The workflow of this application has two distinct but overlapping parts, one of them is for the graded task the other one is for the more open-ended question.
INPUT: JSON with the same structure as source_products_<>_json
PARALLELIZE the work into 5 parallel branches, one for the given data the other for the scraping, one for each website.
For scraping (IN PARALLEL): Identify the most disintiguishing feature of the input product for the scraping (if the input is some json about a Samsung TV Oled, scrape all the TVs)
For scraping (IN PARALLEL): Once the scraping is done, format the scraped data in a similar fashion to the source json and forward it to the matching algorithm
For all (IN PARALLEL): Run the matching algorithm described in agent.md taking into account details such as priorities.
For all: Reunite the outputs from the 5 branches that are running the matching algorithm into 1 branch in a sensible way.
Submit the result
UI description will be given elsewhere. INPUT: product name
As user is typing give most relevant results from the source DB for the user to choose. Get the user's chosen product with some additional user priorities to filter based on the user's choice.
Run the same pipeline as the graded task from step 1 to step 4; This pipeline should take into account the user's specifications for the matching algorithm and maybe even for the scraping.
Return the output to the user, and allow the user to filter/order the outputs based on stuff like (price, or brand)