feat: ttl_to_building_description function

README.md

@@ -6,7 +6,7 @@
 
 BrickLLM is a Python library for generating RDF files following the BrickSchema ontology using Large Language Models (LLMs).
 
-## 🚀 Features
+## 🧰 Features
 
 - Generate BrickSchema-compliant RDF files from natural language descriptions of buildings and facilities
 - Support for multiple LLM providers (OpenAI, Anthropic, Fireworks)

brickllm/__init__.py

@@ -1,11 +1,17 @@
 from .configs import GraphConfig
-from .schemas import ElemListSchema, RelationshipsSchema, TTLSchema
+from .schemas import (
+    ElemListSchema,
+    RelationshipsSchema,
+    TTLSchema,
+    TTLToBuildingPromptSchema,
+)
 from .states import State, StateLocal
 
 __all__ = [
     "ElemListSchema",
     "RelationshipsSchema",
     "TTLSchema",
+    "TTLToBuildingPromptSchema",
     "State",
     "StateLocal",
     "GraphConfig",

brickllm/graphs/abstract_graph.py

@@ -1,12 +1,13 @@
 import os
 from abc import ABC, abstractmethod
-from typing import Any, Dict, List, Union
+from typing import Any, Dict, List, Tuple, Union
 
 from langchain.chat_models.base import BaseChatModel
 from langgraph.graph import StateGraph
 from PIL import Image
 
 from ..helpers.llm_models import _get_model
+from ..utils import ttl_to_building_prompt
 
 
 class AbstractBrickSchemaGraph(ABC):
@@ -17,6 +18,8 @@ def __init__(self, model: Union[str, BaseChatModel] = "openai"):
         self.config = {"configurable": {"thread_id": "1", "llm_model": self.model}}
         self.result = None
         self.ttl_output = None
+        self.generated_building_description = None
+        self.generated_key_elements = None
 
     @abstractmethod
     def build_graph(self):
@@ -81,3 +84,11 @@ def save_ttl_output(self, output_file: str = "brick_schema_output.ttl") -> None:
 
         with open(output_file, "w") as f:
             f.write(self.ttl_output)
+
+    def ttl_to_building_description(self) -> Tuple[str, List[str]]:
+        if self.ttl_output is None:
+            raise ValueError("No TTL output found. Please run the graph first.")
+        self.generated_building_description, self.generated_key_elements = (
+            ttl_to_building_prompt(self.ttl_output, self.model)
+        )
+        return self.generated_building_description, self.generated_key_elements

brickllm/helpers/__init__.py

@@ -6,6 +6,7 @@
     prompt_template_local,
     schema_to_ttl_instructions,
     ttl_example,
+    ttl_to_user_prompt,
 )
 
 __all__ = [
@@ -16,4 +17,5 @@
     "schema_to_ttl_instructions",
     "ttl_example",
     "prompt_template_local",
+    "ttl_to_user_prompt",
 ]

brickllm/helpers/prompts.py

@@ -9,7 +9,7 @@
     You are now asked to identify the elements presents in the user prompt, even if not explicitly mentioned.\n
     USER PROMPT: {prompt} \n
     ELEMENTS: {elements_dict} \n
-    """
+    """  # noqa
 
 get_elem_children_instructions: str = """
     You are a BrickSchema ontology expert and you are provided with a user prompt which describes a building or facility.\n
@@ -22,7 +22,7 @@
 
     USER PROMPT: {prompt} \n
     ELEMENTS HIERARCHY: {elements_list} \n
-    """
+    """  # noqa
 
 get_relationships_instructions: str = """
     You are a BrickSchema ontology expert and are provided with a detailed description of a building or facility.\n
@@ -35,7 +35,7 @@
     If an element has no relationships, add an empty string in place of the missing component ("Room.1","").\n
     Hierarchical structure: {building_structure}\n
     USER PROMPT: {prompt}
-"""
+"""  # noqa
 
 ttl_example: str = """
     @prefix bldg: <urn:Building#> .
@@ -84,7 +84,7 @@
 
     bldg:livingroom a brick:Room ;
         brick:isPartOf bldg:Milano_Residence_1 .
-"""
+"""  # noqa
 
 schema_to_ttl_instructions: str = """
     You are a BrickSchema ontology expert and you are provided with a user prompt which describes a building or facility.\n
@@ -102,7 +102,23 @@
     USER DESCRIPTION: {prompt}\n
 
