File size: 8,880 Bytes
4bb745d
19f4fce
4bb745d
 
f5254ad
 
 
4bb745d
 
170741d
 
 
4bb745d
f5254ad
 
 
4bb745d
 
 
 
 
 
 
 
 
 
19f4fce
f5254ad
 
 
4bb745d
 
f5254ad
4bb745d
 
170741d
 
4bb745d
f5254ad
 
4bb745d
2abc521
4bb745d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f5254ad
 
 
 
 
 
 
 
 
2abc521
4bb745d
170741d
4bb745d
 
170741d
4bb745d
 
 
 
 
 
 
 
 
 
 
 
 
 
2abc521
 
 
4bb745d
 
 
f5254ad
4bb745d
 
 
170741d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f5254ad
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4bb745d
2abc521
4bb745d
19f4fce
 
 
12de659
4bb745d
 
 
 
f5254ad
 
 
 
 
4bb745d
f5254ad
4bb745d
f5254ad
 
 
 
170741d
 
 
 
 
 
 
f5254ad
 
 
4bb745d
f5254ad
 
 
 
 
4bb745d
 
 
170741d
 
4bb745d
 
 
170741d
4bb745d
 
 
f5254ad
 
 
 
19f4fce
4bb745d
 
 
2abc521
 
 
 
 
 
19f4fce
4bb745d
 
 
 
 
13c5bb4
4bb745d
 
 
19f4fce
4bb745d
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
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
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
import streamlit as st
import os
import pathlib
from typing import List

# local imports
from models.llms import load_llm, integrated_llms
from models.embeddings import hf_embed_model, openai_embed_model
from models.llamaCustom import LlamaCustom
from models.llamaCustomV2 import LlamaCustomV2

# from models.vector_database import pinecone_vector_store
from utils.chatbox import show_previous_messages, show_chat_input
from utils.util import validate_openai_api_key

# llama_index
from llama_index.core import (
    SimpleDirectoryReader,
    Document,
    VectorStoreIndex,
    StorageContext,
    Settings,
    load_index_from_storage,
)
from llama_index.core.memory import ChatMemoryBuffer
from llama_index.core.base.llms.types import ChatMessage

# huggingface
from huggingface_hub import HfApi

SAVE_DIR = "uploaded_files"
VECTOR_STORE_DIR = "vectorStores"
HF_REPO_ID = "zhtet/RegBotBeta"

# global
# Settings.embed_model = hf_embed_model
Settings.embed_model = openai_embed_model

# huggingface api
hf_api = HfApi()


def init_session_state():
    if "llama_messages" not in st.session_state:
        st.session_state.llama_messages = [
            {"role": "assistant", "content": "How can I help you today?"}
        ]

    # TODO: create a chat history for each different document
    if "llama_chat_history" not in st.session_state:
        st.session_state.llama_chat_history = [
            ChatMessage.from_str(role="assistant", content="How can I help you today?")
        ]

    if "llama_custom" not in st.session_state:
        st.session_state.llama_custom = None

    if "openai_api_key" not in st.session_state:
        st.session_state.openai_api_key = ""

    if "replicate_api_token" not in st.session_state:
        st.session_state.replicate_api_token = ""

    if "hf_token" not in st.session_state:
        st.session_state.hf_token = ""


# @st.cache_resource
def get_index(
    filename: str,
) -> VectorStoreIndex:
    """This function loads the index from storage if it exists, otherwise it creates a new index from the document."""
    try:
        index_path = pathlib.Path(f"{VECTOR_STORE_DIR}/{filename.replace('.', '_')}")
        if pathlib.Path.exists(index_path):
            print("Loading index from storage ...")
            storage_context = StorageContext.from_defaults(persist_dir=index_path)
            index = load_index_from_storage(storage_context=storage_context)

        else:
            reader = SimpleDirectoryReader(input_files=[f"{SAVE_DIR}/{filename}"])
            docs = reader.load_data(show_progress=True)
            index = VectorStoreIndex.from_documents(
                documents=docs,
                show_progress=True,
            )
            index.storage_context.persist(
                persist_dir=f"vectorStores/{filename.replace('.', '_')}"
            )

    except Exception as e:
        print(f"Error: {e}")
        raise e
    return index


# def get_pinecone_index(filename: str) -> VectorStoreIndex:
#     """Thie function loads the index from Pinecone if it exists, otherwise it creates a new index from the document."""
#     reader = SimpleDirectoryReader(input_files=[f"{SAVE_DIR}/{filename}"])
#     docs = reader.load_data(show_progress=True)
#     storage_context = StorageContext.from_defaults(vector_store=pinecone_vector_store)
#     index = VectorStoreIndex.from_documents(
#         documents=docs, show_progress=True, storage_context=storage_context
#     )

