amazon-bedrock-haystack 6.3.0__tar.gz → 6.4.0__tar.gz

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/CHANGELOG.md

@@ -1,5 +1,18 @@
 # Changelog
 
+## [integrations/amazon_bedrock-v6.3.0] - 2026-01-28
+
+### 🌀 Miscellaneous
+
+- Feat: Bedrock - support images in tool results (#2783)
+
+## [integrations/amazon_bedrock-v6.2.1] - 2026-01-15
+
+### 🐛 Bug Fixes
+
+- None value handling of flattened generation kwargs for AmazonBedrockChatGenerator (#2752)
+
+
 ## [integrations/amazon_bedrock-v6.2.0] - 2026-01-13
 
 ### 🚀 Features

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: amazon-bedrock-haystack
-Version: 6.3.0
+Version: 6.4.0
 Summary: An integration of AWS S3 and Bedrock as a Downloader and Generator components.
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/amazon_bedrock#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/downloaders/s3/s3_downloader.py

@@ -11,7 +11,7 @@ from typing import Any
 from botocore.config import Config
 from haystack import component, default_from_dict, default_to_dict, logging
 from haystack.dataclasses import Document
-from haystack.utils.auth import Secret, deserialize_secrets_inplace
+from haystack.utils.auth import Secret
 from haystack.utils.callable_serialization import deserialize_callable, serialize_callable
 
 from haystack_integrations.common.amazon_bedrock.utils import get_aws_session
@@ -233,11 +233,11 @@ class S3Downloader:
 
         return default_to_dict(
             self,
-            aws_access_key_id=self.aws_access_key_id.to_dict() if self.aws_access_key_id else None,
-            aws_secret_access_key=self.aws_secret_access_key.to_dict() if self.aws_secret_access_key else None,
-            aws_session_token=self.aws_session_token.to_dict() if self.aws_session_token else None,
-            aws_region_name=self.aws_region_name.to_dict() if self.aws_region_name else None,
-            aws_profile_name=self.aws_profile_name.to_dict() if self.aws_profile_name else None,
+            aws_access_key_id=self.aws_access_key_id,
+            aws_secret_access_key=self.aws_secret_access_key,
+            aws_session_token=self.aws_session_token,
+            aws_region_name=self.aws_region_name,
+            aws_profile_name=self.aws_profile_name,
             file_root_path=str(self.file_root_path),
             max_workers=self.max_workers,
             max_cache_size=self.max_cache_size,
@@ -260,8 +260,4 @@ class S3Downloader:
             data["init_parameters"]["s3_key_generation_function"] = deserialize_callable(
                 s3_key_generation_function_name
             )
-        deserialize_secrets_inplace(
-            data["init_parameters"],
-            ["aws_access_key_id", "aws_secret_access_key", "aws_session_token", "aws_region_name", "aws_profile_name"],
-        )
         return default_from_dict(cls, data)

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/embedders/amazon_bedrock/document_embedder.py

@@ -6,7 +6,7 @@ from botocore.config import Config
 from botocore.exceptions import ClientError
 from haystack import component, default_from_dict, default_to_dict, logging
 from haystack.dataclasses import Document
-from haystack.utils.auth import Secret, deserialize_secrets_inplace
+from haystack.utils.auth import Secret
 from tqdm import tqdm
 
 from haystack_integrations.common.amazon_bedrock.errors import (
@@ -257,11 +257,11 @@ class AmazonBedrockDocumentEmbedder:
         """
         return default_to_dict(
             self,
-            aws_access_key_id=self.aws_access_key_id.to_dict() if self.aws_access_key_id else None,
-            aws_secret_access_key=self.aws_secret_access_key.to_dict() if self.aws_secret_access_key else None,
-            aws_session_token=self.aws_session_token.to_dict() if self.aws_session_token else None,
-            aws_region_name=self.aws_region_name.to_dict() if self.aws_region_name else None,
-            aws_profile_name=self.aws_profile_name.to_dict() if self.aws_profile_name else None,
+            aws_access_key_id=self.aws_access_key_id,
+            aws_secret_access_key=self.aws_secret_access_key,
+            aws_session_token=self.aws_session_token,
+            aws_region_name=self.aws_region_name,
+            aws_profile_name=self.aws_profile_name,
             model=self.model,
             batch_size=self.batch_size,
             progress_bar=self.progress_bar,
@@ -281,8 +281,4 @@ class AmazonBedrockDocumentEmbedder:
         :returns:
             Deserialized component.
         """
-        deserialize_secrets_inplace(
-            data["init_parameters"],
-            ["aws_access_key_id", "aws_secret_access_key", "aws_session_token", "aws_region_name", "aws_profile_name"],
-        )
         return default_from_dict(cls, data)

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/embedders/amazon_bedrock/document_image_embedder.py

