Spaces:
Running
Running
vaishnav
commited on
Commit
·
e19d910
1
Parent(s):
3a0580c
update gradio sdk and add lfu caching
Browse files- README.md +1 -1
- app.py +11 -16
- caching/lfu.py +43 -0
- llm_setup/llm_setup.py +8 -7
README.md
CHANGED
@@ -4,7 +4,7 @@ emoji: 💬
|
|
4 |
colorFrom: yellow
|
5 |
colorTo: purple
|
6 |
sdk: gradio
|
7 |
-
sdk_version:
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
license: mit
|
|
|
4 |
colorFrom: yellow
|
5 |
colorTo: purple
|
6 |
sdk: gradio
|
7 |
+
sdk_version: 5.17.1
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
license: mit
|
app.py
CHANGED
@@ -13,24 +13,30 @@ config.set_envs() # Set environment variables using the config module
|
|
13 |
store = stores.chroma.ChromaDB(config.EMBEDDINGS)
|
14 |
service = services.scraper.Service(store)
|
15 |
|
16 |
-
|
17 |
# Scrape data and get the store vector retriever
|
18 |
service.scrape_and_get_store_vector_retriever(config.URLS)
|
19 |
|
20 |
# Initialize the LLMService with logger, prompt, and store vector retriever
|
21 |
llm_svc = LLMService(logger, config.SYSTEM_PROMPT, store.get_chroma_instance().as_retriever())
|
22 |
|
23 |
-
def respond(user_input,
|
24 |
if user_input == "clear_chat_history_aisdb_override":
|
25 |
llm_svc.store={}
|
26 |
return "Memory Cache cleared"
|
27 |
response = llm_svc.conversational_rag_chain().invoke(
|
28 |
{"input": user_input},
|
29 |
-
config={"configurable": {"session_id":
|
30 |
)["answer"]
|
31 |
|
32 |
return response
|
33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
def on_reset_button_click():
|
36 |
llm_svc.store={}
|
@@ -40,18 +46,7 @@ if __name__ == '__main__':
|
|
40 |
logging.info("Starting AIVIz Bot")
|
41 |
|
42 |
with gr.Blocks() as demo:
|
43 |
-
gr.
|
44 |
-
gr.Markdown("Welcome! Ask me anything about vessel tracking, AI models.")
|
45 |
-
|
46 |
-
with gr.Row():
|
47 |
-
chat_interface = gr.ChatInterface(fn=respond)
|
48 |
-
|
49 |
-
with gr.Row():
|
50 |
-
reset_button = gr.Button("🔄 Reset Chat Memory Cache")
|
51 |
-
reset_status = gr.Textbox(label="Status", interactive=False)
|
52 |
-
|
53 |
-
# Bind reset button to function
|
54 |
-
reset_button.click(fn=on_reset_button_click, outputs=reset_status)
|
55 |
|
56 |
# Launch the interface
|
57 |
-
demo.launch(
|
|
|
13 |
store = stores.chroma.ChromaDB(config.EMBEDDINGS)
|
14 |
service = services.scraper.Service(store)
|
15 |
|
|
|
16 |
# Scrape data and get the store vector retriever
|
17 |
service.scrape_and_get_store_vector_retriever(config.URLS)
|
18 |
|
19 |
# Initialize the LLMService with logger, prompt, and store vector retriever
|
20 |
llm_svc = LLMService(logger, config.SYSTEM_PROMPT, store.get_chroma_instance().as_retriever())
|
21 |
|
22 |
+
def respond(user_input,session_hash):
|
23 |
if user_input == "clear_chat_history_aisdb_override":
|
24 |
llm_svc.store={}
|
25 |
return "Memory Cache cleared"
|
26 |
response = llm_svc.conversational_rag_chain().invoke(
|
27 |
{"input": user_input},
|
28 |
+
config={"configurable": {"session_id": session_hash}},
|
29 |
)["answer"]
|
30 |
|
31 |
return response
|
32 |
|
33 |
+
def echo(text, chat_history, request: gr.Request):
|
34 |
+
if request:
|
35 |
+
session_hash = request.session_hash
|
36 |
+
return respond(text, session_hash)
|
37 |
+
else:
|
38 |
+
return "No request object received."
