Cap get_terminal_output first-poll and non-prefix responses to a tail#320140
Merged
Conversation
Contributor
📬 CODENOTIFYThe following users are being notified based on files changed in this PR: @anthonykim1Matched files:
|
Contributor
There was a problem hiding this comment.
Pull request overview
This PR further reduces get_terminal_output payload sizes (when chat.tools.terminal.outputDeltas is enabled) by tail-truncating the first poll and non-prefix fallback responses to ~8 KB (line-aligned), preventing Copilot SDK spillover thrash while keeping full-buffer snapshotting for delta detection.
Changes:
- Add an internal 8 KB “tail budget” and tail-formatting helper to cap first-poll and non-prefix-fallback outputs.
- Preserve full-buffer snapshot (length + hash) so unchanged/delta detection continues to work.
- Add unit tests for tail behavior on first poll and on non-prefix fallback.
Show a summary per file
| File | Description |
|---|---|
| src/vs/workbench/contrib/terminalContrib/chatAgentTools/browser/tools/getTerminalOutputTool.ts | Adds tail-capping helpers and applies them to first-poll and non-prefix fallback outputs under the output-deltas experiment. |
| src/vs/workbench/contrib/terminalContrib/chatAgentTools/test/browser/getTerminalOutputTool.test.ts | Adds tests validating tail truncation behavior for large outputs on first poll and non-prefix fallback. |
Copilot's findings
Comments suppressed due to low confidence (1)
src/vs/workbench/contrib/terminalContrib/chatAgentTools/browser/tools/getTerminalOutputTool.ts:117
- When output exceeds the tail budget, the first poll returns only a tail, but the unchanged-marker still says "(… characters already shown)". That text becomes inaccurate/misleading because the tool did not actually return the full buffer previously, only the tail. Consider adjusting the unchanged-marker when
output.lengthexceeds the tail budget so it doesn’t imply the full output was already shown.
if (currentOutputSnapshot.length === previousOutputSnapshot.length && currentOutputSnapshot.hash === previousOutputSnapshot.hash) {
return `Output of terminal ${id} unchanged since previous poll (${output.length} characters already shown). No new output.`;
}
- Files reviewed: 2/2 changed files
- Comments generated: 1
Collaborator
Author
|
/requires-eval-assessment terminalbench2 gpt-5.4,claude-opus-4.6,claude-opus-4.7 |
meganrogge
added
the
~requires-eval-assessment
Contributor
|
⏳ Queued vscode build for
|
…l output was previously returned
meganrogge
added
~requires-eval-assessment
Contributor
|
⏳ Queued vscode build for
|
Contributor
|
🚀 Queued eval-assessment publish build for
|
lramos15
approved these changes
Jun 5, 2026
Contributor
|
🔬 Queued eval-assessment benchmark for
Results will be posted back here when the run completes. |
Contributor
|
✅ Eval-assessment build published.
|
vs-code-engineering
Bot
removed
the
~requires-eval-assessment
Contributor
|
📊 Eval-assessment benchmark complete.
🧪 Results |
Contributor
|
📊 Eval-assessment benchmark complete.
🧪 Results |
Contributor
|
📊 Eval-assessment benchmark complete.
🧪 Results |
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
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.
Fixes the spillover side of microsoft/vscode-internalbacklog#7869.
PR #315543 already shrinks repeated
get_terminal_outputpolls by returning only the delta. But the first poll and the non-prefix fallback still returned the full buffer (up to the ~60 KB upstream cap inoutputHelpers.ts), which is well above the Copilot SDK's ~10 KB spillover threshold and causes the agent to thrash withread_filecalls on the spillover temp file.This change adds a second, tighter cap inside the tool itself: when the full output exceeds 8 KB, return only the last 8 KB (line-aligned). The truncation marker includes both the omitted character count and a hint telling the agent how to recover the head if it needs to: re-run the command and redirect output to a file, then read that file.
Snapshot tracking (length + hash) still uses the full buffer, so subsequent delta polls work unchanged.
Behavior
The tail cap applies on every code path (always on — not gated by an experiment), because the SDK spillover issue affects all users regardless of whether
chat.tools.terminal.outputDeltasis enabled. The delta-tracking semantics introduced by #315543 remain gated by that experiment.unchanged since previous pollmarker (no output).Risks / follow-ups
\nin the last 8 KB fall back to a raw character cut, which could land mid-ANSI-escape. Rare.Tests
getTerminalOutputTool.test.ts— all 11 tests pass, including two added for the new behavior:returns only the tail on first poll when output exceeds the tail budgetreturns only the tail on non-prefix fallback when output exceeds the tail budget