     COMPONENTS DICT: {sensors_dict}\n
-"""
+"""  # noqa
+
+ttl_to_user_prompt: str = """
+    You are a BrickSchema ontology expert tasked with generating a clear and concise description of a building or facility from a TTL script.
+
+    Your output must follow these guidelines:
+    - Focus on the key building characteristics, components and relationships present in the TTL
+    - Maintain technical accuracy and use proper Brick terminology
+    - Keep descriptions clear and well-structured
+    - Only include information explicitly stated in the TTL script
+    - If no TTL content is provided, return an empty string
+
+    TTL script to analyze:
+    <ttl_script>
+    {ttl_script}
+    </ttl_script>
+"""  # noqa
 
 prompt_template_local: str = """
     Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
@@ -113,4 +129,4 @@
     {user_prompt}
 
     ### Response:
-"""
+"""  # noqa

brickllm/schemas.py

@@ -16,3 +16,12 @@ class TTLSchema(BaseModel):
     ttl_output: str = Field(
         ..., description="The generated BrickSchema turtle/rdf script."
     )
+
+
+class TTLToBuildingPromptSchema(BaseModel):
+    building_description: str = Field(
+        ..., description="The generated building description."
+    )
+    key_elements: List[str] = Field(
+        ..., description="The generated list of key elements."
+    )

brickllm/utils/__init__.py

@@ -22,6 +22,7 @@
     validate_ttl,
 )
 from .rdf_parser import extract_rdf_graph
+from .ttl_to_prompt import ttl_to_building_prompt
 
 __all__ = [
     "find_parents",
@@ -44,4 +45,5 @@
     "general_query",
     "validate_ttl",
     "extract_rdf_graph",
+    "ttl_to_building_prompt",
 ]

brickllm/utils/ttl_to_prompt.py

@@ -0,0 +1,27 @@
+from typing import List, Tuple, Union
+
+from langchain.chat_models.base import BaseChatModel
+from langchain_community.llms import Ollama
+from langchain_core.messages import HumanMessage, SystemMessage
+
+from ..helpers import ttl_to_user_prompt
+from ..schemas import TTLToBuildingPromptSchema
+
+
+def ttl_to_building_prompt(
+    ttl_file: str, llm: Union[Ollama, BaseChatModel]
+) -> Tuple[str, List[str]]:
+
+    # Enforce structured output
+    structured_llm = llm.with_structured_output(TTLToBuildingPromptSchema)
+
+    # System message
+    system_message = ttl_to_user_prompt.format(ttl_script=ttl_file)
+
+    # Generate question
+    answer = structured_llm.invoke(
+        [SystemMessage(content=system_message)]
+        + [HumanMessage(content="Generate the TTL.")]
+    )
+
+    return answer.building_description, answer.key_elements

examples/example_custom_llm.py

@@ -34,3 +34,12 @@
 
 # save the result to a file
 brick_graph.save_ttl_output("my_building_custom.ttl")
+
+# Generate the building description from the generated ttl file
+building_description, key_elements = brick_graph.ttl_to_building_description()
+
+print("Generated building description:")
+print(building_description)
+print("--------------------------------")
+print("Generated key elements:")
+print(key_elements)

examples/example_openai.py

@@ -32,3 +32,12 @@
 
 # save the result to a file
 brick_graph.save_ttl_output("my_building.ttl")
+
+# Generate the building description from the generated ttl file
+building_description, key_elements = brick_graph.ttl_to_building_description()
+
+print("Generated building description:")
+print(building_description)
+print("--------------------------------")
+print("Generated key elements:")
+print(key_elements)