|
39 |
+
|
40 |
|
41 |
def on_reset_button_click():
|
42 |
llm_svc.store={}
|
|
|
46 |
logging.info("Starting AIVIz Bot")
|
47 |
|
48 |
with gr.Blocks() as demo:
|
49 |
+
gr.ChatInterface(fn=echo, type="messages")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
|
51 |
# Launch the interface
|
52 |
+
demo.launch()
|
caching/lfu.py
ADDED
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from collections import defaultdict, OrderedDict
|
2 |
+
|
3 |
+
class LFUCache:
|
4 |
+
def __init__(self, capacity: int):
|
5 |
+
self.capacity = capacity
|
6 |
+
self.data = {} # session_id -> (value, freq)
|
7 |
+
self.freq_map = defaultdict(OrderedDict) # freq -> {session_id: None}
|
8 |
+
self.min_freq = 0
|
9 |
+
|
10 |
+
def _update_freq(self, session_id):
|
11 |
+
value, freq = self.data[session_id]
|
12 |
+
del self.freq_map[freq][session_id]
|
13 |
+
if not self.freq_map[freq]:
|
14 |
+
del self.freq_map[freq]
|
15 |
+
if self.min_freq == freq:
|
16 |
+
self.min_freq += 1
|
17 |
+
|
18 |
+
new_freq = freq + 1
|
19 |
+
self.data[session_id] = (value, new_freq)
|
20 |
+
self.freq_map[new_freq][session_id] = None
|
21 |
+
|
22 |
+
def get(self, session_id):
|
23 |
+
if session_id not in self.data:
|
24 |
+
return None
|
25 |
+
self._update_freq(session_id)
|
26 |
+
return self.data[session_id][0]
|
27 |
+
|
28 |
+
def put(self, session_id, value):
|
29 |
+
if self.capacity == 0:
|
30 |
+
return
|
31 |
+
|
32 |
+
if session_id in self.data:
|
33 |
+
self.data[session_id] = (value, self.data[session_id][1])
|
34 |
+
self._update_freq(session_id)
|
35 |
+
else:
|
36 |
+
if len(self.data) >= self.capacity:
|
37 |
+
# Evict the least frequently used item
|
38 |
+
lfu_session_id, _ = self.freq_map[self.min_freq].popitem(last=False)
|
39 |
+
del self.data[lfu_session_id]
|
40 |
+
|
41 |
+
self.data[session_id] = (value, 1)
|
42 |
+
self.freq_map[1][session_id] = None
|
43 |
+
self.min_freq = 1
|
llm_setup/llm_setup.py
CHANGED
@@ -12,7 +12,7 @@ from langchain_core.chat_history import BaseChatMessageHistory
|
|
12 |
from langchain_community.chat_message_histories import ChatMessageHistory
|
13 |
from langchain_core.runnables.history import RunnableWithMessageHistory
|
14 |
from processing.documents import format_documents
|
15 |
-
|
16 |
|
17 |
def _initialize_llm() -> ChatGoogleGenerativeAI:
|
18 |
"""
|
@@ -23,7 +23,7 @@ def _initialize_llm() -> ChatGoogleGenerativeAI:
|
|
23 |
|
24 |
|
25 |
class LLMService:
|
26 |
-
def __init__(self, logger, system_prompt: str, web_retriever: VectorStoreRetriever):
|
27 |
self._conversational_rag_chain = None
|
28 |
self._logger = logger
|
29 |
self.system_prompt = system_prompt
|
@@ -34,7 +34,7 @@ class LLMService:
|
|
34 |
self._initialize_conversational_rag_chain()
|
35 |
|
36 |
### Statefully manage chat history ###
|
37 |
-
self.store =
|
38 |
|
39 |
def _initialize_conversational_rag_chain(self):
|
40 |
"""
|
@@ -55,7 +55,6 @@ class LLMService:
|
|
55 |
)
|
56 |
|
57 |
|
58 |
-
|
59 |
history_aware_retriever = create_history_aware_retriever(
|
60 |
self.llm, self._web_retriever, contextualize_q_prompt)
|
61 |
|
@@ -79,9 +78,11 @@ class LLMService:
|
|
79 |
)
|
80 |
|
81 |
def _get_session_history(self, session_id: str) -> BaseChatMessageHistory:
|
82 |
-
|
83 |
-
|
84 |
-
|
|
|
|
|
85 |
|
86 |
def conversational_rag_chain(self):
|
87 |
"""
|
|
|
12 |
from langchain_community.chat_message_histories import ChatMessageHistory
|
13 |
from langchain_core.runnables.history import RunnableWithMessageHistory
|
14 |
from processing.documents import format_documents
|
15 |
+
from caching.lfu import LFUCache
|
16 |
|
17 |
def _initialize_llm() -> ChatGoogleGenerativeAI:
|
18 |
"""
|
|
|
23 |
|
24 |
|
25 |
class LLMService:
|
26 |
+
def __init__(self, logger, system_prompt: str, web_retriever: VectorStoreRetriever,cache_capacity: int = 50):
|
27 |
self._conversational_rag_chain = None
|
28 |
self._logger = logger
|
29 |
self.system_prompt = system_prompt
|
|
|
34 |
self._initialize_conversational_rag_chain()
|
35 |
|
36 |
### Statefully manage chat history ###
|
37 |
+
self.store = LFUCache(capacity=cache_capacity)
|
38 |
|
39 |
def _initialize_conversational_rag_chain(self):
|
40 |
"""
|
|
|
55 |
)
|
56 |
|
57 |
|
|
|
58 |
history_aware_retriever = create_history_aware_retriever(
|
59 |
self.llm, self._web_retriever, contextualize_q_prompt)
|
60 |
|
|
|
78 |
)
|
79 |
|
80 |
def _get_session_history(self, session_id: str) -> BaseChatMessageHistory:
|
81 |
+
history = self.store.get(session_id)
|
82 |
+
if history is None:
|
83 |
+
history = ChatMessageHistory()
|
84 |
+
self.store.put(session_id, history)
|
85 |
+
return history
|
86 |
|
87 |
def conversational_rag_chain(self):
|
88 |
"""
|