Update app.py
Browse files
app.py
CHANGED
@@ -10,30 +10,70 @@ import os
|
|
10 |
import pytesseract
|
11 |
from PIL import Image
|
12 |
pytesseract.pytesseract.tesseract_cmd = r"tesseract.exe"
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
|
|
|
|
|
|
|
|
29 |
|
30 |
ipc_tool = Tool(
|
31 |
name="IPC Information Retrieval",
|
32 |
func=retrieve_info,
|
33 |
-
description="
|
34 |
)
|
35 |
|
36 |
-
|
|
|
|
|
|
|
|
|
37 |
llm = ChatGoogleGenerativeAI(
|
38 |
model="gemini-1.5-pro",
|
39 |
temperature=0.25,
|
@@ -56,7 +96,7 @@ llm = ChatGoogleGenerativeAI(
|
|
56 |
""",
|
57 |
)
|
58 |
|
59 |
-
agent_tools = [ipc_tool]
|
60 |
|
61 |
agent = initialize_agent(
|
62 |
tools=agent_tools,
|
@@ -145,6 +185,5 @@ with gr.Blocks(theme=gr.themes.Soft()) as iface:
|
|
145 |
|
146 |
|
147 |
iface.launch(
|
148 |
-
show_error=True
|
149 |
-
prevent_thread_lock=True
|
150 |
)
|
|
|
10 |
import pytesseract
|
11 |
from PIL import Image
|
12 |
pytesseract.pytesseract.tesseract_cmd = r"tesseract.exe"
|
13 |
+
model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
|
14 |
+
index = faiss.read_index('IPC_index.faiss')
|
15 |
+
index2 = faiss.read_index('CrpC_index.faiss')
|
16 |
+
|
17 |
+
|
18 |
+
# Step 3: Retrieval with Citations using PDF filename
|
19 |
+
def retrieve_info_with_citation(query, top_k=5):
|
20 |
+
query_embedding = model.encode([query])
|
21 |
+
D, I = index.search(query_embedding, k=top_k)
|
22 |
+
|
23 |
+
results = []
|
24 |
+
for i in range(min(top_k, len(I[0]))):
|
25 |
+
if D[0][i] < 1.0: # Relevance threshold
|
26 |
+
chunk_index = I[0][i]
|
27 |
+
citation = f"Source: IPC"
|
28 |
+
results.append((match, citation))
|
29 |
+
else:
|
30 |
+
break
|
31 |
+
|
32 |
+
if results:
|
33 |
+
return results
|
34 |
+
else:
|
35 |
+
return [("I'm sorry, I couldn't find relevant information.", "Source: N/A")]
|
36 |
+
|
37 |
+
|
38 |
+
def retrieve_info_with_citation2(query, top_k=5):
|
39 |
+
query_embedding = model.encode([query])
|
40 |
+
D, I = index.search(query_embedding, k=top_k)
|
41 |
+
|
42 |
+
results = []
|
43 |
+
for i in range(min(top_k, len(I[0]))):
|
44 |
+
if D[0][i] < 1.0: # Relevance threshold
|
45 |
+
chunk_index = I[0][i]
|
46 |
+
citation = f"Source: CrPC"
|
47 |
+
results.append((match, citation))
|
48 |
+
else:
|
49 |
+
break
|
50 |
|
51 |
+
if results:
|
52 |
+
return results
|
53 |
+
else:
|
54 |
+
return [("I'm sorry, I couldn't find relevant information.", "Source: N/A")]
|
55 |
+
|
56 |
+
def retrieve_info(query):
|
57 |
+
results = retrieve_info_with_citation(query)
|
58 |
+
formatted_results = "\n\n".join([f"{i+1}. {match}\n{citation}" for i, (match, citation) in enumerate(results)])
|
59 |
+
return formatted_results
|
60 |
|
61 |
+
def retrieve_info2(query):
|
62 |
+
results = retrieve_info_with_citation2(query)
|
63 |
+
formatted_results = "\n\n".join([f"{i+1}. {match}\n{citation}" for i, (match, citation) in enumerate(results)])
|
64 |
+
return formatted_results
|
65 |
|
66 |
ipc_tool = Tool(
|
67 |
name="IPC Information Retrieval",
|
68 |
func=retrieve_info,
|
69 |
+
description="Retrieve information from the Indian Penal Code Related to query keyword(s)."
|
70 |
)
|
71 |
|
72 |
+
crpc_tool=Tool(
|
73 |
+
name="CrPC Information Retrieval",
|
74 |
+
func=retrieve_info2,
|
75 |
+
description="Retrieve information from the Code of Criminal Procedure(CrPC) Related to query keyword(s)."
|
76 |
+
)
|
77 |
llm = ChatGoogleGenerativeAI(
|
78 |
model="gemini-1.5-pro",
|
79 |
temperature=0.25,
|
|
|
96 |
""",
|
97 |
)
|
98 |
|
99 |
+
agent_tools = [ipc_tool,crpc_tool]
|
100 |
|
101 |
agent = initialize_agent(
|
102 |
tools=agent_tools,
|
|
|
185 |
|
186 |
|
187 |
iface.launch(
|
188 |
+
show_error=True
|
|
|
189 |
)
|