Uh oh!
There was an error while loading. Please reload this page.
fix: bedrock reasoning issue - #25303
Merged
Merged
Conversation
rekram1-node
commented
May 1, 2026
CollaboratorAuthor
/review |
Contributor
lgtm |
Uh oh!
There was an error while loading. Please reload this page.
66 tasks
oleksii-honchar pushed a commit
to oleksii-honchar/better-opencode
that referenced
this pull request
May 6, 2026
isaacfinnegan pushed a commit
to isaacfinnegan/whispercode
that referenced
this pull request
Jun 9, 2026
AIALRA-0 pushed a commit
to AIALRA-0/opencode-turn-engine
that referenced
this pull request
Jun 10, 2026
AIALRA-0 pushed a commit
to AIALRA-0/opencode-turn-engine
that referenced
this pull request
Jun 10, 2026
ESRE-dev added a commit
to ESRE-dev/opencode
that referenced
this pull request
Jun 10, 2026
* The branch existed solely as a defense-in-depth filter that dropped reasoning/thinking parts when the compaction model differed from the conversation model, avoiding Anthropic HTTP 400s from orphaned thinking signatures. * Upstream PR anomalyco#25303 (29ec077), landed after our last sync base 26cc537, now performs this strip natively: the live path skips reasoning parts on model mismatch in message-v2.ts:373-380 and compact-v2 does the same in to-llm-message.ts:76-81. Our base only stripped providerMetadata while still emitting the part, so upstream now fixes the exact bug we patched, more robustly. * Keeping the branch would mean re-authoring a maintained feature against the heavily rewritten upstream compaction code for zero behavioral benefit, so it is dropped during the 2026-06-10 catch-up (bd epic opencode-k4t). * Retirement is at the manifest level only and reversible: the ref is archived at tag archive/compaction-thinking-signature-2026-06-10.
avion23 pushed a commit
to avion23/opencode
that referenced
this pull request
Jun 10, 2026
isaacfinnegan pushed a commit
to isaacfinnegan/whispercode
that referenced
this pull request
Jun 15, 2026
This was referenced Jun 16, 2026
This PR leads to mass KV cache invalidation when switching between models (the only way to specify different model params between plan/build/compact). Prior context should never be mass-edited outside of a compaction event, else the user potentially racks up costs with KV cache misses / large new prefill computation. I submitted a PR to retain reasoning while still addressing the signature issue: #32604 |
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 freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
Closes#24648
Instead of passing different model reasoning traces as is without metadata, convert them to assistant text and drop their metadata