Building an Intelligent QA/Chatbot with LangChain and Open Source LLMs
| dc.contributor.author | Bogusz, William | en |
| dc.contributor.author | Mohbat, Cedric | en |
| dc.contributor.author | Liu, James | en |
| dc.contributor.author | Neeser, Andrew | en |
| dc.contributor.author | Sigua, Alex | en |
| dc.date.accessioned | 2024-09-03T18:25:13Z | en |
| dc.date.available | 2024-09-03T18:25:13Z | en |
| dc.date.issued | 2024-05-01 | en |
| dc.description.abstract | We have created a web application enabling access to Intelligent Q/A chatbots, where the end user has access to query language learning models to retrieve context specific information. This web application will provide a collection-based interface, where documents uploaded by the user provide the context for responses by the language learning model to user input. This is accomplished through retrieval augmented generation (RAG) pipeline. As to reduce inaccuracies and fulfill the user needs of the client, the language learning model will notify the user if a query cannot be sufficiently answered given the documents in a collection. As such, our application emphasizes collection management with the functionality to upload (in .txt, .html or .zip format) and delete documents as well as select specific collections, while providing a familiar interface not much different from the web interface for established AI chatbot services such as OpenAI’s ChatGPT or Anthropic’s Claude. The final product also currently encompasses a landing page and user login, with accessibility to a document upload portal for creating document collections. | en |
| dc.format.mimetype | application/pdf | en |
| dc.identifier.uri | https://hdl.handle.net/10919/121057 | en |
| dc.language.iso | en | en |
| dc.rights | Creative Commons Attribution-NonCommercial 4.0 International | en |
| dc.rights.uri | http://creativecommons.org/licenses/by-nc/4.0/ | en |
| dc.title | Building an Intelligent QA/Chatbot with LangChain and Open Source LLMs | en |
| dc.type | Report | en |
| dc.type.dcmitype | Text | en |