Now allowing extra fields in prompts / tools (instead of ignoring the…

…m altogether), but raising warnings when we encounter them. (#21)

Co-authored-by: Thejas N U <77475353+ThejasNU@users.noreply.github.com>

libs/agentc_cli/tests/test_click.py

@@ -13,8 +13,12 @@
 from agentc_core.defaults import DEFAULT_TOOL_CATALOG_NAME
 from agentc_testing.repo import ExampleRepoKind
 from agentc_testing.repo import initialize_repo
+from agentc_testing.server import isolated_server_factory
 from unittest.mock import patch
 
+# This is to keep ruff from falsely flagging this as unused.
+_ = isolated_server_factory
+
 
 @pytest.mark.smoke
 def test_index(tmp_path):

libs/agentc_core/agentc_core/prompt/models.py

@@ -17,6 +17,7 @@
 
 
 class ToolSearchMetadata(pydantic.BaseModel):
+    model_config = pydantic.ConfigDict(extra="allow")
     name: typing.Optional[str] = None
     query: typing.Optional[str] = None
     annotations: typing.Optional[str] = None
@@ -42,6 +43,7 @@ def name_or_query_must_be_specified(self):
 
 class _BaseFactory(abc.ABC):
     class PromptMetadata(pydantic.BaseModel):
+        model_config = pydantic.ConfigDict(extra="allow")
         name: str
         description: str
         record_kind: typing.Literal[RecordKind.RawPrompt, RecordKind.JinjaPrompt]
@@ -80,6 +82,11 @@ class Factory(_BaseFactory):
         def __iter__(self) -> typing.Iterable["RawPromptDescriptor"]:
             front_matter, prompt_text = self._get_prompt_metadata()
             metadata = RawPromptDescriptor.Factory.PromptMetadata.model_validate(front_matter)
+            if metadata.__pydantic_extra__:
+                logger.warning(
+                    f"Extra fields found in {self.filename.name}: {metadata.__pydantic_extra__}. "
+                    f"We will ignore these."
+                )
             descriptor_args = {
                 "name": metadata.name,
                 "description": metadata.description,
@@ -111,6 +118,11 @@ class Factory(_BaseFactory):
         def __iter__(self) -> typing.Iterable["JinjaPromptDescriptor"]:
             front_matter, prompt_text = self._get_prompt_metadata()
             metadata = JinjaPromptDescriptor.Factory.PromptMetadata.model_validate(front_matter)
+            if metadata.__pydantic_extra__:
+                logger.warning(
+                    f"Extra fields found in {self.filename.name}: {metadata.__pydantic_extra__}. "
+                    f"We will ignore these."
+                )
             descriptor_args = {
                 "name": metadata.name,
                 "description": metadata.description,

libs/agentc_core/agentc_core/record/descriptor.py

@@ -45,7 +45,7 @@ def is_tool(self) -> bool:
 class RecordDescriptor(pydantic.BaseModel):
     """This model represents a tool's persistable description or metadata."""
 
-    model_config = pydantic.ConfigDict(validate_assignment=True, use_enum_values=True)
+    model_config = pydantic.ConfigDict(validate_assignment=True, use_enum_values=True, extra="allow")
 
     record_kind: typing.Literal[
         RecordKind.PythonFunction,

libs/agentc_core/agentc_core/tool/descriptor/models.py

@@ -56,7 +56,7 @@ def __iter__(self) -> typing.Iterable["PythonToolDescriptor"]:
             for _, tool in inspect.getmembers(imported_module):
                 if not is_tool(tool):
                     continue
-                yield PythonToolDescriptor(
+                record_descriptor = PythonToolDescriptor(
                     record_kind=RecordKind.PythonFunction,
                     name=get_name(tool),
                     description=get_description(tool),
@@ -65,6 +65,12 @@ def __iter__(self) -> typing.Iterable["PythonToolDescriptor"]:
                     version=self.version,
                     annotations=get_annotations(tool),
                 )
+                if record_descriptor.__pydantic_extra__:
+                    logger.warning(
+                        f"Extra fields found in {self.filename.name} for tool {get_name(tool)}: "
+                        f"{record_descriptor.__pydantic_extra__.keys()}. We will ignore these."
+                    )
+                yield record_descriptor
 