@@ -16,7 +16,7 @@ from haystack.components.converters.image.image_utils import (
     _PDFPageInfo,
 )
 from haystack.dataclasses import ByteStream
-from haystack.utils.auth import Secret, deserialize_secrets_inplace
+from haystack.utils.auth import Secret
 from tqdm import tqdm
 
 from haystack_integrations.common.amazon_bedrock.errors import (
@@ -178,11 +178,11 @@ class AmazonBedrockDocumentImageEmbedder:
             file_path_meta_field=self.file_path_meta_field,
             root_path=self.root_path,
             model=self.model,
-            aws_access_key_id=self.aws_access_key_id.to_dict() if self.aws_access_key_id else None,
-            aws_secret_access_key=self.aws_secret_access_key.to_dict() if self.aws_secret_access_key else None,
-            aws_session_token=self.aws_session_token.to_dict() if self.aws_session_token else None,
-            aws_region_name=self.aws_region_name.to_dict() if self.aws_region_name else None,
-            aws_profile_name=self.aws_profile_name.to_dict() if self.aws_profile_name else None,
+            aws_access_key_id=self.aws_access_key_id,
+            aws_secret_access_key=self.aws_secret_access_key,
+            aws_session_token=self.aws_session_token,
+            aws_region_name=self.aws_region_name,
+            aws_profile_name=self.aws_profile_name,
             progress_bar=self.progress_bar,
             boto3_config=self.boto3_config,
             image_size=self.image_size,
@@ -200,17 +200,6 @@ class AmazonBedrockDocumentImageEmbedder:
         :returns:
             Deserialized component.
         """
-        init_params = data["init_parameters"]
-        deserialize_secrets_inplace(
-            init_params,
-            keys=[
-                "aws_access_key_id",
-                "aws_secret_access_key",
-                "aws_session_token",
-                "aws_region_name",
-                "aws_profile_name",
-            ],
-        )
         return default_from_dict(cls, data)
 
     @component.output_types(documents=list[Document])

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/embedders/amazon_bedrock/text_embedder.py

@@ -4,7 +4,7 @@ from typing import Any
 from botocore.config import Config
 from botocore.exceptions import ClientError
 from haystack import component, default_from_dict, default_to_dict, logging
-from haystack.utils.auth import Secret, deserialize_secrets_inplace
+from haystack.utils.auth import Secret
 
 from haystack_integrations.common.amazon_bedrock.errors import (
     AmazonBedrockConfigurationError,
@@ -180,11 +180,11 @@ class AmazonBedrockTextEmbedder:
         """
         return default_to_dict(
             self,
-            aws_access_key_id=self.aws_access_key_id.to_dict() if self.aws_access_key_id else None,
-            aws_secret_access_key=self.aws_secret_access_key.to_dict() if self.aws_secret_access_key else None,
-            aws_session_token=self.aws_session_token.to_dict() if self.aws_session_token else None,
-            aws_region_name=self.aws_region_name.to_dict() if self.aws_region_name else None,
-            aws_profile_name=self.aws_profile_name.to_dict() if self.aws_profile_name else None,
+            aws_access_key_id=self.aws_access_key_id,
+            aws_secret_access_key=self.aws_secret_access_key,
+            aws_session_token=self.aws_session_token,
+            aws_region_name=self.aws_region_name,
+            aws_profile_name=self.aws_profile_name,
             model=self.model,
             boto3_config=self.boto3_config,
             **self.kwargs,
@@ -200,8 +200,4 @@ class AmazonBedrockTextEmbedder:
         :returns:
             Deserialized component.
         """
-        deserialize_secrets_inplace(
-            data["init_parameters"],
-            ["aws_access_key_id", "aws_secret_access_key", "aws_session_token", "aws_region_name", "aws_profile_name"],
-        )
         return default_from_dict(cls, data)

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/generators/amazon_bedrock/chat/chat_generator.py

@@ -13,7 +13,7 @@ from haystack.tools import (
     flatten_tools_or_toolsets,
     serialize_tools_or_toolset,
 )
-from haystack.utils.auth import Secret, deserialize_secrets_inplace
+from haystack.utils.auth import Secret
 from haystack.utils.callable_serialization import deserialize_callable, serialize_callable
 
 from haystack_integrations.common.amazon_bedrock.errors import (
@@ -27,6 +27,7 @@ from haystack_integrations.components.generators.amazon_bedrock.chat.utils impor
     _parse_completion_response,
     _parse_streaming_response,
     _parse_streaming_response_async,
+    _validate_and_format_cache_point,
     _validate_guardrail_config,
 )
 
