feat: Add default OutputAdapter filters (post 2.0)

haystack/components/converters/output_adapter_filters.py

@@ -0,0 +1,111 @@
+import json
+from typing import Any, Dict, List
+
+from haystack.dataclasses import ChatMessage, ChatRole
+
+# pylint: disable=pointless-string-statement
+"""
+## Custom Filters for OutputAdapter
+
+This module defines a collection of custom filters for use with the `OutputAdapter` in processing and transforming
+data within a pipeline. These filters allow for dynamic data manipulation based on the specific needs of the
+pipeline components.
+
+### Defined Filters:
+
+- `change_role`: Adjusts the role of the last message in a list of chat messages.
+- `prepare_fc_params`: Prepares function call parameters for OpenAI functions.
+- `tojson`: Converts a string in JSON format to a Python dictionary.
+
+### Usage:
+
+To utilize these filters, import the `ALL_FILTERS` dictionary and pass it to the `OutputAdapter` as the
+`custom_filters` argument. You can either specify a subset of these filters or use all of them.
+
+### Examples:
+
+#### a) Using a Subset of Custom Filters:
+
+```python
+from haystack.components.converters import OutputAdapter
+from haystack.components.converters.output_adapter_filters import ALL_FILTERS
+
+# Use only the 'prepare_fc_params' filter
+custom_filters_subset = {"prepare_fc_params": ALL_FILTERS["prepare_fc_params"]}
+
+# Example pipeline component setup with one custom filter
+pipe.add_component("a2", OutputAdapter(template="{{fc_kwargs | prepare_fc_params}}",
+                                       output_type=Dict[str, Any],
+                                       custom_filters=custom_filters_subset))
+```
+
+#### b) Using All Filters:
+
+```python
+from haystack.components.converters import OutputAdapter
+from haystack.components.converters.output_adapter_filters import ALL_FILTERS
+
+# Example pipeline setup using all available custom filters
+pipe = Pipeline()
+pipe.add_component("fc_llm", OpenAIChatGenerator(model="gpt-3.5-turbo-0125"))
+pipe.add_component("spec_to_functions", OpenAPIServiceToFunctions())
+pipe.add_component("openapi_container", OpenAPIServiceConnector())
+pipe.add_component("a1", OutputAdapter(template="{{documents[0].content | tojson}}",
+                                       output_type=Dict[str, Any],
+                                       custom_filters=ALL_FILTERS))
+pipe.add_component("a2", OutputAdapter(template="{{fc_kwargs | prepare_fc_params}}",
+                                       output_type=Dict[str, Any],
+                                       custom_filters=ALL_FILTERS))
+# Additional components utilizing different filters from ALL_FILTERS
+```
+
+This approach enables flexible and dynamic data manipulation within the Haystack pipeline, providing the capability
+to adjust data formats and values between different components seamlessly.
+
+### Note:
+
+When use above below listed custom filters or your own developed filters, ensure that the data passed through the
+pipelines, OutputAdapters, and filters is properly sanitized and validated, especially if it originates from untrusted
+sources, to mitigate security risks.
+"""
+
+
+def change_role(messages: List[ChatMessage], role: ChatRole) -> List[ChatMessage]:
+    """
+    Change the role of the last message in a list of chat messages.
+
+    :param messages: List of chat messages
+    :param role: New role for the last message
+    :return: List of chat messages with the last message's role changed to the specified role
+    """
+    messages[-1].role = role
+    return messages
+
+
+def prepare_fc_params(openai_functions_schema: Dict[str, Any]) -> Dict[str, Any]:
+    """
+    Prepare function call parameters for OpenAI functions given the OpenAI functions schema.
+
+    :param openai_functions_schema: OpenAI functions schema
+    :return: Prepared function call parameters
+    """
+    if openai_functions_schema:
+        return {
+            "tools": [{"type": "function", "function": openai_functions_schema}],
+            "tool_choice": {"type": "function", "function": {"name": openai_functions_schema["name"]}},
+        }
+    else:
+        return {}
+
+
+def tojson(string_like_json: Any) -> Dict[str, Any]:
+    """
+    Convert a string in JSON format to a Python dictionary.
+
+    :param string_like_json: String in JSON format
+    :return: Python dictionary representation of the JSON string
+    """
+    return json.loads(str(string_like_json))
+
+
+ALL_FILTERS = {"change_role": change_role, "prepare_fc_params": prepare_fc_params, "tojson": tojson}

releasenotes/notes/add-output-adapter-predefined-custom-filters-b589b37c560641d5.yaml

@@ -0,0 +1,4 @@
+---
+features:
+ - |
+   Introduce predefined custom filters for OutputAdapter to easily enhance pipeline data manipulation capabilities.

test/components/converters/test_output_adapter.py

@@ -5,6 +5,7 @@
 from haystack import Pipeline, component
 from haystack.components.converters import OutputAdapter
 from haystack.components.converters.output_adapter import OutputAdaptationException
+from haystack.components.converters.output_adapter_filters import ALL_FILTERS
 
 
 def custom_filter_to_sede(value):
@@ -147,3 +148,18 @@ def run(self):
         result = pipe.run(data={})
         assert result
         assert result["output_adapter"]["output"] == {"framework": "Haystack"}
+
+    def test_sede_pipeline_with_custom_filters(self):
+        pipe = Pipeline()
+        pipe.add_component(
+            name="output_adapter",
+            instance=OutputAdapter(template="{{ irrelevant}}", output_type=str, custom_filters=ALL_FILTERS),
+        )
+        pipeline_serialized = pipe.dumps()
+        assert "custom_filters" in pipeline_serialized
+        assert "prepare_fc_params" in pipeline_serialized
+        assert "tojson" in pipeline_serialized
+        assert "change_role" in pipeline_serialized
+
+        deserialized_pipe = Pipeline.loads(pipeline_serialized)
+        assert deserialized_pipe.get_component("output_adapter").custom_filters == ALL_FILTERS