mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-06 08:16:43 +00:00
5b2e345670
The `/v1/messages` endpoint's AnthropicStreamWrapper silently dropped the trigger chunk of every new content block on a detected block transition (e.g. text -> thinking, thinking -> text, text -> tool_use). The code emitted `content_block_stop` -> `content_block_start` and then returned without re-enqueuing `processed_chunk`, relying on an incorrect assumption that `content_block_start` already carries the trigger chunk's content. In practice, `_translate_streaming_openai_chunk_to_anthropic_content_block()` returns an empty `TextBlock(text="")` for text transitions, so the first characters of every new text/thinking block were lost on the wire. Symptoms for Bedrock Converse reasoning providers (MiniMax, Kimi, Claude extended thinking) using `/v1/messages` + `stream=true`: - Responses start mid-sentence (leading characters missing). - If the model emits the text as a single Bedrock chunk, the text block is streamed with zero `content_block_delta` events, causing clients like Claude Code CLI to see an empty response. - Non-streaming on the same deployment returns the full text correctly. Fix: after emitting the synthetic `content_block_stop` + `content_block_start` pair, re-enqueue `processed_chunk` when it is a non-empty `content_block_delta`. A small `_trigger_delta_has_content` helper inspects the standard Anthropic delta variants (`text`, `thinking`, `partial_json`, `signature`) so that empty tool_use openers (whose tool name is already carried by `content_block_start`) are intentionally skipped. Applied to both the sync (`__next__`) and async (`__anext__`) paths. Tests: - Add `test_first_chunk_on_block_transition.py` with a focused regression test exercising a text -> thinking -> text sequence (the exact Bedrock Converse reasoning-model pattern). The new tests fail on the unpatched `main` and pass after the fix, for both sync and async iterators. - Update the existing `test_parallel_tool_calls.py::test_anthropic_stream_wrapper_interleaved_tool_calls_and_text` expected sequence: two `content_block_delta` events for the interleaved text chunks were previously missing because the wrapper dropped them; the test now also asserts the text content is preserved verbatim. |
||
|---|---|---|
| .. | ||
| agent_tests | ||
| audio_tests | ||
| basic_proxy_startup_tests | ||
| batches_tests | ||
| benchmarks | ||
| code_coverage_tests | ||
| documentation_tests | ||
| enterprise | ||
| guardrails_tests | ||
| image_gen_tests | ||
| litellm | ||
| litellm-proxy-extras | ||
| litellm_core_utils | ||
| litellm_utils_tests | ||
| llm_responses_api_testing | ||
| llm_translation | ||
| load_tests | ||
| local_testing | ||
| logging_callback_tests | ||
| mcp_tests | ||
| multi_instance_e2e_tests | ||
| ocr_tests | ||
| old_proxy_tests/tests | ||
| openai_endpoints_tests | ||
| otel_tests | ||
| pass_through_tests | ||
| pass_through_unit_tests | ||
| proxy_admin_ui_tests | ||
| proxy_e2e_anthropic_messages_tests | ||
| proxy_e2e_azure_batches_tests | ||
| proxy_security_tests | ||
| proxy_unit_tests | ||
| router_unit_tests | ||
| scim_tests | ||
| search_tests | ||
| spend_tracking_tests | ||
| store_model_in_db_tests | ||
| test_litellm | ||
| unified_google_tests | ||
| vector_store_tests | ||
| windows_tests | ||
| __init__.py | ||
| gettysburg.wav | ||
| large_text.py | ||
| openai_batch_completions.jsonl | ||
| README.MD | ||
| test_budget_management.py | ||
| test_callbacks_on_proxy.py | ||
| test_config.py | ||
| test_debug_warning.py | ||
| test_default_encoding_non_root.py | ||
| test_end_users.py | ||
| test_entrypoint.py | ||
| test_fallbacks.py | ||
| test_gpt5_azure_temperature_support.py | ||
| test_health.py | ||
| test_keys.py | ||
| test_litellm_proxy_responses_config.py | ||
| test_logging.conf | ||
| test_models.py | ||
| test_new_vector_store_endpoints.py | ||
| test_openai_endpoints.py | ||
| test_organizations.py | ||
| test_otel_thread_leak.py | ||
| test_passthrough_endpoints.py | ||
| test_presidio_latency.py | ||
| test_proxy_server_non_root.py | ||
| test_ratelimit.py | ||
| test_resource_cleanup.py | ||
| test_service_logger_otel.py | ||
| test_spend_logs.py | ||
| test_team.py | ||
| test_team_logging.py | ||
| test_team_members.py | ||
| test_users.py | ||
In total litellm runs 1000+ tests
[02/20/2025] Update:
To make it easier to contribute and map what behavior is tested,
we've started mapping the litellm directory in tests/test_litellm
This folder can only run mock tests.