fix(compressor): use text char sum for multimodal token estimation in _find_tail_cut_by_tokens#16369
Merged
Conversation
… _find_tail_cut_by_tokens
_find_tail_cut_by_tokens called len(content) to estimate message tokens.
When content is a list of blocks (multimodal: text + image_url), len()
returns block count (e.g. 2) rather than character count, so a message
with 500 chars of text was counted as ~10 tokens instead of ~135.
This caused the backward walk to exhaust all messages before hitting the
budget ceiling; the head_end safeguard then forced cut = n - min_tail,
shrinking the protected tail to the bare minimum and preventing effective
compression of long multimodal conversations.
Fix mirrors the existing pattern in _prune_old_tool_results (line 487):
sum(len(p.get("text", "")) for p in raw_content)
if isinstance(raw_content, list) else len(raw_content)
Tests: 3 new cases in TestTokenBudgetTailProtection — regression guard
(confirms the test fails with the bug), plain-string regression guard,
and image-only block edge case.
Fixes #16087.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…t list
raw_content from message["content"] can be a list that contains bare
strings, not only dicts. The previous `p.get("text", "")` call raised
AttributeError on string items, crashing context compression for any
session that had a message with mixed content.
Guard with isinstance checks: dict → .get("text"), str → len(p),
fallback → len(str(p)). Adds a regression test covering the bare-string
case that would have AttributeError'd on the pre-fix code.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The bare-string isinstance guard added in 80ae262 covered _find_tail_cut_by_tokens (line 1084) but missed the identical pattern in _calculate_protect_tail_boundary (line 487, the protect-tail scan loop). Both loops call .get("text", "") on every list item in message["content"]; both crash with AttributeError when that list contains a bare string. Apply the same dict/str/fallback isinstance guard to the protect-tail path. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Contributor
Author
|
Post-merge clarification on the actual trigger: The reporter's stated repro (
So Where the miscount actually bites: list-shaped The fix is still correct — list-shaped content arithmetic was wrong regardless of origin — but the user-attached-image scenario from the issue is theoretical, not observable. |
19 tasks
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
3 participants
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.
Multimodal messages are now sized by text-char count, not block count — so tail protection stops at the token budget instead of exhausting the message list.
Salvages #16113 by @briandevans onto current main. Closes #16087 (reported by @SimbaKingjoe). Also closes #16117 (same bug, narrower fix).
Root cause
_find_tail_cut_by_tokenscalledlen(content) // _CHARS_PER_TOKEN + 10. Whencontentis a list of blocks ([{"type": "text", "text": "..."}, {"type": "image_url", ...}]),len()returned block count (~2), so every multimodal message scored ~10 tokens regardless of size. The backward walk exhausted the list before hitting the budget ceiling, the head_end safeguard forcedcut = n - min_tail, and compression had nothing left to summarize.Changes
agent/context_compressor.py— sum text-char lengths across list content blocks in both_find_tail_cut_by_tokensand_prune_old_tool_results, with isinstance guards for dict / str / fallback so bare-string items don't AttributeError.tests/agent/test_context_compressor.py— 4 regression tests: multimodal char-sum, plain-string unchanged, image-only block, mixed list with bare strings.Validation
tests/agent/test_context_compressor.py_find_tail_cut_by_tokensw/ real ContextCompressor