Skip to content

Latest commit

 

History

History
58 lines (39 loc) · 2.86 KB

core-functionalities.md

File metadata and controls

58 lines (39 loc) · 2.86 KB

AIEP Core Functionalities Guide

This guide provides an overview of the three main functionalities of the AIEP application: File Uploading, Generating Summaries, and Interacting with the Chatbot.

1. Uploading Files

Users can upload files and retrieve all uploaded files through specific endpoints in the backend. Follow these steps:

Upload Files

  • Endpoint: /jobs/create
  • Location of Code: app-backend/app/jobs.py
  • Testing: Use the FastAPI playground at AIEP API Docs.
  • Sample Files: Test with sample files available in the Google Drive folder.

Retrieve All Uploaded Files

  • Endpoint: /jobs/all-jobs
  • Instructions: This endpoint returns a list of all uploaded files. You can also test it in the FastAPI playground.

Notes

  • Ensure you’ve processed the file before accessing the summary or interacting with it in the CMS.

2. Generating Summaries

Summaries are generated automatically after a file upload job is completed.

Accessing Summaries

  • CMS Location: Once the job is finished, you can view the processed summaries in the Payload CMS.
  • Note: Do not directly access or modify items in the MongoDB database. Always use the CMS for any data management.

Frontend Display

  • Summary Cards Structure: Summaries are displayed in the frontend with a structure similar to the schema defined in app-backend/app/extract.py.
  • Frontend Data Handling: When a processed job is selected, its data is stored in a zustand store located at app-frontend/src/store/jobStore.js.

Developer Tip

For frontend development, use the summary schema from extract.py for consistent data structure.

3. Chatbot Interaction

The chatbot allows users to interact with uploaded content through semantic retrieval.

Implementation Details

  • Code Location: app-frontend/src/app/[locale]/api/chat/route.js
  • Backend Integration: Initially, this functionality was integrated with AstraDB, but the current on-prem setup does not support this.
  • Current Database Solution: In the on-prem Dockerized version, Qdrant is being explored as an open-source, self-hosted vector database solution.

Data Embeddings with Qdrant

  • Job Processing: The /jobs/create endpoint loads processed output data as embeddings into the Qdrant collection, using job_id as collection_id.
  • Local Testing: When running locally, view items in Qdrant at localhost:6333/dashboard.

Data Retrieval

  • Endpoint: /rag/doc-search — retrieves semantically related data.
  • Testing: You can test this endpoint via the hosted UI playground; however, it is currently not integrated in the frontend.