@@ -142,6 +143,12 @@ class AmazonBedrockChatGenerator:
       and `aws_region_name` as environment variables or pass them as
      [Secret](https://docs.haystack.deepset.ai/docs/secret-management) arguments. Make sure the region you set
     supports Amazon Bedrock.
+
+    This component supports prompt caching. You can use the `tools_cachepoint_config` parameter to configure the cache
+    point for tools.
+    To cache messages, you can use the `cachePoint` key in `ChatMessage.meta` attribute.
+    Example: `ChatMessage.from_user("Long message...", meta={"cachePoint": {"type": "default"}})`
+    For more information, see the [Amazon Bedrock documentation](https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html).
     """
 
     def __init__(
@@ -160,6 +167,7 @@ class AmazonBedrockChatGenerator:
         tools: ToolsType | None = None,
         *,
         guardrail_config: dict[str, str] | None = None,
+        tools_cachepoint_config: dict[str, str] | None = None,
     ) -> None:
         """
         Initializes the `AmazonBedrockChatGenerator` with the provided parameters. The parameters are passed to the
@@ -201,6 +209,10 @@ class AmazonBedrockChatGenerator:
             See the
             [Guardrails Streaming documentation](https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails-streaming.html)
             for more information.
+        :param tools_cachepoint_config: Optional configuration to use prompt caching for tools.
+            The dictionary must match the
+            [CachePointBlock schema](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_CachePointBlock.html).
+            Example: `{"type": "default", "ttl": "5m"}`
 
 
         :raises ValueError: If the model name is empty or None.
@@ -225,6 +237,10 @@ class AmazonBedrockChatGenerator:
         _validate_guardrail_config(guardrail_config=guardrail_config, streaming=streaming_callback is not None)
         self.guardrail_config = guardrail_config
 
+        self.tools_cachepoint_config = (
+            _validate_and_format_cache_point(tools_cachepoint_config) if tools_cachepoint_config else None
+        )
+
         def resolve_secret(secret: Secret | None) -> str | None:
             return secret.resolve_value() if secret else None
 
@@ -299,17 +315,18 @@ class AmazonBedrockChatGenerator:
         callback_name = serialize_callable(self.streaming_callback) if self.streaming_callback else None
         return default_to_dict(
             self,
-            aws_access_key_id=self.aws_access_key_id.to_dict() if self.aws_access_key_id else None,
-            aws_secret_access_key=self.aws_secret_access_key.to_dict() if self.aws_secret_access_key else None,
-            aws_session_token=self.aws_session_token.to_dict() if self.aws_session_token else None,
-            aws_region_name=self.aws_region_name.to_dict() if self.aws_region_name else None,
-            aws_profile_name=self.aws_profile_name.to_dict() if self.aws_profile_name else None,
+            aws_access_key_id=self.aws_access_key_id,
+            aws_secret_access_key=self.aws_secret_access_key,
+            aws_session_token=self.aws_session_token,
+            aws_region_name=self.aws_region_name,
+            aws_profile_name=self.aws_profile_name,
             model=self.model,
             generation_kwargs=self.generation_kwargs,
             streaming_callback=callback_name,
             boto3_config=self.boto3_config,
             tools=serialize_tools_or_toolset(self.tools),
             guardrail_config=self.guardrail_config,
+            tools_cachepoint_config=self.tools_cachepoint_config,
         )
 
     @classmethod
@@ -331,10 +348,6 @@ class AmazonBedrockChatGenerator:
         serialized_callback_handler = init_params.get("streaming_callback")
         if serialized_callback_handler:
             data["init_parameters"]["streaming_callback"] = deserialize_callable(serialized_callback_handler)
-        deserialize_secrets_inplace(
-            data["init_parameters"],
-            ["aws_access_key_id", "aws_secret_access_key", "aws_session_token", "aws_region_name", "aws_profile_name"],
-        )
         deserialize_tools_or_toolset_inplace(data["init_parameters"], key="tools")
         return default_from_dict(cls, data)
 
@@ -389,7 +402,7 @@ class AmazonBedrockChatGenerator:
         tool_config = merged_kwargs.pop("toolConfig", None)
         if flattened_tools:
             # Format Haystack tools to Bedrock format
-            tool_config = _format_tools(flattened_tools)
+            tool_config = _format_tools(flattened_tools, tools_cachepoint_config=self.tools_cachepoint_config)
 
         # Any remaining kwargs go to additionalModelRequestFields
         additional_fields = merged_kwargs if merged_kwargs else None

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/generators/amazon_bedrock/chat/utils.py

@@ -40,7 +40,9 @@ FINISH_REASON_MAPPING: dict[str, FinishReason] = {
 
 
 # Haystack to Bedrock util methods
-def _format_tools(tools: list[Tool] | None = None) -> dict[str, Any] | None:
+def _format_tools(
+    tools: list[Tool] | None = None, tools_cachepoint_config: dict[str, dict[str, str]] | None = None
+) -> dict[str, Any] | None:
     """
     Format Haystack Tool(s) to Amazon Bedrock toolConfig format.
 
