Utterly disappointed by @airindia customer support. After delivering our luggage 5 days late and damaged, they have been delaying for the past 3 mons on our fair compensation.
Air India twitter account delaying by repeating the same message. And email is no help.
Expected more
Extremely disappointed with @airindia services.
Flight landed in New Delhi from Vancouver 3 days ago, and they have still yet to give an update for our lost 4 check-in bags.
Customer service cuts call whenever questioned.
@AirIndiaX@rajeshdogra7
Do something ASAP.
@Chaotic5555 @_nightsweekends@_buildspace I will probably end up with the Google Maps API. Its just the easiest to work with and works well with React Native. The other option I know of is MapBox. Not aware of any other API's🤔
I thought this would be a good community to get feedback on an idea I have been working on. So used the same template :)
Open to suggestions, questions, and feedback!
cc: @_nightsweekends@_buildspace
#SOSDay51 No update today either😥. Just back-and-forth with the professor on understanding the idea. Have started thinking about how the vision system would incorporate into all of this. There would need to be some sort of feedback loop within the LLM responses. So just thinking
#SOSDay50 No update today sadly. I talked to the professor, and we will be discussing the methodology for my memory system tomorrow. Hopefully, that helps. I am considering transitioning from the Memory module to the Vision system (using a vision model). Lets see🤷♂️.
#SOSDay49 Honestly. This Memory module rocking my brain. It just isn't what I am looking for and it being a blackbox llm, it makes it 10x harder to solve. I am reaching out to a professor who has worked in NLP. Hopefully that helps 🤞
#SOSDay48 I am back!!
Seeing the code from fresh eyes helped quite a bit. Started off with cleaning up some of the code, and found an obvious bug (storing the wrong pair) in the memory module. Currently working on fixing it and making it make sense.
#SOSDay35 to #SOSDay47. For anyone who's been following along my journey of creating Friday. Just wanted to let yall know I will not be posting until the 15th as I will be gone for a camp without access to my laptop. Will continue where I left off when I come back!
See you then!
#SOSDay34 Started off with watching the lecture for nights & weekends S5. Implemented the memory module with both retrieve and store. Still needs testing on different scenarios but is a bit slow so need to look into optimization. Starting off with preventing duplicates in memory.
#SOSDay33 Continued where I left off yesterday. Just writing out comments of where certain parts of the flow will go in the application. One step at a time...
(I love my gifs)
#SOSDay32 Not a lot but started the implementation of query to response flows with memory. This is gonna take a second but we will get it eventually.
Long day today.
#SOSDay30. Starting off, can't believe I posted about something for 30 days straight. Noice. In terms Friday, no shipping today but created flow charts for the user query involving the memory system. Turns out breaking a task into multiple steps leads to better results for LLMs
#SOSDay29 No shipping today. More video watching and planning the layout of the memory system and how/what I want to store inside memory.
Man this is confusing😕
#SOSDay28 Turns out a long-memory system with LLMs is a whole researching field so its a bit more difficult then I thought it would be. Found resources online that could help create a simple system, so going to use that to learn memory systems and build it over the next few days
I might be overcomplicating it, but thinking about using Phi3 to provide a short summary of the full interaction between the user and the assistant from first query to the end, and storing that in memory. So when a similar query comes along, it has a bit better/summarize context.
#SOSDay27
Welp the memory module isn't going well. I am currently storing every interaction between the user and the assistant as a document. Which returns the wrong context when asking a new question. I still need to add some checks based on distance between embeddings but yea.