24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208 | def get_relationships(state: State, config: Dict[str, Any]) -> Dict[str, Any]:
"""
Determine relationships between building components using a language model.
Args:
state (State): The current state containing the user prompt and element hierarchy.
config (dict): Configuration dictionary containing the language model.
Returns:
dict: A dictionary containing the grouped sensor paths.
"""
custom_logger.eurac("🔗 Identifying related entities from the user prompt")
user_prompt = state["user_prompt"]
hierarchical_structure = state["elem_hierarchy"]
# Convert building structure to a JSON string for better readability
hierarchical_structure_json = json.dumps(hierarchical_structure, indent=2)
# Get the model name from the config
llm = config.get("configurable", {}).get("llm_model")
# Enforce structured output
structured_llm = llm.with_structured_output(RelationshipsSchema)
# System message
system_message = get_relationships_instructions.format(
prompt=user_prompt
)
answer = structured_llm.invoke(
[SystemMessage(content=system_message)]
)
# Check the format of the entities in order to not break the graph.
relationships = []
for relationship in answer.relationships:
relationships.append((relationship[0].replace(" ", "_"), (relationship[1].replace(" ", "_"))))
# Build the skeleton of the entities as a dict
entities_dict = {}
for relationship in relationships:
if relationship[0] not in entities_dict and relationship[0] != "":
entities_dict[relationship[0]] = {
"name": relationship[0],
"type": None
}
if relationship[1] not in entities_dict and relationship[1] != "":
entities_dict[relationship[1]] = {
"name": relationship[1],
"type": None
}
# Obtain all Brick entities to check if the entity type that will be identified is true
def extract_keys(data, keys=None):
if keys is None:
keys = []
for key, value in data.items():
keys.append(key)
if isinstance(value, dict): # If the value is a nested dictionary, recurse
extract_keys(value, keys)
return keys
all_brick_entities = extract_keys(brick_hierarchy)
custom_logger.eurac("🔗 Identifying entity types")
for entity in entities_dict:
system_message = find_entities_type.format(
user_prompt=user_prompt,
hierarchical_structure=hierarchical_structure_json,
entity=entity,
relationships=relationships
)
structured_llm = llm.with_structured_output(EntityType)
for attempt in range(MAX_RETRIES):
answer = structured_llm.invoke([SystemMessage(content=system_message)])
concept = answer.ontology_concept
if concept == entity:
entity_type = concept.split(".")[0]
elif "." in concept:
entity_type = concept.split(".")[-1]
elif concept not in all_brick_entities:
custom_logger.warning(f"Entity not recognized (attempt {attempt + 1}). Retrying...")
continue
else:
entity_type = concept
break
else:
# Fallback if all retries failed
custom_logger.warning(f"Max retries reached for entity: {entity}. Using fallback type.")
entity_type = entity.split(".")[0]
entities_dict[entity]["type"] = entity_type
# Adding the possible relationships for each entity extracted by the Brick ontology
for entity in entities_dict:
try:
entities_dict[entity]["relationships"] = find_rel_constraint(entities_dict[entity]["type"])
except Exception:
# Drop the entity if it does not have a type
print("Type not recognized for entity: ", entity)
# Extract the super classes for each entity from the Brick ontology
for entity in entities_dict:
super_classes = find_parents(brick_hierarchy, entities_dict[entity]["type"])[1]
entities_dict[entity]["super_classes"] = super_classes
custom_logger.eurac("🔗 Getting relationships between building components")
triples = []
for relationship in relationships:
subject = relationship[0]
object = relationship[1]
all_relationships = entities_dict[subject]["relationships"]
object_type = entities_dict[object]["type"]
object_superclasses = entities_dict[object]["super_classes"]
object_constraint = [object_type] + object_superclasses
# Extract the relationships that are valid for the entity type and its super classes
valid_relationships = {}
for rel_type in all_relationships:
if any([constraint in all_relationships[rel_type]["constraints"] for constraint in object_constraint]):
valid_relationships[rel_type] = all_relationships[rel_type]["definition"]
if len(valid_relationships.keys()) == 0:
valid_relationships = get_rel_definition()
system_message = find_relationship.format(
user_prompt=user_prompt,
tuple=(subject, object),
relationships=valid_relationships
)
structured_llm = llm.with_structured_output(Relationship)
answer = structured_llm.invoke(
[SystemMessage(content=system_message)]
+ [HumanMessage(content="Find the most appropriate relationships based on the user prompt provided.")]
)
valid = False
for attempt in range(MAX_RETRIES):
if answer.relationship in valid_relationships:
valid = True
break
else:
custom_logger.warning(f"The relationship is not valid. Trying again (attempt {attempt + 1}).")
answer = structured_llm.invoke([SystemMessage(content=system_message)])
if not valid:
custom_logger.error("Failed to get a valid relationship after maximum retries.")
if answer.relationship not in list(valid_relationships.keys()):
answer.relationship = None
triples.append((f"{subject}", f"{answer.relationship}", f"{object}"))
# Append the triples identified to the graph
graph = Graph()
BLDG = Namespace("urn:Building#")
BRICK = Namespace("https://brickschema.org/schema/Brick#")
RDF = Namespace("http://www.w3.org/1999/02/22-rdf-syntax-ns#")
graph.bind("bldg", BLDG)
graph.bind("brick", BRICK)
graph.bind("rdf", RDF)
for triple in triples:
if triple[1] is not None:
graph.add((BLDG[triple[0]], BRICK[triple[1]], BLDG[triple[2]]))
for entity in entities_dict:
if not entities_dict[entity]["type"] is None:
graph.add((BLDG[entity], RDF.type, BRICK[entities_dict[entity]["type"]]))
return {"relationships": triples, "graph": graph}
|