@@ -51,13 +53,16 @@ def _format_tools(tools: list[Tool] | None = None) -> dict[str, Any] | None:
     if not tools:
         return None
 
-    tool_specs = []
+    tool_specs: list[dict[str, Any]] = []
     for tool in tools:
         tool_specs.append(
             {"toolSpec": {"name": tool.name, "description": tool.description, "inputSchema": {"json": tool.parameters}}}
         )
 
-    return {"tools": tool_specs} if tool_specs else None
+    if tools_cachepoint_config:
+        tool_specs.append({"cachePoint": tools_cachepoint_config})
+
+    return {"tools": tool_specs}
 
 
 def _convert_image_content_to_bedrock_format(image_content: ImageContent) -> dict[str, Any]:
@@ -181,20 +186,23 @@ def _repair_tool_result_messages(bedrock_formatted_messages: list[dict[str, Any]
         original_idx = None
         for tool_call_id in tool_call_ids:
             for idx, tool_result in tool_result_messages:
-                tool_result_contents = [c for c in tool_result["content"] if "toolResult" in c]
+                tool_result_contents = [c for c in tool_result["content"] if "toolResult" in c or "cachePoint" in c]
                 for content in tool_result_contents:
-                    if content["toolResult"]["toolUseId"] == tool_call_id:
+                    if "toolResult" in content and content["toolResult"]["toolUseId"] == tool_call_id:
                         regrouped_tool_result.append(content)
                         # Keep track of the original index of the last tool result message
                         original_idx = idx
+                    elif "cachePoint" in content and content not in regrouped_tool_result:
+                        regrouped_tool_result.append(content)
+
         if regrouped_tool_result and original_idx is not None:
             repaired_tool_result_prompts.append((original_idx, {"role": "user", "content": regrouped_tool_result}))
 
     # Remove the tool result messages from bedrock_formatted_messages
     bedrock_formatted_messages_minus_tool_results: list[tuple[int, Any]] = []
     for idx, msg in enumerate(bedrock_formatted_messages):
-        # Assumes the content of tool result messages only contains 'toolResult': {...} objects (e.g. no 'text')
-        if msg.get("content") and "toolResult" not in msg["content"][0]:
+        # Filter out messages that contain toolResult (they are handled by repaired_tool_result_prompts)
+        if msg.get("content") and not any("toolResult" in c for c in msg["content"]):
             bedrock_formatted_messages_minus_tool_results.append((idx, msg))
 
     # Add the repaired tool result messages and sort to maintain the correct order
@@ -251,6 +259,32 @@ def _format_text_image_message(message: ChatMessage) -> dict[str, Any]:
     return {"role": message.role.value, "content": bedrock_content_blocks}
 
 
+def _validate_and_format_cache_point(cache_point: dict[str, str] | None) -> dict[str, dict[str, str]] | None:
+    """
+    Validate and format a cache point dictionary.
+
+    Schema available at https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_CachePointBlock.html
+
+    :param cache_point: Cache point dictionary to validate and format.
+    :returns: Dictionary in Bedrock cachePoint format or None if no cache point is provided.
+    :raises ValueError: If cache point is not valid.
+    """
+    if not cache_point:
+        return None
+
+    if "type" not in cache_point or cache_point["type"] != "default":
+        err_msg = "Cache point must have a 'type' key with value 'default'."
+        raise ValueError(err_msg)
+    if not set(cache_point).issubset({"type", "ttl"}):
+        err_msg = "Cache point can only contain 'type' and 'ttl' keys."
+        raise ValueError(err_msg)
+    if "ttl" in cache_point and cache_point["ttl"] not in ("5m", "1h"):
+        err_msg = "Cache point 'ttl' must be one of '5m', '1h'."
+        raise ValueError(err_msg)
+
+    return {"cachePoint": cache_point}
+
+
 def _format_messages(messages: list[ChatMessage]) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
     """
     Format a list of Haystack ChatMessages to the format expected by Bedrock API.
@@ -264,21 +298,30 @@ def _format_messages(messages: list[ChatMessage]) -> tuple[list[dict[str, Any]],
               non_system_messages is a list of properly formatted message dictionaries.
     """
     # Separate system messages, tool calls, and tool results
-    system_prompts = []
+    system_prompts: list[dict[str, Any]] = []
     bedrock_formatted_messages = []
     for msg in messages:
+        cache_point = _validate_and_format_cache_point(msg.meta.get("cachePoint"))
         if msg.is_from(ChatRole.SYSTEM):
             # Assuming system messages can only contain text
             # Don't need to track idx since system_messages are handled separately
             system_prompts.append({"text": msg.text})
-        elif msg.tool_calls:
-            bedrock_formatted_messages.append(_format_tool_call_message(msg))
+            if cache_point:
+                system_prompts.append(cache_point)
+            continue
+
+        if msg.tool_calls:
+            formatted_msg = _format_tool_call_message(msg)
         elif msg.tool_call_results:
-            bedrock_formatted_messages.append(_format_tool_result_message(msg))
+            formatted_msg = _format_tool_result_message(msg)
         else:
-            bedrock_formatted_messages.append(_format_text_image_message(msg))
+            formatted_msg = _format_text_image_message(msg)
+        if cache_point:
+            formatted_msg["content"].append(cache_point)
+        bedrock_formatted_messages.append(formatted_msg)
 
     repaired_bedrock_formatted_messages = _repair_tool_result_messages(bedrock_formatted_messages)
+
     return system_prompts, repaired_bedrock_formatted_messages
 