#     return index


def get_chroma_index(filename: str) -> VectorStoreIndex:
    """This function loads the index from Chroma if it exists, otherwise it creates a new index from the document."""
    pass


def check_api_key(model_name: str, source: str):
    if source.startswith("openai"):
        if not st.session_state.openai_api_key:
            with st.expander("OpenAI API Key", expanded=True):
                openai_api_key = st.text_input(
                    label="Enter your OpenAI API Key:",
                    type="password",
                    help="Get your key from https://platform.openai.com/account/api-keys",
                    value=st.session_state.openai_api_key,
                )

                if openai_api_key and st.spinner("Validating OpenAI API Key ..."):
                    result = validate_openai_api_key(openai_api_key)
                    if result["status"] == "success":
                        st.session_state.openai_api_key = openai_api_key
                        st.success(result["message"])
                    else:
                        st.error(result["message"])
                        st.info("You can still select a different model to proceed.")
                        st.stop()

    elif source.startswith("replicate"):
        if not st.session_state.replicate_api_token:
            with st.expander("Replicate API Token", expanded=True):
                replicate_api_token = st.text_input(
                    label="Enter your Replicate API Token:",
                    type="password",
                    help="Get your key from https://replicate.ai/account",
                    value=st.session_state.replicate_api_token,
                )

                # TODO: need to validate the token

                if replicate_api_token:
                    st.session_state.replicate_api_token = replicate_api_token
                    # set the environment variable
                    os.environ["REPLICATE_API_TOKEN"] = replicate_api_token

    elif source.startswith("huggingface"):
        if not st.session_state.hf_token:
            with st.expander("Hugging Face Token", expanded=True):
                hf_token = st.text_input(
                    label="Enter your Hugging Face Token:",
                    type="password",
                    help="Get your key from https://huggingface.co/settings/token",
                    value=st.session_state.hf_token,
                )

                if hf_token:
                    st.session_state.hf_token = hf_token
                    # set the environment variable
                    os.environ["HF_TOKEN"] = hf_token


init_session_state()

st.set_page_config(page_title="Llama", page_icon="🦙")

st.header("California Drinking Water Regulation Chatbot - RegBot with LlamaIndex Demo")

tab1, tab2 = st.tabs(["Config", "Chat"])

with tab1:
    selected_llm_name = st.selectbox(
        label="Select a model:",
        options=[f"{key} | {value}" for key, value in integrated_llms.items()],
    )
    model_name, source = selected_llm_name.split("|")

    check_api_key(model_name=model_name.strip(), source=source.strip())

    selected_file = st.selectbox(
        label="Choose a file to chat with: ", options=os.listdir(SAVE_DIR)
    )

    if st.button("Clear all api keys"):
        st.session_state.openai_api_key = ""
        st.session_state.replicate_api_token = ""
        st.session_state.hf_token = ""
        st.success("All API keys cleared!")
        st.rerun()

    if st.button("Submit", key="submit", help="Submit the form"):
        with st.status("Loading ...", expanded=True) as status:
            try:
                st.write("Loading Model ...")
                llama_llm = load_llm(
                    model_name=model_name.strip(), source=source.strip()
                )
                if llama_llm is None:
                    raise ValueError("Model not found!")
                Settings.llm = llama_llm

                st.write("Processing Data ...")
                index = get_index(selected_file)
                # index = get_pinecone_index(selected_file)

                st.write("Finishing Up ...")
                llama_custom = LlamaCustom(model_name=selected_llm_name, index=index)
                # llama_custom = LlamaCustomV2(model_name=selected_llm_name, index=index)
                st.session_state.llama_custom = llama_custom

                status.update(label="Ready to query!", state="complete", expanded=False)
            except Exception as e:
                status.update(label="Error!", state="error", expanded=False)
                st.error(f"Error: {e}")
                st.stop()

with tab2:
    messages_container = st.container(height=300)
    show_previous_messages(framework="llama", messages_container=messages_container)
    show_chat_input(
        disabled=False,
        framework="llama",
        model=st.session_state.llama_custom,
        messages_container=messages_container,
    )

    def clear_history():
        messages_container.empty()
        st.session_state.llama_messages = [
            {"role": "assistant", "content": "How can I help you today?"}
        ]

        st.session_state.llama_chat_history = [
            ChatMessage.from_str(role="assistant", content="How can I help you today?")
        ]

    if st.button("Clear Chat History"):
        clear_history()
        st.rerun()