From 34735d3cfdf10411fc472288bbe582747921c342 Mon Sep 17 00:00:00 2001
From: Alkid <alkid1baci@gmail.com>
Date: Fri, 29 Nov 2024 16:36:10 +0100
Subject: [PATCH] added LLM generated response using the RAG assistance

---
 general_working_directory/llm_with_rag.py | 79 +++++++++++++++++++++++
 1 file changed, 79 insertions(+)
 create mode 100644 general_working_directory/llm_with_rag.py

diff --git a/general_working_directory/llm_with_rag.py b/general_working_directory/llm_with_rag.py
new file mode 100644
index 0000000..7fcf1dc
--- /dev/null
+++ b/general_working_directory/llm_with_rag.py
@@ -0,0 +1,79 @@
+from openai import OpenAI
+import pandas as pd
+import numpy as np
+from numpy.linalg import norm
+from owlapy.iri import IRI
+from owlapy.owl_individual import OWLNamedIndividual
+from owlapy.owl_ontology_manager import OntologyManager
+from owlapy.owl_property import OWLDataProperty
+from owlapy.owl_reasoner import StructuralReasoner
+
+manager = OntologyManager()
+ontology = manager.load_ontology(IRI.create("file://fashionpedia-third-generation.owl"))
+reasoner = StructuralReasoner(ontology=ontology)
+dprop1 = OWLDataProperty(IRI.create("http://example.org/hasDescription"))
+dprop2 = OWLDataProperty(IRI.create("http://example.org/hasLLMDescription"))
+
+llm_client = OpenAI(base_url="http://tentris-ml.cs.upb.de:8501/v1", api_key="token-tentris-upb")
+
+
+def get_result(query, docs):
+    return llm_client.chat.completions.create(
+        model="tentris",
+        messages=[
+            {
+                "role": "user",
+                "content":
+                    [
+                        {
+                            "type": "text",
+                            "text": "You are a apparel-loving AI and your focus is to give information about apparels. You should find similar points on the information provided to you and present them in a short paragraph tailored to the following query: "
+                                    f"'{query}'"
+                                    f"The information is as follows: {docs}"
+                        }
+                    ]
+            }
+        ],
+        temperature=0.1,
+        seed=1
+    ).choices[0].message.content
+
+
+query = input("What would you like to wear?\n")
+
+df = pd.read_csv("embeddings_third_kg.csv", index_col=0, nrows=None)
+iris = df.index.values.tolist()
+
+embbeding_client = OpenAI(base_url="http://tentris-ml.cs.upb.de:8502/v1", api_key="token-tentris-upb")
+
+docs = np.array(df.values)
+qr = np.array(embbeding_client.embeddings.create(input=[query], model="tentris").data[0].embedding)
+
+docs_norms = docs / norm(docs, axis=1, keepdims=True)
+qr_norms = qr / norm(qr)
+
+cosine_similarities = (docs_norms @ qr_norms).flatten()
+
+best_match_index = np.argmax(cosine_similarities)
+best_similarity = cosine_similarities[best_match_index]
+
+indexes = np.argpartition(cosine_similarities, -10)[-10:]
+merged_documents = ""
+for index in indexes:
+
+    iri = iris[index]
+    image_ind = OWLNamedIndividual(iri)
+
+    llm_description = str(list(reasoner.data_property_values(image_ind, dprop2))[0].get_literal())
+    if len(llm_description) > 800:
+        llm_description = llm_description[:800]
+
+    all_descriptions = ""
+    for d in list(reasoner.data_property_values(image_ind, dprop1)):
+        all_descriptions = all_descriptions + d.get_literal() + "\n"
+
+    merged_documents += llm_description + " \n" + all_descriptions
+
+result = get_result(query, merged_documents)
+
+print(result)
\ No newline at end of file