 
@@ -310,6 +353,9 @@ def _parse_completion_response(response_body: dict[str, Any], model: str) -> lis
                     "prompt_tokens": response_body.get("usage", {}).get("inputTokens", 0),
                     "completion_tokens": response_body.get("usage", {}).get("outputTokens", 0),
                     "total_tokens": response_body.get("usage", {}).get("totalTokens", 0),
+                    "cache_read_input_tokens": response_body.get("usage", {}).get("cacheReadInputTokens", 0),
+                    "cache_write_input_tokens": response_body.get("usage", {}).get("cacheWriteInputTokens", 0),
+                    "cache_details": response_body.get("usage", {}).get("CacheDetails", {}),
                 },
             }
             # guardrail trace
@@ -461,6 +507,9 @@ def _convert_event_to_streaming_chunk(
                 "prompt_tokens": usage.get("inputTokens", 0),
                 "completion_tokens": usage.get("outputTokens", 0),
                 "total_tokens": usage.get("totalTokens", 0),
+                "cache_read_input_tokens": usage.get("cacheReadInputTokens", 0),
+                "cache_write_input_tokens": usage.get("cacheWriteInputTokens", 0),
+                "cache_details": usage.get("cacheDetails", {}),
             }
         if "trace" in event_meta:
             chunk_meta["trace"] = event_meta["trace"]

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/generators/amazon_bedrock/generator.py

@@ -8,7 +8,7 @@ from botocore.config import Config
 from botocore.exceptions import ClientError
 from haystack import component, default_from_dict, default_to_dict, logging
 from haystack.dataclasses import StreamingChunk
-from haystack.utils import Secret, deserialize_callable, deserialize_secrets_inplace, serialize_callable
+from haystack.utils import Secret, deserialize_callable, serialize_callable
 
 from haystack_integrations.common.amazon_bedrock.errors import (
     AmazonBedrockConfigurationError,
@@ -287,11 +287,11 @@ class AmazonBedrockGenerator:
         callback_name = serialize_callable(self.streaming_callback) if self.streaming_callback else None
         return default_to_dict(
             self,
-            aws_access_key_id=self.aws_access_key_id.to_dict() if self.aws_access_key_id else None,
-            aws_secret_access_key=self.aws_secret_access_key.to_dict() if self.aws_secret_access_key else None,
-            aws_session_token=self.aws_session_token.to_dict() if self.aws_session_token else None,
-            aws_region_name=self.aws_region_name.to_dict() if self.aws_region_name else None,
-            aws_profile_name=self.aws_profile_name.to_dict() if self.aws_profile_name else None,
+            aws_access_key_id=self.aws_access_key_id,
+            aws_secret_access_key=self.aws_secret_access_key,
+            aws_session_token=self.aws_session_token,
+            aws_region_name=self.aws_region_name,
+            aws_profile_name=self.aws_profile_name,
             model=self.model,
             max_length=self.max_length,
             streaming_callback=callback_name,
@@ -310,10 +310,6 @@ class AmazonBedrockGenerator:
         :returns:
             Deserialized component.
         """
-        deserialize_secrets_inplace(
-            data["init_parameters"],
-            ["aws_access_key_id", "aws_secret_access_key", "aws_session_token", "aws_region_name", "aws_profile_name"],
-        )
         init_params = data.get("init_parameters", {})
         serialized_callback_handler = init_params.get("streaming_callback")
         if serialized_callback_handler:

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/src/haystack_integrations/components/rankers/amazon_bedrock/ranker.py

@@ -3,7 +3,7 @@ from typing import Any
 
 from botocore.exceptions import ClientError
 from haystack import Document, component, default_from_dict, default_to_dict, logging
-from haystack.utils import Secret, deserialize_secrets_inplace
+from haystack.utils import Secret
 
 from haystack_integrations.common.amazon_bedrock.errors import (
     AmazonBedrockConfigurationError,
@@ -133,11 +133,11 @@ class AmazonBedrockRanker:
         return default_to_dict(
             self,
             model=self.model_name,
-            aws_access_key_id=self.aws_access_key_id.to_dict() if self.aws_access_key_id else None,
-            aws_secret_access_key=self.aws_secret_access_key.to_dict() if self.aws_secret_access_key else None,
-            aws_session_token=self.aws_session_token.to_dict() if self.aws_session_token else None,
-            aws_region_name=self.aws_region_name.to_dict() if self.aws_region_name else None,
-            aws_profile_name=self.aws_profile_name.to_dict() if self.aws_profile_name else None,
+            aws_access_key_id=self.aws_access_key_id,
+            aws_secret_access_key=self.aws_secret_access_key,
+            aws_session_token=self.aws_session_token,
+            aws_region_name=self.aws_region_name,
+            aws_profile_name=self.aws_profile_name,
             top_k=self.top_k,
             max_chunks_per_doc=self.max_chunks_per_doc,
             meta_fields_to_embed=self.meta_fields_to_embed,
@@ -154,10 +154,6 @@ class AmazonBedrockRanker:
         :returns:
             The deserialized component.
         """
-        deserialize_secrets_inplace(
-            data["init_parameters"],
-            ["aws_access_key_id", "aws_secret_access_key", "aws_session_token", "aws_region_name", "aws_profile_name"],
-        )
         return default_from_dict(cls, data)
 
     def _prepare_bedrock_input_docs(self, documents: list[Document]) -> list[str]:

