File size: 4,338 Bytes
3ad9a49
 
 
114ce4a
f3576a5
 
 
 
 
 
 
 
 
 
4e65999
 
114ce4a
4e65999
 
 
 
 
f3576a5
4e65999
 
 
f3576a5
4e65999
3ad9a49
e19a241
f3576a5
 
 
e19a241
f3576a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4e65999
 
 
 
 
 
 
 
 
 
 
f3576a5
 
 
 
 
 
 
 
 
 
114ce4a
3ad9a49
114ce4a
f3576a5
 
 
 
 
 
 
 
114ce4a
4e65999
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3ad9a49
4e65999
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# AI assistant with a RAG system to query information from the CAMELS cosmological simulations using Langchain
# Author: Pablo Villanueva Domingo

import gradio as gr
from langchain import hub
from langchain_chroma import Chroma
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_community.embeddings import HuggingFaceInstructEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_mistralai import ChatMistralAI
from langchain_community.document_loaders import WebBaseLoader
from langchain_core.rate_limiters import InMemoryRateLimiter

# Load documentation from urls
def get_docs():

    # Get urls
    urlsfile = open("urls.txt")
    urls = urlsfile.readlines()
    urls = [url.replace("\n","") for url in urls]
    urlsfile.close()

    # Load, chunk and index the contents of the blog.
    loader = WebBaseLoader(urls)
    docs = loader.load()

    return docs

# Join content pages for processing
def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)

# Create a RAG chain
def RAG(llm, docs, embeddings):

    # Split text
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    splits = text_splitter.split_documents(docs)

    # Create vector store
    vectorstore = Chroma.from_documents(documents=splits, embedding=embeddings)

    # Retrieve and generate using the relevant snippets of the documents
    retriever = vectorstore.as_retriever()

    # Prompt basis example for RAG systems
    prompt = hub.pull("rlm/rag-prompt")

    # Create the chain
    rag_chain = (
        {"context": retriever | format_docs, "question": RunnablePassthrough()}
        | prompt
        | llm
        | StrOutputParser()
    )

    return rag_chain

# Define a limiter to avoid rate limit issues with MistralAI
rate_limiter = InMemoryRateLimiter(
    requests_per_second=0.1,  # <-- MistralAI free. We can only make a request once every second
    check_every_n_seconds=0.01,  # Wake up every 100 ms to check whether allowed to make a request,
    max_bucket_size=10,  # Controls the maximum burst size.
)

# Get docs
docs = get_docs()
print("Pages loaded:",len(docs))

# LLM model
llm = ChatMistralAI(model="mistral-large-latest", rate_limiter=rate_limiter)

# Embeddings
embed_model = "sentence-transformers/multi-qa-distilbert-cos-v1"
# embed_model = "nvidia/NV-Embed-v2"
embeddings = HuggingFaceInstructEmbeddings(model_name=embed_model)

# RAG chain
rag_chain = RAG(llm, docs, embeddings)

# Function to handle prompt and query the RAG chain
def handle_prompt(message, history):
    try:
        # Stream output
        out=""
        for chunk in rag_chain.stream(message):
            out += chunk
            yield out
    except:
        raise gr.Error("Requests rate limit exceeded")


if __name__=="__main__":

    # Predefined messages and examples
    description = "AI powered assistant which answers any question related to the [CAMELS simulations](https://www.camel-simulations.org/)."
    greetingsmessage = "Hi, I'm the CAMELS DocBot, I'm here to assist you with any question related to the CAMELS simulations."
    example_questions = [
                        "How can I read a halo file?",
                        "Which simulation suites are included in CAMELS?",
                        "Which are the largest volumes in CAMELS simulations?",
                        "Write a complete snippet of code getting the power spectrum of a simulation"
                        ]

    # Define customized Gradio chatbot
    chatbot = gr.Chatbot([{"role":"assistant", "content":greetingsmessage}],
                        type="messages",
                        avatar_images=["ims/userpic.png","ims/camelslogo.jpg"],
                        height="60vh")

    # Define Gradio interface
    demo = gr.ChatInterface(handle_prompt,
                            type="messages",
                            title="CAMELS DocBot",
                            fill_height=True,
                            examples=example_questions,
                            theme=gr.themes.Soft(),
                            description=description,
                            cache_examples=False,
                            chatbot=chatbot)

    demo.launch()