Spaces:
Running
Running
File size: 6,830 Bytes
b1c8f17 eb92a4f b5b2e6a b1c8f17 2e76cf7 22917d7 7183db2 2c7b10d b1c8f17 7bd26ee b1c8f17 2e76cf7 b1c8f17 576d9e2 b5b2e6a 270e05e b1c8f17 270e05e dca52fb 7bd26ee 24a6044 7bd26ee 1fc729a bdd570c 1fc729a bdd570c eb0c972 1fc729a 00eb6f3 1fc729a e7a043e 1fc729a 7183db2 7a5d35d 7183db2 1fc729a 7a5d35d 1fc729a 7a5d35d 545c7b4 1fc729a 7a5d35d e7a043e 1fc729a 6fe2220 e7a043e 1fc729a b1c8f17 00eb6f3 e7a043e b1c8f17 1fc729a c8510e0 7bd26ee |
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 |
from fastapi import FastAPI, HTTPException, Request, Query
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from typing import List, Dict, Any
from helper_functions_api import md_to_html, search_brave, fetch_and_extract_content, limit_tokens, together_response, insert_data
import os
from dotenv import load_dotenv, find_dotenv
from datetime import datetime, timedelta
from fastapi_cache import FastAPICache
from fastapi_cache.backends.inmemory import InMemoryBackend
from fastapi_cache.decorator import cache
import asyncio
import re
# Load environment variables from .env file
#load_dotenv("keys.env")
app = FastAPI()
@app.on_event("startup")
async def startup():
FastAPICache.init(InMemoryBackend(), prefix="fastapi-cache")
TOGETHER_API_KEY = os.getenv('TOGETHER_API_KEY')
BRAVE_API_KEY = os.getenv('BRAVE_API_KEY')
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
HELICON_API_KEY = os.getenv("HELICON_API_KEY")
SUPABASE_USER = os.environ['SUPABASE_USER']
SUPABASE_PASSWORD = os.environ['SUPABASE_PASSWORD']
# Groq model names
llm_default_small = "llama3-8b-8192"
llm_default_medium = "llama3-70b-8192"
# Together Model names (fallback)
llm_fallback_small = "meta-llama/Llama-3-8b-chat-hf"
llm_fallback_medium = "meta-llama/Llama-3-70b-chat-hf"
SysPromptJson = "You are now in the role of an expert AI who can extract structured information from user request. Both key and value pairs must be in double quotes. You must respond ONLY with a valid JSON file. Do not add any additional comments."
SysPromptList = "You are now in the role of an expert AI who can extract structured information from user request. All elements must be in double quotes. You must respond ONLY with a valid python List. Do not add any additional comments."
SysPromptDefault = "You are an expert AI, complete the given task. Do not add any additional comments."
SysPromptMd = "You are an expert AI who can create a structured report using information provided in the context from user request.The report should be in markdown format consists of markdown tables structured into subtopics. Do not add any additional comments."
sys_prompts = {
"offline": {
"Chat": "You are an expert AI, complete the given task. Do not add any additional comments.",
"Full Text Report": "You are an expert AI who can create a detailed report from user request. The report should be in markdown format. Do not add any additional comments.",
"Tabular Report": "You are an expert AI who can create a structured report from user request.The report should be in markdown format structured into subtopics/tables/lists. Do not add any additional comments.",
"Tables only": "You are an expert AI who can create a structured tabular report from user request.The report should be in markdown format consists of only markdown tables. Do not add any additional comments.",
},
"online": {
"Chat": "You are an expert AI, complete the given task using the provided context. Do not add any additional comments.",
"Full Text Report": "You are an expert AI who can create a detailed report using information scraped from the internet. You should decide which information is relevant to the given task and use it to create a report. The report should be in markdown format. Do not add any additional comments.",
"Tabular Report": "You are an expert AI who can create a structured report using information scraped from the internet. You should decide which information is relevant to the given task and use it to create a report. The report should be in markdown format structured into subtopics/tables/lists. Do not add any additional comments.",
"Tables only": "You are an expert AI who can create a structured tabular report using information scraped from the internet. You should decide which information is relevant to the given task. The report should be in markdown format consists of only markdown tables. Do not add any additional comments.",
},
}
class QueryModel(BaseModel):
topic: str = Query(default="market research", description="input query to generate Report")
description: str = Query(default="", description="additional context for report")
user_id: str = Query(default="", description="unique user id")
user_name: str = Query(default="", description="user name")
internet: bool = Query(default=True, description="Enable Internet search")
output_format: str = Query(default="Tabular Report", description="Output format for the report",
enum=["Chat", "Full Text Report", "Tabular Report", "Tables only"])
data_format: str = Query(default="Structured data", description="Type of data to extract from the internet",
enum=["No presets", "Structured data", "Quantitative data"])
#@cache(expire=604800)
async def generate_report(query: QueryModel):
query_str = query.topic
description = query.description
user_id = query.user_id
internet = "online" if query.internet else "offline"
sys_prompt_output_format = sys_prompts[internet][query.output_format]
data_format = query.data_format
optimized_search_query = ""
all_text_with_urls = [("", "")]
if query.internet:
search_query = re.sub(r'[^\w\s]', '', description).strip()
try:
urls, optimized_search_query = await search_brave(search_query, num_results=4)
all_text_with_urls = fetch_and_extract_content(data_format, urls, query_str)
additional_context = limit_tokens(str(all_text_with_urls))
prompt = f"#### COMPLETE THE TASK: {description} #### IN THE CONTEXT OF ### CONTEXT: {query_str} USING THE #### SCRAPED DATA:{additional_context}"
except Exception as e:
print(e)
query.internet = False
print("failed to search/scrape results, falling back to LLM response")
if not query.internet:
prompt = f"#### COMPLETE THE TASK: {description} #### IN THE CONTEXT OF ### CONTEXT: {query_str}"
md_report = together_response(prompt, model=llm_default_medium, SysPrompt=sys_prompt_output_format)
if user_id != "test":
insert_data(user_id, query_str, description, str(all_text_with_urls), md_report)
references_html = {}
for text, url in all_text_with_urls:
references_html[url] = str(md_to_html(text))
return {
"report": md_to_html(md_report),
"references": references_html,
"search_query": optimized_search_query
}
@app.post("/generate_report")
async def api_generate_report(request: Request, query: QueryModel):
return await generate_report(query)
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],) |