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/tests/test_chat_generator.py

@@ -39,6 +39,10 @@ MODELS_TO_TEST_WITH_THINKING = [
     "us.anthropic.claude-sonnet-4-20250514-v1:0",
 ]
 
+MODELS_TO_TEST_WITH_PROMPT_CACHING = [
+    "amazon.nova-micro-v1:0"  # cheap, fast model
+]
+
 
 def hello_world():
     return "Hello, World!"
@@ -164,6 +168,7 @@ class TestAmazonBedrockChatGenerator:
                 "boto3_config": boto3_config,
                 "tools": None,
                 "guardrail_config": {"guardrailIdentifier": "test", "guardrailVersion": "test"},
+                "tools_cachepoint_config": None,
             },
         }
 
@@ -298,6 +303,7 @@ class TestAmazonBedrockChatGenerator:
                             }
                         ],
                         "guardrail_config": None,
+                        "tools_cachepoint_config": None,
                     },
                 }
             },
@@ -945,6 +951,28 @@ class TestAmazonBedrockChatGeneratorInference:
         assert "trace" in results["replies"][0].meta
         assert "guardrail" in results["replies"][0].meta["trace"]
 
+    @pytest.mark.parametrize("streaming_callback", [None, print_streaming_chunk])
+    @pytest.mark.parametrize("model_name", MODELS_TO_TEST_WITH_PROMPT_CACHING)
+    def test_prompt_caching_live_run_with_user_message(self, model_name, streaming_callback):
+        generator = AmazonBedrockChatGenerator(model=model_name, streaming_callback=streaming_callback)
+
+        system_message = ChatMessage.from_system("Always respond with: 'Life is beautiful' (and nothing else).")
+
+        user_message = ChatMessage.from_user(
+            "User message that should be long enough to cache. " * 100, meta={"cachePoint": {"type": "default"}}
+        )
+        messages = [system_message, user_message]
+        result = generator.run(messages=messages)
+
+        assert "replies" in result
+        assert len(result["replies"]) == 1
+        usage = result["replies"][0].meta["usage"]
+
+        # tests run in parallel based on the workflow matrix, so this request should either hit the cache (read tokens)
+        # or populate it (write tokens)
+        assert usage["cache_read_input_tokens"] > 1000 or usage["cache_write_input_tokens"] > 1000
+        assert "cache_details" in usage
+
     @pytest.mark.parametrize("model_name", [MODELS_TO_TEST_WITH_TOOLS[0]])  # just one model is enough
     def test_pipeline_with_amazon_bedrock_chat_generator(self, model_name, tools):
         """

{amazon_bedrock_haystack-6.3.0 → amazon_bedrock_haystack-6.4.0}/tests/test_chat_generator_utils.py

@@ -22,6 +22,7 @@ from haystack_integrations.components.generators.amazon_bedrock.chat.utils impor
     _format_tools,
     _parse_completion_response,
     _parse_streaming_response,
+    _validate_and_format_cache_point,
     _validate_guardrail_config,
 )
 
@@ -59,7 +60,7 @@ def tools():
 
 class TestAmazonBedrockChatGeneratorUtils:
     def test_format_tools(self, tools):
-        formatted_tool = _format_tools(tools)
+        formatted_tool = _format_tools(tools, tools_cachepoint_config={"type": "default"})
         assert formatted_tool == {
             "tools": [
                 {
@@ -84,7 +85,8 @@ class TestAmazonBedrockChatGeneratorUtils:
                         },
                     }
                 },
-            ]
+                {"cachePoint": {"type": "default"}},
+            ],
         }
 
     def test_format_messages(self):
@@ -121,6 +123,52 @@ class TestAmazonBedrockChatGeneratorUtils:
             {"role": "assistant", "content": [{"text": "The weather in Paris is sunny and 25°C."}]},
         ]
 
