ghost: fetch: go back to DONE as finish_reason for non-server-terminated choices #3122
+288
−1
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
Contributor
ulugbekna
commented
Jan 23, 2026
ulugbekna
commented
- ghost: fix: align finishReason behavior in fetchAndStreamCompletions2
- ghost: fetch: go back to DONE as finish_reason for non-server-terminated choices
## Problem
There was a user-observable behavior difference between the old
\`fetchAndStreamCompletions\` and the new \`fetchAndStreamCompletions2\`
functions. This difference affected whether cached follow-up suggestions
were used or a new multiline fetch was triggered after accepting a
completion.
## Root Cause
When the stream ends without all completions being explicitly finished
by the server, the two implementations produced different \`finishReason\`
values in the \`APIChoice\`:
| Scenario | Old (SSEProcessor) | New (convertStreamToApiChoices) |
|-------------------------------|---------------------|----------------------------------|
| Stream ends, no finish_reason | 'stop' (default) | 'DONE' (was hardcoded) |
## Impact
The \`hasAcceptedCurrentCompletion\` function in \`current.ts\` (line 79)
uses the \`finishReason\` to determine whether a completion was fully
accepted:
\`\`\`typescript
return exactMatch && finishReason === 'stop';
\`\`\`
This function ONLY returns \`true\` when \`finishReason === 'stop'\`.
With the old behavior:
- \`finishReason\` = 'stop' → \`hasAcceptedCurrentCompletion\` returns \`true\`
- This triggers multiline follow-up completion mode in \`ghostTextStrategy.ts\`
With the new (broken) behavior:
- \`finishReason\` = 'DONE' → \`hasAcceptedCurrentCompletion\` returns \`false\`
- This incorrectly uses cached suggestions instead of triggering new fetch
## Investigation Method
1. Traced the flow in \`SSEProcessor.processSSE\` (stream.ts):
- When stream ends via [DONE], calls \`finishSolutions\` with reason 'DONE'
- However, \`prepareSolutionForReturn\` calls \`convertToAPIJsonData\`
- \`convertToAPIJsonData\` defaults \`finish_reason\` to 'stop':
\`finish_reason: streamingData.finish_reason ?? 'stop'\`
2. Traced the flow in \`convertStreamToApiChoices\` (fetch.ts):
- Final loop for unfinished completions had hardcoded 'DONE'
- This bypassed the default 'stop' semantics
3. Verified via \`hasAcceptedCurrentCompletion\` tests in \`current.test.ts\`:
- Tests confirm 'stop' → true, 'content_filter'/'snippy' → false
- 'DONE' would also return false, breaking the multiline flow
## Fix
Changed line 921 in \`convertStreamToApiChoices\` from:
\`'DONE', // should match original ghost-text fetcher behavior\`
to:
\`completion.accumulator.finishReason ?? 'stop', // Matches SSEProcessor\`
This ensures the new implementation produces the same default 'stop'
\`finishReason\` as the old implementation when the server doesn't provide
an explicit finish_reason."
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.