DocuChat is a full-stack web application developed using C#, .NET 8.0, and Blazor. The goal is to provide a cost-effective chat interface for users to interact with documents. It will prioritize data accuracy, minimize errors, and aim to deliver a high-quality user experience.

• Parses SEC documents and extracts relevant data. (build in progress)
• Use LLM and RAG to develop a chatbot capable of interpreting and responding to queries based on information extracted from SEC documents. (coming soon)

1. Ensure you have .NET 8 SDK installed on your machine. You can download it from the official .NET website.

2. Clone the repository which will open the solution:

   git clone https://github.com/your-repo/DocumentAPI.git 

3. Open the user secret file and paste this in the file. I am using Mac, the user secret file in Rider IDE is by right-click the project -> tools -> .NET user secrets.

   {"SecUserAgent": "Personal-Project/1.0 (+{{your email}@gmail.com)"}

4. Run the project.

   

   ParseDocuments

   curl --location 'http://localhost:5084/api/sec/sec-parser' \
   --header 'accept: */*' \
   --header 'Content-Type: application/json' \
   --header 'X-CALLING-APP: CompanyA' \
   --data '{
   "secDocumentUrls": [
   "https://www.sec.gov/Archives/edgar/data/320193/000032019319000119/a10-k20199282019.htm",
   "https://www.sec.gov/Archives/edgar/data/320193/000032019323000106/aapl-20230930.htm",
   "https://www.sec.gov/Archives/edgar/data/320193/000032019322000108/aapl-20220924.htm",
   "https://www.sec.gov/Archives/edgar/data/320193/000032019321000105/aapl-20210925.htm",
   "https://www.sec.gov/Archives/edgar/data/320193/000032019320000096/aapl-20200926.htm",
   "https://www.sec.gov/Archives/edgar/data/789019/000156459020034944/msft-10k_20200630.htm",
   "https://www.sec.gov/Archives/edgar/data/1318605/000162828024002390/tsla-20231231.htm"
   ],
   "secDocumentTypeEnum": 1
   }'

   

   BatchGetDocumentUrls

   curl --location 'http://localhost:5084/api/sec/batch-get-sec-urls?formType=1&startDate=2019-04-30&endDate=2024-04-30' \
   --header 'accept: */*'
   }'

   

1. Data
   • Schema: https://www.sec.gov/files/form10-k.pdf / https://sec-api.io/docs/sec-filings-item-extraction-api
   • Scraping rule: https://www.sec.gov/os/webmaster-faq#developers
   • Search: https://www.sec.gov/edgar/search/#/q=Microsoft&category=custom&forms=10-K
2. Requirement
   • Start from 10K and 10Q forms.
   • Extract the specific sections from the forms.
   • Do a load testing, say 200 documents.

1. Algorithms
   • Levenshtein Distance for Measuring Text Similarity
2. Libraries
   • HtmlAgilityPack library to parse the HTML documents
   • Carter library for routing and handling requests, so I don't need to write my own filters from scratch and can have more time to focus on the business.
   • Polly library that provides resilience strategies in fluent-to-express policies such as Retry, WaitAndRetry, and CircuitBreaker, etc.
   • ...