+    def test_format_messages_with_cache_point(self):
+        meta = {"cachePoint": {"type": "default"}}
+
+        messages = [
+            ChatMessage.from_system("\\nYou are a helpful assistant, be super brief in your responses.", meta=meta),
+            ChatMessage.from_user("What is the weather in Paris?", meta=meta),
+            ChatMessage.from_assistant(
+                tool_calls=[ToolCall(id="123", tool_name="weather", arguments={"city": "Paris"})], meta=meta
+            ),
+            ChatMessage.from_tool(
+                tool_result="Sunny and 25°C",
+                origin=ToolCall(id="123", tool_name="weather", arguments={"city": "Paris"}),
+                meta=meta,
+            ),
+            ChatMessage.from_assistant("The weather in Paris is sunny and 25°C.", meta=meta),
+        ]
+        formatted_system_prompts, formatted_messages = _format_messages(messages)
+        assert formatted_system_prompts == [
+            {"text": "\\nYou are a helpful assistant, be super brief in your responses."},
+            {"cachePoint": {"type": "default"}},
+        ]
+        assert formatted_messages == [
+            {
+                "role": "user",
+                "content": [{"text": "What is the weather in Paris?"}, {"cachePoint": {"type": "default"}}],
+            },
+            {
+                "role": "assistant",
+                "content": [
+                    {"toolUse": {"toolUseId": "123", "name": "weather", "input": {"city": "Paris"}}},
+                    {"cachePoint": {"type": "default"}},
+                ],
+            },
+            {
+                "role": "user",
+                "content": [
+                    {"toolResult": {"toolUseId": "123", "content": [{"text": "Sunny and 25°C"}]}},
+                    {"cachePoint": {"type": "default"}},
+                ],
+            },
+            {
+                "role": "assistant",
+                "content": [{"text": "The weather in Paris is sunny and 25°C."}, {"cachePoint": {"type": "default"}}],
+            },
+        ]
+
     def test_format_messages_tool_result_with_image(self):
         base64_image = (
             "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8/5+hHgAHggJ/PchI7wAAAABJRU5ErkJggg=="
@@ -420,7 +468,14 @@ class TestAmazonBedrockChatGeneratorUtils:
         text_response = {
             "output": {"message": {"role": "assistant", "content": [{"text": "This is a test response"}]}},
             "stopReason": "end_turn",
-            "usage": {"inputTokens": 10, "outputTokens": 20, "totalTokens": 30},
+            "usage": {
+                "inputTokens": 10,
+                "outputTokens": 20,
+                "totalTokens": 30,
+                "cacheReadInputTokens": 1000,
+                "cacheWriteInputTokens": 0,
+                "cacheDetails": {},
+            },
         }
 
         replies = _parse_completion_response(text_response, model)
@@ -430,7 +485,14 @@ class TestAmazonBedrockChatGeneratorUtils:
         assert replies[0].meta == {
             "model": model,
             "finish_reason": "stop",
-            "usage": {"prompt_tokens": 10, "completion_tokens": 20, "total_tokens": 30},
+            "usage": {
+                "prompt_tokens": 10,
+                "completion_tokens": 20,
+                "total_tokens": 30,
+                "cache_read_input_tokens": 1000,
+                "cache_write_input_tokens": 0,
+                "cache_details": {},
+            },
             "index": 0,
         }
 
@@ -457,7 +519,14 @@ class TestAmazonBedrockChatGeneratorUtils:
         assert replies[0].meta == {
             "model": model,
             "finish_reason": "tool_calls",
-            "usage": {"prompt_tokens": 15, "completion_tokens": 25, "total_tokens": 40},
+            "usage": {
+                "prompt_tokens": 15,
+                "completion_tokens": 25,
+                "total_tokens": 40,
+                "cache_read_input_tokens": 0,
+                "cache_write_input_tokens": 0,
+                "cache_details": {},
+            },
             "index": 0,
         }
 
@@ -488,7 +557,14 @@ class TestAmazonBedrockChatGeneratorUtils:
         assert replies[0].meta == {
             "model": model,
             "finish_reason": "stop",
-            "usage": {"prompt_tokens": 25, "completion_tokens": 35, "total_tokens": 60},
+            "usage": {
+                "prompt_tokens": 25,
+                "completion_tokens": 35,
+                "total_tokens": 60,
+                "cache_read_input_tokens": 0,
+                "cache_write_input_tokens": 0,
+                "cache_details": {},
+            },
             "index": 0,
         }
 
@@ -556,7 +632,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                 "model": "anthropic.claude-3-5-sonnet-20240620-v1:0",
                 "index": 0,
                 "finish_reason": "tool_calls",
-                "usage": {"prompt_tokens": 366, "completion_tokens": 134, "total_tokens": 500},
+                "usage": {
+                    "prompt_tokens": 366,
+                    "completion_tokens": 134,
+                    "total_tokens": 500,
+                    "cache_read_input_tokens": 0,
+                    "cache_write_input_tokens": 0,
+                    "cache_details": {},
+                },
             },
         )
         assert replies[0] == expected_message
@@ -610,6 +693,7 @@ class TestAmazonBedrockChatGeneratorUtils:
                 "totalTokens": 558,
                 "cacheReadInputTokens": 0,
                 "cacheWriteInputTokens": 0,
+                "cacheDetails": {},
             },
             "metrics": {"latencyMs": 4811},
         }
@@ -646,7 +730,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                 "model": "arn:aws:bedrock:us-east-1::inference-profile/us.anthropic.claude-3-7-sonnet-20250219-v1:0",
                 "index": 0,
                 "finish_reason": "tool_calls",
-                "usage": {"prompt_tokens": 412, "completion_tokens": 146, "total_tokens": 558},
+                "usage": {
+                    "prompt_tokens": 412,
+                    "completion_tokens": 146,
+                    "total_tokens": 558,
+                    "cache_read_input_tokens": 0,
+                    "cache_write_input_tokens": 0,
+                    "cache_details": {},
+                },
             },
         )
         assert replies[0] == expected_message
@@ -712,7 +803,14 @@ class TestAmazonBedrockChatGeneratorUtils:
             "model": model,
             "finish_reason": "content_filter",
             "index": 0,
-            "usage": {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0},
+            "usage": {
+                "prompt_tokens": 0,
+                "completion_tokens": 0,
+                "total_tokens": 0,
+                "cache_read_input_tokens": 0,
+                "cache_write_input_tokens": 0,
+                "cache_details": {},
+            },
             "trace": trace,
         }
 
