From caf9c1890ec2506024436cca841ef33698ba3ecd Mon Sep 17 00:00:00 2001 From: milan Date: Wed, 19 Aug 2026 21:33:20 +0000 Subject: [PATCH] fix(bedrock): enable tool search for Opus 4.8 and gen 5 Claude on Invoke Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../anthropic_claude3_transformation.py | 3 +- litellm/llms/bedrock/common_utils.py | 56 +++++++++++++++++++ .../anthropic_claude3_transformation.py | 42 +------------- ...odel_prices_and_context_window_backup.json | 18 ++++++ model_prices_and_context_window.json | 18 ++++++ ...ations_anthropic_claude3_transformation.py | 30 ++++++++++ .../test_anthropic_claude3_transformation.py | 9 +++ 7 files changed, 135 insertions(+), 41 deletions(-) diff --git a/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py b/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py index b8b07af59c6..b4eb1bdd17d 100644 --- a/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py +++ b/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py @@ -22,6 +22,7 @@ from litellm.llms.bedrock.common_utils import ( normalize_custom_field_on_tools, normalize_tool_input_schema_types_for_bedrock_invoke, pop_bedrock_invoke_output_config_format, + supports_tool_search_on_bedrock, ) from litellm.types.llms.anthropic import ANTHROPIC_TOOL_SEARCH_BETA_HEADER from litellm.types.llms.openai import AllMessageValues @@ -275,7 +276,7 @@ class AmazonAnthropicClaudeConfig(AmazonInvokeConfig, AnthropicConfig): if tool_search_used and not (programmatic_tool_calling_used or input_examples_used): beta_set.discard(ANTHROPIC_TOOL_SEARCH_BETA_HEADER) - if "opus-4" in model.lower() or "opus_4" in model.lower(): + if supports_tool_search_on_bedrock(model): beta_set.add("tool-search-tool-2025-10-19") auto_beta_list: Final = filter_and_transform_beta_headers( diff --git a/litellm/llms/bedrock/common_utils.py b/litellm/llms/bedrock/common_utils.py index 4ad20772ed0..ec6da833ae0 100644 --- a/litellm/llms/bedrock/common_utils.py +++ b/litellm/llms/bedrock/common_utils.py @@ -21,6 +21,7 @@ import httpx import litellm from litellm import verbose_logger +from litellm.llms.anthropic.common_utils import AnthropicModelInfo from litellm.llms.base_llm.anthropic_messages.transformation import ( BaseAnthropicMessagesConfig, ) @@ -742,6 +743,61 @@ def is_claude_4_5_on_bedrock(model: str) -> bool: ) +_BEDROCK_TOOL_SEARCH_MODEL_PATTERNS: Final = ( + "opus-4.5", + "opus_4.5", + "opus-4-5", + "opus_4_5", + "sonnet-4.5", + "sonnet_4.5", + "sonnet-4-5", + "sonnet_4_5", + "opus-4.6", + "opus_4.6", + "opus-4-6", + "opus_4_6", + "sonnet-4.6", + "sonnet_4.6", + "sonnet-4-6", + "sonnet_4_6", + "opus-4.7", + "opus_4.7", + "opus-4-7", + "opus_4_7", + "opus-4.8", + "opus_4.8", + "opus-4-8", + "opus_4_8", + "opus-5", + "opus_5", + "sonnet-5", + "sonnet_5", + "haiku-4.5", + "haiku_4.5", + "haiku-4-5", + "haiku_4_5", +) + + +def supports_tool_search_on_bedrock(model: str) -> bool: + """ + Whether Bedrock InvokeModel accepts the ``tool-search-tool-2025-10-19`` + beta for ``model``. + + The model map's ``supports_tool_search`` flag is authoritative when + ``model`` resolves to an entry that sets it; the name patterns cover ids + the map cannot resolve (ARNs, unlisted regional variants). + + Ref: https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool + """ + catalog: Final = AnthropicModelInfo._get_provider_resolved_capability(model, "supports_tool_search", "bedrock") + if catalog is not None: + return catalog + + model_lower: Final = model.lower() + return any(pattern in model_lower for pattern in _BEDROCK_TOOL_SEARCH_MODEL_PATTERNS) + + _BEDROCK_MODEL_VERSION_SUFFIX_RE: Final = re.compile(r"-v\d+(?::\d+)?$") diff --git a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py index 0161f4fadc9..bbea50da5fe 100644 --- a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py +++ b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py @@ -37,6 +37,7 @@ from litellm.llms.bedrock.common_utils import ( normalize_custom_field_on_tools, normalize_tool_input_schema_types_for_bedrock_invoke, pop_bedrock_invoke_output_config_format, + supports_tool_search_on_bedrock, ) from litellm.llms.bedrock.request_metadata import ( bedrock_request_metadata_headers, @@ -390,46 +391,7 @@ class AmazonAnthropicClaudeMessagesConfig( Returns: True if the model supports tool search on Bedrock """ - catalog: Final = AnthropicModelInfo._get_provider_resolved_capability(model, "supports_tool_search", "bedrock") - if catalog is not None: - return catalog - - model_lower: Final = model.lower() - - supported_patterns: Final = [ - # Opus 4.5 - "opus-4.5", - "opus_4.5", - "opus-4-5", - "opus_4_5", - # Sonnet 4.5 - "sonnet-4.5", - "sonnet_4.5", - "sonnet-4-5", - "sonnet_4_5", - # Opus 4.6 - "opus-4.6", - "opus_4.6", - "opus-4-6", - "opus_4_6", - # sonnet 4.6 - "sonnet-4.6", - "sonnet_4.6", - "sonnet-4-6", - "sonnet_4_6", - # Opus 4.7 - "opus-4.7", - "opus_4.7", - "opus-4-7", - "opus_4_7", - # Haiku 4.5 - "haiku-4.5", - "haiku_4.5", - "haiku-4-5", - "haiku_4_5", - ] - - return any(pattern in model_lower for pattern in supported_patterns) + return supports_tool_search_on_bedrock(model) def _get_tool_search_beta_header_for_bedrock( self, diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 07f9027313b..acfcc29844b 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -1535,6 +1535,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1571,6 +1572,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1607,6 +1609,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1643,6 +1646,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1679,6 +1683,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1715,6 +1720,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1751,6 +1757,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1788,6 +1795,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1825,6 +1833,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1862,6 +1871,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1899,6 +1909,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1936,6 +1947,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2006,6 +2018,7 @@ "cache_read_input_token_cost": 2e-07, "input_cost_per_token": 2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2043,6 +2056,7 @@ "cache_read_input_token_cost": 2e-07, "input_cost_per_token": 2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2080,6 +2094,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2117,6 +2132,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2154,6 +2170,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2191,6 +2208,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 07f9027313b..acfcc29844b 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -1535,6 +1535,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1571,6 +1572,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1607,6 +1609,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1643,6 +1646,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1679,6 +1683,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1715,6 +1720,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1751,6 +1757,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1788,6 +1795,7 @@ "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1825,6 +1833,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1862,6 +1871,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1899,6 +1909,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -1936,6 +1947,7 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2006,6 +2018,7 @@ "cache_read_input_token_cost": 2e-07, "input_cost_per_token": 2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2043,6 +2056,7 @@ "cache_read_input_token_cost": 2e-07, "input_cost_per_token": 2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2080,6 +2094,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2117,6 +2132,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2154,6 +2170,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, @@ -2191,6 +2208,7 @@ "cache_read_input_token_cost": 2.2e-07, "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", + "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, diff --git a/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py b/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py index 4c4c0e17a38..a852ee6f679 100644 --- a/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py +++ b/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py @@ -545,3 +545,33 @@ def test_output_format_removed_from_bedrock_invoke_request(): assert ( "output_format" not in result ), f"output_format should be removed for Bedrock Invoke, got keys: {result.keys()}" + + +@pytest.mark.parametrize( + "model, expected", + [ + pytest.param("us.anthropic.claude-sonnet-5-v99:9", True, id="sonnet_5"), + pytest.param("us.anthropic.claude-opus-5-v99:9", True, id="opus_5"), + pytest.param("us.anthropic.claude-opus-4-8-v99:9", True, id="opus_4_8"), + pytest.param("us.anthropic.claude-haiku-4-5-v99:9", True, id="haiku_4_5"), + pytest.param("us.anthropic.claude-opus-4-1-v99:9", False, id="opus_4_1_unsupported"), + ], +) +def test_bedrock_chat_invoke_tool_search_beta_header(model, expected): + """The chat Invoke path previously hardcoded an ``opus-4`` name check, which + both missed tool-search capable families (Haiku 4.5, Sonnet 5, Opus 5) and + wrongly sent the beta for Opus 4.1, which Anthropic documents as + unsupported. It now shares ``supports_tool_search_on_bedrock`` with the + /v1/messages path.""" + config = AmazonAnthropicClaudeConfig() + + betas = config._compute_bedrock_invoke_beta_headers( + model=model, + messages=[{"role": "user", "content": "test"}], + optional_params={ + "tools": [{"type": "tool_search_tool_regex_20251119", "name": "tool_search_tool_regex"}], + }, + headers={}, + ) + + assert ("tool-search-tool-2025-10-19" in betas) is expected diff --git a/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py b/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py index ce81edf3101..38834b62b4e 100644 --- a/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py +++ b/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py @@ -2714,6 +2714,9 @@ def test_filter_and_transform_beta_headers_passes_context_management_for_bedrock "us.anthropic.claude-haiku-4-5-20251001-v1:0", "us.anthropic.claude-sonnet-4-5-20250929-v1:0", "us.anthropic.claude-opus-4-7", + "us.anthropic.claude-opus-4-8", + "us.anthropic.claude-opus-5", + "us.anthropic.claude-sonnet-5", ], ) def test_bedrock_messages_tool_search_adds_beta_header(local_beta_headers_config, model): @@ -2726,6 +2729,9 @@ def test_bedrock_messages_tool_search_adds_beta_header(local_beta_headers_config Opus 4.7, so the beta was silently dropped for those models and every tool-search request failed. Verified live 2026-08-11: Bedrock returns 200 with ``server_tool_use`` for all three models once the beta is sent. + Opus 4.8, Opus 5, and Sonnet 5 verified live 2026-08-19: Bedrock + InvokeModel returns 200 with ``server_tool_use`` and + ``tool_search_tool_result`` blocks. """ from litellm.types.router import GenericLiteLLMParams @@ -2782,6 +2788,9 @@ def test_bedrock_messages_tool_search_model_map_flag_is_authoritative(local_mode "model, expected", [ pytest.param("us.anthropic.claude-opus-4-6-v99:9", True, id="unmapped_id_falls_back_to_patterns"), + pytest.param("us.anthropic.claude-opus-5-v99:9", True, id="unmapped_opus_5_falls_back_to_patterns"), + pytest.param("us.anthropic.claude-sonnet-5-v99:9", True, id="unmapped_sonnet_5_falls_back_to_patterns"), + pytest.param("us.anthropic.claude-opus-4-8-v99:9", True, id="unmapped_opus_4_8_falls_back_to_patterns"), pytest.param("anthropic.claude-3-5-sonnet-20240620-v1:0", False, id="mapped_entry_without_flag_no_pattern"), ], )