File size: 3,105 Bytes
9c6fc06
 
 
 
f7ad073
9c6fc06
 
 
 
f7ad073
9c6fc06
 
 
f7ad073
9c6fc06
f7ad073
 
 
 
9c6fc06
 
a5674a9
9c6fc06
 
 
 
 
 
 
 
 
 
97fb706
9c6fc06
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5a0c146
9c6fc06
 
 
 
 
 
 
 
 
 
 
 
 
 
f4b3f52
9c6fc06
e7b7a08
5ef98b9
7c15df0
5ef98b9
 
36414cf
5ef98b9
f4b3f52
35a1789
f4b3f52
9c6fc06
 
 
f4b3f52
 
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
import os

import gradio as gr

from openai import OpenAI

from langchain_community.embeddings.sentence_transformer import SentenceTransformerEmbeddings
from langchain_community.vectorstores import Chroma

client = OpenAI(api_key=os.environ['OPENAI_API_KEY'])

embedding_model = SentenceTransformerEmbeddings(model_name='thenlper/gte-small')

tesla_10k_collection = 'tesla-10k-2019-to-2023'

vectorstore_persisted = Chroma(
    collection_name=tesla_10k_collection,
    persist_directory='./tesla_db',
    embedding_function=embedding_model
)

retriever = vectorstore_persisted.as_retriever(
    search_type='similarity',
    search_kwargs={'k': 5}
)

qna_system_message = """
You are an assistant to a financial services firm who answers user queries on annual reports.
Users will ask questions delimited by triple backticks, that is, ```.
User input will have the context required by you to answer user questions.
This context will begin with the token: ###Context.
The context contains references to specific portions of a document relevant to the user query.
Please answer only using the context provided in the input. However, do not mention anything about the context in your answer. 
If the answer is not found in the context, respond "I don't know".
"""

qna_user_message_template = """
###Context
Here are some documents that are relevant to the question.
{context}
```
{question}
```
"""

def predict(user_input):

    relevant_document_chunks = retriever.get_relevant_documents(user_input)
    context_list = [d.page_content for d in relevant_document_chunks]
    context_for_query = ".".join(context_list)
    
    prompt = [
        {'role':'system', 'content': qna_system_message},
        {'role': 'user', 'content': qna_user_message_template.format(
            context=context_for_query,
            question=user_input
            )
        }
    ]

    try:
        response = client.chat.completions.create(
            model="gpt-4o-mini",
            messages=prompt,
            temperature=0
        )

        prediction = response.choices[0].message.content

    except Exception as e:
        prediction = e
        
    return prediction


textbox = gr.Textbox(placeholder="Enter your query here", lines=6)

demo = gr.Interface(
    inputs=textbox, fn=predict, outputs="text",
    title="AMA on Tesla 10-K statements",
    description="This web API presents an interface to ask questions on contents of the Tesla 10-K reports for the period 2019 - 2023.",
    article="Note that questions that are not relevant to the Tesla 10-K report will not be answered.",
    examples=[["What was the total revenue of the company in 2022?", "$ 81.46 Billion"],
              ["Summarize the Management Discussion and Analysis section of the 2021 report in 50 words.", ""],
              ["What was the company's debt level in 2020?", ""],
              ["Identify 5 key risks identified in the 2019 10k report? Respond with bullet point summaries.", ""]
             ],
    cache_examples=False,
    concurrency_limit=16
)


demo.queue()
demo.launch(auth=("demouser", os.getenv('PASSWD')))