th-server-12212: chat_stream 5xx-retry (core 0.16.2) + gpt-oss chips-parse#222
Merged
Conversation
Workspace cargo check clean across server + operator + ingestion + 5 adapters (no E0053). Confirms core 0.16.2 live on crates.io. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
extract_suggested_replies now also parses a trailing markdown "Suggested replies:" list, so models that ignore the <suggested_replies> marker (gpt-oss-120b — the TP agent's default model) still populate chips instead of shipping empty suggestedNextActions. Conservative: fires only when the header is followed by a pure bullet/numbered list to the end of the reply. Ships alongside the core 0.16.2 chat_stream 5xx-retry (dep already bumped in af6d3f2). Lockstep version bump to 1.22.12 across all published artifacts. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Cuts server 1.22.12, bundling two fixes for the chat/empty-reply + chips symptoms surfaced on the Transformation Posture agent (gpt-oss-120b).
1. chat_stream 5xx-retry (core 0.16.2)
chat_stream(every chat turn) retried only transient transport errors, then bailed immediately on any non-success HTTP status. A transient gateway 5xx (groq/LiteLLM 502/503 — as happened in the recent valkey→api-prime→litellm incident) propagated as anAGENT_ERRORand the widget rendered an empty reply. Core 0.16.2 (already published) makeschat_streamretrypolicy.retry_on_statusbefore reading any stream bytes, mirroringchat(). This bumps the workspace core dep 0.16.1→0.16.2.2. gpt-oss markdown-list chips-parse
gpt-oss-120b often ignores the
<suggested_replies>marker and ends its reply with a markdownSuggested replies:list, soextract_suggested_repliesfound no marker → empty chips. It now also parses that trailing list (conservative: only when the header is followed by a pure bullet/numbered list to the end). 4 new unit tests on the real gpt-oss shape.Validation (no Rust CI in this repo)
Local:
cargo check --workspaceclean with core 0.16.2;cargo test -p ...-server --lib suggestions::14/14;cargo test -p ...-core --lib chat_stream_retries(in core PR #79) pass; clippy clean for changed code.Lockstep version bump to 1.22.12 across all published artifacts via
version:bump.