@@ -784,7 +882,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                     "model": model,
                     "index": 0,
                     "finish_reason": "tool_calls",
-                    "usage": {"prompt_tokens": 364, "completion_tokens": 71, "total_tokens": 435},
+                    "usage": {
+                        "prompt_tokens": 364,
+                        "completion_tokens": 71,
+                        "total_tokens": 435,
+                        "cache_read_input_tokens": 0,
+                        "cache_write_input_tokens": 0,
+                        "cache_details": {},
+                    },
                 },
             )
         ]
@@ -852,7 +957,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                 meta={
                     "model": model,
                     "received_at": ANY,
-                    "usage": {"prompt_tokens": 364, "completion_tokens": 71, "total_tokens": 435},
+                    "usage": {
+                        "prompt_tokens": 364,
+                        "completion_tokens": 71,
+                        "total_tokens": 435,
+                        "cache_read_input_tokens": 0,
+                        "cache_write_input_tokens": 0,
+                        "cache_details": {},
+                    },
                 },
                 component_info=c_info,
             ),
@@ -976,7 +1088,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                     "model": "arn:aws:bedrock:us-east-1::inference-profile/us.anthropic.claude-sonnet-4-20250514-v1:0",
                     "index": 0,
                     "finish_reason": "tool_calls",
-                    "usage": {"prompt_tokens": 412, "completion_tokens": 104, "total_tokens": 516},
+                    "usage": {
+                        "prompt_tokens": 412,
+                        "completion_tokens": 104,
+                        "total_tokens": 516,
+                        "cache_read_input_tokens": 0,
+                        "cache_write_input_tokens": 0,
+                        "cache_details": {},
+                    },
                     "completion_start_time": ANY,
                 },
             )
@@ -1062,7 +1181,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                     "model": model,
                     "index": 0,
                     "finish_reason": "stop",
-                    "usage": {"prompt_tokens": 461, "completion_tokens": 138, "total_tokens": 599},
+                    "usage": {
+                        "prompt_tokens": 461,
+                        "completion_tokens": 138,
+                        "total_tokens": 599,
+                        "cache_read_input_tokens": 0,
+                        "cache_write_input_tokens": 0,
+                        "cache_details": {},
+                    },
                     "completion_start_time": ANY,
                 },
             )
@@ -1140,7 +1266,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                     "model": "anthropic.claude-3-5-sonnet-20240620-v1:0",
                     "index": 0,
                     "finish_reason": "tool_calls",
-                    "usage": {"prompt_tokens": 366, "completion_tokens": 83, "total_tokens": 449},
+                    "usage": {
+                        "prompt_tokens": 366,
+                        "completion_tokens": 83,
+                        "total_tokens": 449,
+                        "cache_read_input_tokens": 0,
+                        "cache_write_input_tokens": 0,
+                        "cache_details": {},
+                    },
                     "completion_start_time": ANY,
                 },
             )
@@ -1216,7 +1349,14 @@ class TestAmazonBedrockChatGeneratorUtils:
                     "model": model,
                     "index": 0,
                     "finish_reason": "content_filter",
-                    "usage": {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0},
+                    "usage": {
+                        "prompt_tokens": 0,
+                        "completion_tokens": 0,
+                        "total_tokens": 0,
+                        "cache_read_input_tokens": 0,
+                        "cache_write_input_tokens": 0,
+                        "cache_details": {},
+                    },
                     "trace": trace,
                 },
             )
@@ -1450,3 +1590,25 @@ class TestAmazonBedrockChatGeneratorUtils:
                 },
                 streaming=False,
             )
+
+    def test_validate_and_format_cache_point(self):
+        cache_point = _validate_and_format_cache_point(None)
+        assert cache_point is None
+
+        cache_point = _validate_and_format_cache_point({})
+        assert cache_point is None
+
+        cache_point = _validate_and_format_cache_point({"type": "default"})
+        assert cache_point == {"cachePoint": {"type": "default"}}
+
+        cache_point = _validate_and_format_cache_point({"type": "default", "ttl": "5m"})
+        assert cache_point == {"cachePoint": {"type": "default", "ttl": "5m"}}
+
+        with pytest.raises(ValueError, match=r"Cache point must have a 'type' key with value 'default'."):
+            _validate_and_format_cache_point({"invalid": "config"})
+
+        with pytest.raises(ValueError, match=r"Cache point must have a 'type' key with value 'default'."):
+            _validate_and_format_cache_point({"type": "invalid"})
+
+        with pytest.raises(ValueError, match=r"Cache point can only contain 'type' and 'ttl' keys."):
+            _validate_and_format_cache_point({"type": "default", "invalid": "config"})