 
 class SQLPPQueryToolDescriptor(RecordDescriptor):
@@ -76,7 +82,7 @@ class SQLPPQueryToolDescriptor(RecordDescriptor):
 
     class Factory(_BaseFactory):
         class Metadata(pydantic.BaseModel, JSONSchemaValidatingMixin):
-            model_config = pydantic.ConfigDict(frozen=True, use_enum_values=True)
+            model_config = pydantic.ConfigDict(frozen=True, use_enum_values=True, extra="allow")
 
             # Below, we enumerate all fields that appear in a .sqlpp file.
             name: str
@@ -110,6 +116,11 @@ def __iter__(self) -> typing.Iterable["SQLPPQueryToolDescriptor"]:
                 elif len(matches) != 1:
                     logger.warning("More than one multi-line comment found. Using first comment.")
                 metadata = SQLPPQueryToolDescriptor.Factory.Metadata.model_validate(yaml.safe_load(matches[0]))
+                if metadata.__pydantic_extra__:
+                    logger.warning(
+                        f"Extra fields found in {self.filename.name}: {metadata.__pydantic_extra__}. "
+                        f"We will ignore these."
+                    )
 
             # Now, generate a single SQL++ tool descriptor.
             yield SQLPPQueryToolDescriptor(
@@ -145,7 +156,7 @@ class VectorSearchMetadata(pydantic.BaseModel):
 
     class Factory(_BaseFactory):
         class Metadata(pydantic.BaseModel, JSONSchemaValidatingMixin):
-            model_config = pydantic.ConfigDict(frozen=True, use_enum_values=True)
+            model_config = pydantic.ConfigDict(frozen=True, use_enum_values=True, extra="allow")
 
             # Below, we enumerate all fields that appear in a .yaml file for semantic search.
             record_kind: typing.Literal[RecordKind.SemanticSearch]
@@ -181,6 +192,11 @@ def name_should_be_valid_identifier(cls, v: str):
         def __iter__(self) -> typing.Iterable["SemanticSearchToolDescriptor"]:
             with self.filename.open("r") as fp:
                 metadata = SemanticSearchToolDescriptor.Factory.Metadata.model_validate(yaml.safe_load(fp))
+                if metadata.__pydantic_extra__:
+                    logger.warning(
+                        f"Extra fields found in {self.filename.name}: {metadata.__pydantic_extra__}. "
+                        f"We will ignore these."
+                    )
                 yield SemanticSearchToolDescriptor(
                     record_kind=RecordKind.SemanticSearch,
                     name=metadata.name,
@@ -340,10 +356,11 @@ def default(self, obj):
 
     class Factory(_BaseFactory):
         class Metadata(pydantic.BaseModel):
-            model_config = pydantic.ConfigDict(frozen=True, use_enum_values=True)
+            model_config = pydantic.ConfigDict(frozen=True, use_enum_values=True, extra="allow")
 
             # Note: we cannot validate this model in isolation (we need the referencing descriptor as well).
             class OpenAPIMetadata(pydantic.BaseModel):
+                model_config = pydantic.ConfigDict(extra="allow")
                 filename: typing.Optional[str | None] = None
                 url: typing.Optional[str | None] = None
                 operations: list["HTTPRequestToolDescriptor.OperationMetadata"]
@@ -356,6 +373,11 @@ class OpenAPIMetadata(pydantic.BaseModel):
         def __iter__(self) -> typing.Iterable["HTTPRequestToolDescriptor"]:
             with self.filename.open("r") as fp:
                 metadata = HTTPRequestToolDescriptor.Factory.Metadata.model_validate(yaml.safe_load(fp))
+                if metadata.__pydantic_extra__:
+                    logger.warning(
+                        f"Extra fields found in {self.filename.name}: {metadata.__pydantic_extra__}. "
+                        f"We will ignore these."
+                    )
                 for operation in metadata.open_api.operations:
                     operation_handle = HTTPRequestToolDescriptor.validate_operation(
                         source_filename=self.filename,