amitupadhyay-ai commited on
Commit
f2f4514
·
verified ·
1 Parent(s): 7ea840e

User → Retrieval → Context → LLM → Final Answer

Browse files
Files changed (1) hide show
  1. app.py +32 -18
app.py CHANGED
@@ -1,30 +1,34 @@
1
  import gradio as gr
2
  from sentence_transformers import SentenceTransformer
 
3
  import numpy as np
4
 
5
- # Load embedding model (lightweight + free)
6
- model = SentenceTransformer("all-MiniLM-L6-v2")
7
 
 
 
8
 
9
- # STEP 1: Load data
 
10
  def load_data():
11
  with open("data.txt", "r") as f:
12
  return f.read()
13
 
14
 
15
- # STEP 2: Chunking
16
  def chunk_text(text, chunk_size=200):
17
  return [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
18
 
19
 
20
- # STEP 3: Create embeddings
21
  def embed_chunks(chunks):
22
- return model.encode(chunks)
23
 
24
 
25
- # STEP 4: Similarity search
26
  def retrieve(query, chunks, embeddings):
27
- query_embedding = model.encode([query])[0]
28
 
29
  scores = []
30
  for i, emb in enumerate(embeddings):
@@ -37,7 +41,7 @@ def retrieve(query, chunks, embeddings):
37
  return [chunk for _, chunk in scores[:2]]
38
 
39
 
40
- # STEP 5: Full pipeline
41
  def ask_ai(query):
42
  text = load_data()
43
  chunks = chunk_text(text)
@@ -45,23 +49,33 @@ def ask_ai(query):
45
 
46
  relevant_chunks = retrieve(query, chunks, embeddings)
47
 
48
- response = f"Question: {query}\n\nRelevant Context:\n"
49
- for i, chunk in enumerate(relevant_chunks, 1):
50
- response += f"{i}. {chunk}\n\n"
 
 
 
 
 
 
 
 
 
 
 
51
 
52
- response += "Final Answer:\n"
53
- response += " ".join(relevant_chunks)
54
 
55
- return response
56
 
57
 
58
  # UI
59
  demo = gr.Interface(
60
  fn=ask_ai,
61
- inputs=gr.Textbox(lines=3, placeholder="Ask anything..."),
62
  outputs=gr.Textbox(lines=15),
63
- title="AIWisdom — Semantic RAG",
64
- description="Step 3: Embeddings-based retrieval (real AI)",
65
  )
66
 
67
  demo.launch()
 
1
  import gradio as gr
2
  from sentence_transformers import SentenceTransformer
3
+ from transformers import pipeline
4
  import numpy as np
5
 
6
+ # Embedding model
7
+ embed_model = SentenceTransformer("all-MiniLM-L6-v2")
8
 
9
+ # LLM (free, small but works)
10
+ llm = pipeline("text2text-generation", model="google/flan-t5-base")
11
 
12
+
13
+ # Load data
14
  def load_data():
15
  with open("data.txt", "r") as f:
16
  return f.read()
17
 
18
 
19
+ # Chunking
20
  def chunk_text(text, chunk_size=200):
21
  return [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
22
 
23
 
24
+ # Embed chunks
25
  def embed_chunks(chunks):
26
+ return embed_model.encode(chunks)
27
 
28
 
29
+ # Retrieve relevant chunks
30
  def retrieve(query, chunks, embeddings):
31
+ query_embedding = embed_model.encode([query])[0]
32
 
33
  scores = []
34
  for i, emb in enumerate(embeddings):
 
41
  return [chunk for _, chunk in scores[:2]]
42
 
43
 
44
+ # Final RAG pipeline
45
  def ask_ai(query):
46
  text = load_data()
47
  chunks = chunk_text(text)
 
49
 
50
  relevant_chunks = retrieve(query, chunks, embeddings)
51
 
52
+ context = "\n".join(relevant_chunks)
53
+
54
+ prompt = f"""
55
+ You are an AI systems expert.
56
+
57
+ Use the context below to answer the question.
58
+
59
+ Context:
60
+ {context}
61
+
62
+ Question: {query}
63
+
64
+ Answer clearly:
65
+ """
66
 
67
+ response = llm(prompt, max_new_tokens=200)
 
68
 
69
+ return response[0]["generated_text"]
70
 
71
 
72
  # UI
73
  demo = gr.Interface(
74
  fn=ask_ai,
75
+ inputs=gr.Textbox(lines=3, placeholder="Ask anything about AI systems..."),
76
  outputs=gr.Textbox(lines=15),
77
+ title="AIWisdom — Full RAG System",
78
+ description="Step 4: Retrieval + LLM reasoning (real AI system)",
79
  )
80
 
81
  demo.launch()