test(streaming): integration suite for SSE pipeline and translator#331
Merged
4 commits merged intomainfrom Apr 28, 2026
Merged
Conversation
added 4 commits
April 28, 2026 22:41
CI re-runs the full test suite (incl. doctests) on every PR via the .github/workflows/ci.yml tests job, so local pre-push duplication adds ~20 min per push without catching anything new. Pre-push hooks should be fast-fail; expensive checks belong on the CI server. Closes audit finding: silent productivity tax (pre-push duplication).
Documents the three-state intent (true/false/absent) of ProviderConfig.is_enabled and the dependency on deny_unknown_fields (added in the next commit) to reject typos like enbaled = false at parse time. Behaviour is unchanged; this is purely contractual clarity to support the silent-typo-killer audit. Closes audit finding: silent typo killer on provider config.
Adds #[serde(deny_unknown_fields)] to AppConfig and the major sub-structs (ProviderConfig, ModelConfig, TierConfig, RouterConfig, ScoringConfig, CacheConfig, BudgetConfig, DlpConfig, SecurityConfig). Without this guard, a typo like enbaled = false in a [[providers]] block silently parses (the unknown key is dropped) and the provider remains enabled with the wrong intent. With the guard, parsing fails loudly and the operator gets an actionable error pointing at the offending key. Tested with the full nextest suite (1268 tests) plus all doctests: no fixture, preset or example carries a stale field, so this is a pure tightening with no migration cost. Closes audit finding: silent typo killer on TOML config.
Each entry in DENIED_SECTIONS / DENIED_KEYS now carries a short justification table covering why it can not be hot-reloaded — either because the data is sensitive (credentials, DLP rules) or because the consumer is constructed once at process start (TLS listener, secret backend, TEE attestation, FIPS gate). Adds tee, fips, server.tls and secrets.backend to the deny-list so the documented "static-init" rationale matches actual behaviour. Also emits an INFO log on every denied attempt telling the operator to restart instead of expecting the silent reload to apply. Adds two unit tests covering the new deny entries (tee/fips sections and server.tls / secrets.backend keys) and asserts that sibling keys in the same sections remain editable. Closes audit finding: hot-reload UX (silent ignore of denied edits).
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.
Summary
tests/integration/streaming_test.rs(17 tests) covering the streaming path that the audit flagged as a critical gap.AnthropicToOpenAIStreamtranslator.LoggingSseStream, plus chunked HTTP reassembly against amockitoupstream.Coverage
Anthropic streaming format
anthropic_message_start_event_emitted_firstanthropic_content_block_start_then_deltas_then_stopanthropic_message_delta_carries_stop_reasonanthropic_message_stop_terminates_streamOpenAI streaming format
openai_chunk_first_emits_role_assistantopenai_chunks_carry_content_in_delta_contentopenai_final_chunk_has_finish_reasonopenai_data_done_terminates_streamTranslation Anthropic to OpenAI
anthropic_to_openai_translation_preserves_token_orderanthropic_tool_use_streaming_translates_to_openai_tool_calls_streamingopenai_chunk_role_user_round_trips_to_anthropic_user_messagestreaming_translation_handles_partial_json_in_tool_callsError and edge cases
streaming_returns_error_on_upstream_disconnectstreaming_cancellation_propagates_to_upstreamstreaming_mid_stream_error_emits_error_event_to_clientstreaming_token_count_accurate_on_partial_streamHTTP-level
translator_handles_chunked_sse_response_from_mock_server(mockito +SseStreamreassembly)Test plan
cargo check --testscargo fmt --checkcargo clippy --tests -- -D warningscargo nextest run --test lib integration::streaming_test::-- 17 passedcargo nextest run --test lib(full integration suite) -- 234 passed