search

rmusser01 / tldw

Too Long, Didn't Watch(TL/DW): Your Personal Research Multi-Tool - Open Source NotebookLM
Apache License 2.0
45 stars 2 forks source link

Improvement: Implement the option to perform re-prompting against a transcription stored within the SQLite DB as part of the UI #27

Open rmusser01 opened 1 month ago

rmusser01 commented 1 month ago

As a user of the application where all prompts, URLs, and summaries are recorded in a SQLite DB,

I would like to be able to review prior URLs, associated transcriptions, and generated summaries for that transcription.

Further, I would like to be able to re-prompt or re-summarize, already generated transcriptions, through buttons in the GUI.

As a user, I should be able to review what transcriptions I have, and be able to select one (or multiple) for re-summarization according to my current summarization prompt text from within the GUI, and have the resulting summarizations added to the SQLite DB, and stored appropriately. (I'm not a DBA...)

So that, at a later point in time, as a user, I may come back to the application, look up a prior transcript + summarization, and re-perform summarization with a different prompt, or with a different LLM target.

This issue is reliant on Issue #26 to be completed first.

rmusser01 commented 1 month ago

https://github.com/danielmiessler/fabric

rmusser01 commented 1 month ago

Revisiting this issue, I have created an additional tab which will act as a control interface for configuring + launching the the llamafile inference engine, while also acting as a UI stand-in for the Llamafile inference engine, so that a user may only interact with the singular browser window, as opposed to having multiple open.

Have implemented an iframe solution using the default port for llamafile (8080), easy enough to modify by hand for now for those interested. - Modify the line <iframe src="http://127.0.0.1:8080" title= to be able to change the local port for now.

rmusser01 commented 4 days ago

Partially done. Added a new tab which allows the user to select a media item within the DB and have the title/content/summary shipped off with a prompt/custom prompt.

Need to modify the API calls to add ability to specify user/system prompt.