Merge branch 'main' into feature/chinese-document-splitter

haystack/tools/parameters_schema_utils.py

@@ -7,15 +7,11 @@ from dataclasses import MISSING, fields, is_dataclass
 from inspect import getdoc
 from typing import Any, Callable, Dict, List, Sequence, Tuple, Union, get_args, get_origin
 
+from docstring_parser import parse
 from pydantic import BaseModel, Field, create_model
 
 from haystack import logging
 from haystack.dataclasses import ChatMessage
-from haystack.lazy_imports import LazyImport
-
-with LazyImport(message="Run 'pip install docstring-parser'") as docstring_parser_import:
-    from docstring_parser import parse
-
 
 logger = logging.getLogger(__name__)
 
@@ -33,7 +29,6 @@ def _get_param_descriptions(method: Callable) -> Tuple[str, Dict[str, str]]:
     if not docstring:
         return "", {}
 
-    docstring_parser_import.check()
     parsed_doc = parse(docstring)
     param_descriptions = {}
     for param in parsed_doc.params:

pyproject.toml

@@ -57,6 +57,7 @@ dependencies = [
   "numpy",
   "python-dateutil",
   "jsonschema",             # JsonSchemaValidator, Tool
+  "docstring-parser",       # ComponentTool
   "haystack-experimental",
 ]
 
@@ -129,9 +130,6 @@ dependencies = [
   # Structured logging
   "structlog",
 
-  # ComponentTool
-  "docstring-parser",
-
   # needed in link content fetcher tests
   "httpx[http2]",
 

releasenotes/notes/doc-parser-core-dependency-a1b2c3d4e5f6g7h8.yaml

@@ -0,0 +1,4 @@
+---
+enhancements:
+  - |
+    Made doc-parser a core dependency since ComponentTool that uses it is one of the core Tool components.