Adebayo Musturpha
Region: NG
Tuesday 15 September 2026 19:18:06 GMT
5225
526
44
25
Music
Download
Comments
Testy :
Easy, before the start of operation daily all food available should be logged in the database with the quantity available, after each sales the cashier can remove the quantity sold out of the available quantity in the database( of course it can be automated l just trying to be plain ) after the remover it must take effect on the database, anytime the ai is asked which food is available the AI query the database to check for the food which still have quantity and of course the prices and even the quantity left in case the customer is asking for more than is available So let’s say we have 100 plate of jollof for sale at each sales the amount of plate sold must be deducted from the 100 anytime the AI is called it checked the database for the amount left and of course with the price 📌 Oya where the connection
2026-09-16 07:21:19
0
AUWAL AHMADU :
The safest approach is to make sure the AI never guesses whether an item is available. The real-time database should always be the source of truth. When a customer asks about an item, the backend first checks the restaurant’s inventory or POS system using a database query. The result—such as the item name, price, and whether it is in stock—is then passed to the AI as context. I would also add strict guardrails so the AI can only respond based on the retrieved data. For example, if the database says “Out of Stock,” the AI should tell the customer that the item is unavailable instead of assuming it is available. For a very large catalog, such as 1 million+ food items, I would use a combination of keyword and semantic search with something like Elasticsearch or a vector database such as Pinecone or Qdrant. This would help match queries like “spicy grilled chicken” to the correct menu item. For frequently requested items, Redis caching could also be used to reduce database load and improve response time. So the main principle is: the database determines availability, while the AI only explains the verified result to the user.
2026-09-16 14:43:52
4
THE-REGINALD :
I will design a RAG system using supabase as the database
2026-09-16 04:50:50
5
DUG :
You don't need AI for that. Simple logic and pattern match can do the job very well. Even on WhatsApp
2026-09-16 16:51:46
0
Victor123 :
Never let the ai know the list, like don't store the inventory in the AI, use function calling when somebody asks if a particular food is available the ai doesn't guess it simply takes the name and runs a function that checks if the food is available (this part might take some time but with some brainstorming it can be faster), let the data base do all the serious maths your acting back end queries the data base using indexing or vector search to find the item amongst the millions, then the database then spits back a flat response Available: true //with the price Unavailable: false Lock the system prompts, if the data base says it's unavailable, it's unavailable nothing else to it, if it's available it's available simple, the ai is just a receptionist no need to carry all menu in its head Mr Bayo it's a new pc I want not the job, I'm still a student it'll go a long way, thank U
2026-09-15 19:39:09
7
Dúdúyẹmi :
You connect the AI agent to the company live database. set up RAG system for the AI to always retrieve updated information whenever it needs it.
2026-09-16 16:28:22
0
leotech :
I wouldn’t train the AI to memorize your food database. I’d connect the AI directly to your existing system database. When a customer asks whether a food is available, the AI checks the live database and gives the customer the result. If the system can’t confirm it, the AI won’t guess. That way, even if you have millions of food records and the stock changes every minute, the answer comes from your actual restaurant data
2026-09-15 22:41:47
5
Jweb💚💯 :
It’s very simple, we don’t need to build an Ai from scratch, they are existing Models that give access to their api key, what happens in the backend is we receive the prompt from the customer send it to the Ai model to give us the intent of the customer send, if it’s buying, asking for price asking for product availability e.t.c we would already have functions of different intent on the backend, once the Ai interprets the intent it now triggers the function to call and perform that task, if let’s say the intent of customer is product availability, the Ai trigger the function that would go and get the product quantity and check if it’s available, i have built a project like that, and i use Django and Django rest framework
2026-09-16 10:04:42
0
thecodedan :
I believe id use tool/function calling.. the tool will call the db api to fetch the food data and display it to the user .. I believe that should work
2026-09-15 19:37:59
0
Billy :
you'll make sure to update the food service in the database and you'll connect the AI to it and make sure that before the AI give the list of available food it goes through that data and make sure it's available before giving the list
2026-09-15 19:27:27
2
Professor Egusi :
Just use vector db like rag and Gemini api
2026-09-16 16:53:37
0
Build_Stack :
The database should have categories to filter thereby AI should confirm that by asking then have call function for checking current stock count
2026-09-16 13:35:23
0
GABBY :
u have to have an agent then a hardware that checks it constantly to see if it's complete
2026-09-16 10:33:06
0
D E X T R A✱ :
Regex search
2026-09-16 10:15:25
0
Abdull Rockib :
I will put some guard drail and the IA most request from the database before answering so he will not hallucinate
2026-09-16 06:34:07
0
Westlove | Software Engineer :
streaming RAG
2026-09-16 09:13:41
0
vventures💙 :
Ai will read the database is very easy💯
2026-09-16 05:58:19
0
Lee joongi 이준기 :
The Ai would need a structure where every morning before work a human would upload the the availability and the recent amounts then it can work throughout the day..... Or week or depending on how the company restocks or the amount of food they prepare in a week... Then the workflow continues
2026-09-16 01:25:24
0
Salex :
I have an AI I built that can’t hallucinate I designed that purposely for cyber security and the easiest way is to use real time data beast for fast query, and all the problems is solved
2026-09-16 04:47:58
0
Aether :
yo just made my first ai model a few weeks back
2026-09-16 03:36:01
0
MrChuks / Ads Expert :
I’ll use Lovable and Claude. Thank you
2026-09-16 17:29:38
0
𝔼𝔻𝕀𝕀𝕄 🐾 :
First off,Connect the AI agent to the company’s real database/API, then you'd let the AI query it whenever it needs factual information so it only answer from the returned data and if there’s no match, say it doesn’t have it. So: AI is for the conversation 🧠 Database is equal to the source of truth 📦 Tools/API is basically connection 🔌 Validation also prevents hallucinations 🛡️
2026-09-16 06:19:52
0
MS AUtomation SOlutions :
Since they have thier menu; that gets updated when a food is ordered or when they want to restock ( more like a catalogue ). The AI can check or query the catalogue ( it might be in form of a sheet or thier app )
2026-09-16 05:37:45
0
Elisha Shadowbox :
is simple boss
2026-09-16 12:05:58
0
To see more videos from user @adebayo_musturpha, please go to the Tikwm
homepage.
