Feature/concept refines text (Pipelex#120)

Author: lchoquel

.cursor/rules/standards.mdc

@@ -7,6 +7,10 @@ alwaysApply: true
 
 This document outlines the coding standards and quality control procedures that must be followed when contributing to this project.
 
+## Style
+
+Always use type hints. Use the types with Uppercase first letter for types like Dict[], List[] etc.
+
 ## Code Quality Checks
 
 ### Linting and Type Checking

mkdocs.yml

@@ -39,8 +39,8 @@ extra:
       link: https://github.com/Pipelex/pipelex
       name: Pipelex on GitHub
     - icon: fontawesome/brands/python
-      link: https://pypi.org/project/kajson/
-      name: kajson on PyPI
+      link: https://pypi.org/project/pipelex/
+      name: pipelex on PyPI
     - icon: fontawesome/brands/twitter
       link: https://x.com/PipelexAI
       name: Pipelex on X

pipelex/core/concept_factory.py

@@ -72,7 +72,11 @@ def make_from_details_dict(
             raise ConceptFactoryError(f"Concept '{code}' in domain '{domain_code}' has no definition")
         details_dict["definition"] = concept_definition
         details_dict["domain"] = domain_code
-        details_dict["refines"] = ConceptFactory.make_refines(domain=domain_code, refines=details_dict.pop("refines", []))
+        refines = ConceptFactory.make_refines(domain=domain_code, refines=details_dict.pop("refines", []))
+        if not refines and not details_dict.get("structure"):
+            # No structure? this refines Text
+            refines = [NativeConcept.TEXT.code]
+        details_dict["refines"] = refines
         concept_blueprint = ConceptBlueprint.model_validate(details_dict)
         the_concept = ConceptFactory.make_concept_from_blueprint(domain=domain_code, code=code, concept_blueprint=concept_blueprint)
         return the_concept
@@ -90,19 +94,18 @@ def make_concept_from_definition(
             structure_class_name = code
         else:
             structure_class_name = TextContent.__name__
-        # TODO: why use a dict to create a concept? it makes no sense
-        # TODO: don't wall make_concept_code from the factory, the code received here must already be a valid concept code
-        concept_dict = {
-            "domain": domain_code,
-            "code": ConceptCodeFactory.make_concept_code(domain_code, code),
-            "definition": definition,
-            "structure_class_name": structure_class_name,
-        }
+
         try:
-            the_concept = Concept.model_validate(concept_dict)
+            the_concept = Concept(
+                code=ConceptCodeFactory.make_concept_code(domain_code, code),
+                domain=domain_code,
+                definition=definition,
+                structure_class_name=structure_class_name,
+                refines=[NativeConcept.TEXT.code],
+            )
+            return Concept.model_validate(the_concept)
         except ValidationError as e:
             raise ConceptFactoryError(f"Error validating concept: {e}") from e
-        return the_concept
 
     @classmethod
     def make_concept_from_blueprint(

pipelex/core/stuff.py

@@ -140,6 +140,11 @@ def as_text(self) -> TextContent:
         """Get content as TextContent if applicable."""
         return self.content_as(TextContent)
 
+    @property
+    def as_str(self) -> str:
+        """Get content as string if applicable."""
+        return self.as_text.text
+
     @property
     def as_image(self) -> ImageContent:
         """Get content as ImageContent if applicable."""

pipelex/libraries/llm_deck/base_llm_deck.toml

@@ -21,7 +21,7 @@ best-grok = "grok-3"
 ####################################################################################################
 # LLM Presets — General purpose
 
-cheap_llm_for_text = { llm_handle = "gpt-4o-mini", temperature = 0.5, max_tokens = 50 }
+cheap_llm_for_text = { llm_handle = "gpt-4o-mini", temperature = 0.5 }
 cheap_llm_for_short_text = { llm_handle = "gpt-4o-mini", temperature = 0.5, max_tokens = 50 }
 cheap_llm_for_object = { llm_handle = "gpt-4o-mini", temperature = 0.5 }
 cheap_llm_to_structure = { llm_handle = "gpt-4o-mini", temperature = 0.1 }

pyproject.toml

@@ -24,7 +24,7 @@ dependencies = [
     "instructor>=1.8.3",
     "jinja2>=3.1.4",
     "json2html>=1.3.0",
-    "kajson==0.2.0",
+    "kajson==0.2.2",
     "markdown>=3.6",
     "networkx>=3.4.2",
     "openai>=1.60.1",