Upload main.py
Browse files
main.py
ADDED
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from langchain_openai import ChatOpenAI
|
2 |
+
from langchain_core.prompts import ChatPromptTemplate
|
3 |
+
from langchain_core.output_parsers import StrOutputParser
|
4 |
+
from langchain_core.runnables import RunnablePassthrough
|
5 |
+
from langchain_community.utilities.duckduckgo_search import DuckDuckGoSearchAPIWrapper
|
6 |
+
from langchain.schema.runnable import RunnableLambda
|
7 |
+
import requests
|
8 |
+
from bs4 import BeautifulSoup
|
9 |
+
from dotenv import load_dotenv
|
10 |
+
import os
|
11 |
+
import json
|
12 |
+
import streamlit as st
|
13 |
+
|
14 |
+
load_dotenv()
|
15 |
+
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
|
16 |
+
os.environ["LANGCHAIN_TRACING_V2"] = "true"
|
17 |
+
os.environ["LANGCHAIN_API_KEY"] = os.getenv("LANGCHAIN_API_KEY")
|
18 |
+
|
19 |
+
|
20 |
+
RESULTS_PER_QUESTION = 3
|
21 |
+
ddg_search = DuckDuckGoSearchAPIWrapper()
|
22 |
+
|
23 |
+
def web_search(query: str, num_results: int = RESULTS_PER_QUESTION):
|
24 |
+
results = ddg_search.results(query, num_results)
|
25 |
+
return [r["link"] for r in results]
|
26 |
+
|
27 |
+
summary_template = """
|
28 |
+
{text}
|
29 |
+
|
30 |
+
---------------------------
|
31 |
+
Using the above text, answer in short the following question:
|
32 |
+
|
33 |
+
> {question}
|
34 |
+
---------------------------
|
35 |
+
If the question cannot be answered using the text, imply summarize the text. Include all factual information, numbers, stats, etc if available.
|
36 |
+
"""
|
37 |
+
|
38 |
+
summary_prompt = ChatPromptTemplate.from_template(summary_template)
|
39 |
+
|
40 |
+
def scrape_text(url: str):
|
41 |
+
try:
|
42 |
+
response = requests.get(url)
|
43 |
+
if response.status_code == 200:
|
44 |
+
soup = BeautifulSoup(response.text, "html.parser")
|
45 |
+
page_text = soup.get_text(separator = " ", strip = True)
|
46 |
+
return page_text
|
47 |
+
else:
|
48 |
+
return f"Failed to retrieve the webpage: Status code {response.status_code}"
|
49 |
+
except Exception as e:
|
50 |
+
print(e)
|
51 |
+
return f"Failed to retrieve the webpage: {e}"
|
52 |
+
|
53 |
+
url = "https://blog.langchain.dev/announcing-langsmith/"
|
54 |
+
|
55 |
+
scrape_and_summarize_chain = RunnablePassthrough.assign(
|
56 |
+
summary = RunnablePassthrough.assign(
|
57 |
+
text = lambda x: scrape_text(x["url"])[:10000]
|
58 |
+
) | summary_prompt | ChatOpenAI(model = "gpt-4o") | StrOutputParser()
|
59 |
+
) | (lambda x: f"URL : {x["url"]}\n\nSummary:\n\n{x['summary']}")
|
60 |
+
|
61 |
+
web_search_chain = RunnablePassthrough.assign(
|
62 |
+
urls = lambda x: web_search(x["question"]),
|
63 |
+
) | (lambda x: [{"question": x["question"], "url": u} for u in x["urls"]]) | scrape_and_summarize_chain.map()
|
64 |
+
|
65 |
+
search_prompt = ChatPromptTemplate.from_messages(
|
66 |
+
[
|
67 |
+
(
|
68 |
+
"user",
|
69 |
+
"""
|
70 |
+
Write 3 google search queries to search online that form an objective
|
71 |
+
opinion from the following: {question}\n
|
72 |
+
You must respond with a list of strings in the following format:
|
73 |
+
[["query1"], ["query2"], ["query3"]]
|
74 |
+
""",
|
75 |
+
),
|
76 |
+
]
|
77 |
+
)
|
78 |
+
|
79 |
+
search_question_chain = search_prompt | ChatOpenAI(model = "gpt-4o") | StrOutputParser() | json.loads
|
80 |
+
|
81 |
+
full_research_chain = search_question_chain | (lambda x: list(map(lambda y: {"question": y[0]}, x))) | web_search_chain.map()
|
82 |
+
|
83 |
+
WRITER_SYSTEM_PROMPT = "You are an AI critical thinker research assistant. Your sole purpose is to write well written, critically acclaimed, objective and structured reports on given text."
|
84 |
+
|
85 |
+
RESEARCH_REPORT_TEMPLATE = """
|
86 |
+
Information:
|
87 |
+
--------
|
88 |
+
{research_summary}
|
89 |
+
--------
|
90 |
+
Using the above information, answer the following question or topic: "{question}" in a detailed report -- \
|
91 |
+
The report should focus on the answer to the question, should be well structured, informative, \
|
92 |
+
in depth, with facts and numbers if available and a minimum of 1,200 words.
|
93 |
+
You should strive to write the report as long as you can using all relevant and necessary information provided.
|
94 |
+
You must write the report with markdown syntax.
|
95 |
+
You MUST determine your own concrete and valid opinion based on the given information. Do NOT deter to general and meaningless conclusions.
|
96 |
+
Write all used source urls at the end of the report, and make sure to not add duplicated sources, but only one reference for each.
|
97 |
+
You must write the report in apa format.
|
98 |
+
Please do your best, this is very important to my career.
|
99 |
+
"""
|
100 |
+
|
101 |
+
prompt = ChatPromptTemplate.from_messages(
|
102 |
+
[
|
103 |
+
("system", WRITER_SYSTEM_PROMPT),
|
104 |
+
("user", RESEARCH_REPORT_TEMPLATE),
|
105 |
+
]
|
106 |
+
)
|
107 |
+
|
108 |
+
def collapse_list_of_lists(list_of_lists):
|
109 |
+
content = []
|
110 |
+
for l in list_of_lists:
|
111 |
+
content.append("\n\n".join(l))
|
112 |
+
return "\n\n".join(content)
|
113 |
+
|
114 |
+
chain = RunnablePassthrough.assign(
|
115 |
+
research_summary = full_research_chain | collapse_list_of_lists
|
116 |
+
) | prompt | ChatOpenAI(model = "gpt-4o") | StrOutputParser()
|
117 |
+
|
118 |
+
st.set_page_config(page_title="Research Assistant/Report Generation")
|
119 |
+
st.header("Research -> Report")
|
120 |
+
input = st.text_input("Input Question:", key = "input")
|
121 |
+
response = chain.invoke(input)
|
122 |
+
submit = st.button("Ask Question")
|
123 |
+
|
124 |
+
if submit:
|
125 |
+
st.subheader("Response:")
|
126 |
+
st.write(response.strip())
|