line truncation to the autocomplete functionality to improve acceptance rates.
Changes
uselessSuggestionFilter.ts: Added
multilineparameter topostprocessGhostSuggestion. Whenmultiline=false, completions are truncated at the first newline character.types.ts: Added
multiline: booleanfield toFimGhostPromptandHoleFillerGhostPromptinterfaces.FillInTheMiddle.ts: Integrated
shouldCompleteMultilinefrom continuedev to determine when to use single-line mode based on context (cursor position, language, etc.).HoleFiller.ts: Same integration for the hole-filler strategy.
GhostInlineCompletionProvider.ts: Pass the multiline flag through to postprocessing.
How it works
The
shouldCompleteMultilinefunction from continuedev determines whether to allow multi-line completions based on:When single-line mode is active, completions are truncated at the first newline, resulting in shorter, more focused suggestions that typically have higher acceptance rates.
Testing
Added comprehensive tests for the new single-line truncation behavior in
uselessSuggestionFilter.test.ts.