Research Assistant App

Sun, 25 Aug 2024 09:44:02 GMT

 Properties

Key Value
Identifier research-assistant-app
Name Research Assistant App
Type Topic
Creation timestamp Sun, 25 Aug 2024 09:44:02 GMT
Modification timestamp Mon, 09 Sep 2024 08:44:51 GMT

Top-level Flow

  1. Identify or describe a topic of interest
  2. Choose source
    • Wikipedia
    • arXiv
    • Other sources?
  3. Submit topic of interest
    • Semantic search against source to retrieve relevant documents
    • Fetch and store documents (e.g., PDFs) in topic directory
    • Index/embed the retrieved documents
  4. Notify user that topic of interest is available for ((G)RAG) querying
    • Allow the user to view the retrieved documents and accompanying (knowledge) graph
      • The user, at this point, can select / deselect individual PDFs
  5. Provide the user with a query interface
  6. Stream response

Miscellaneous

  • The user's topics of interest and related queries (together with the responses) are stored making it possible to re-visit "old" topics and accompanying queries / responses
  • Topics of interest can be viewed, deleted and re-freshed (the semantic search against the chosen source is re-submitted and the relevant documents are fetched and re-indexed/embedded)
    • The accompanying queries and responses
    • Set of retrieved documents
    • Graph visualization, etcetera

Back to top


Notes
Sun, 25 Aug 2024 10:41:14 GMT
LlamaIndex vs. LangChain

LlamaIndex excels in search and retrieval tasks. It’s a powerful tool for data indexing and querying and a great choice for projects that require advanced search. LlamaIndex enables the handling of large datasets, resulting in quick and accurate information retrieval.

LangChain is a framework with a modular and flexible set of tools for building a wide range of NLP applications. It offers a standard interface for constructing chains, extensive integrations with various tools, and complete end-to-end chains for common application scenarios.


Back to top