Looking for AI models for my SAAS
m looking for a fast, lightweight, and free (or very cheap) AI model for the first stage of my SaaS, which is similar to Notion or Obsidian. The AI should quickly format and organize programming notes, making them clean and structured without delay.
I don’t need anything super smart—just something fast, efficient, and low-cost. It should work instantly so developers can take notes without interruption.
I’m open to free, self-hosted models or cheap API alternatives. If there’s a way to define rules for formatting, that would be a bonus....
Best practices using neverthrow w/ RQ
Like in the title - I wanted to ask if anyone used those two together, and in turn how to get best type inference out of the two - I assume custom hooks like
useNeverthrowQuery that intercept the queryFn / mutationFn and turn it back to promises? Not sure if the API of neverthrow allows for such conversionHow do you manage to debug the server side in VS Code?
I've been trying various configurations, but the breakpoints are unbound for some reason.
configurations I tried:
```
{...
NextJS `npm run build`, warnings and errors don't have colours?
At some point my NextJS t3stack warnings and errors stopped having yellow for warnings and red for errors. Finding the issues has now become a lot slower.
What do I google or how do I check why this is happening?...
T3 Pro Plan Context Window Size
According to the T3 FAQ, the Pro plan provides 1500 standard message credits per month, plus 100 premium credits for Claude. I'm wondering about the context window/token limits. Specifically, what is the maximum number of tokens I can send per message for each model available through T3? Is there a fixed token limit that applies to all models, or does it vary?
Solution:
I received an official response regarding the context window size limits for the T3 Chat models and thought it might be helpful to share:
Most models on T3 Chat support the full context window size. There's no message limit, other than the model's overall context window limit. The exception is Anthropic models (Claude), which are capped at 32K tokens.
Here's a breakdown of the context window sizes for each model:...
Censored responses from AI's
Has anyone tested prompting certain LLM's with censored information that is validated later rather than earlier?
Does T3 chat just delete the previous tokens? or do they just send everything the API has sent and saves it to the local storage and when the api deletes it's response it just adds the deleted to the end?...
Versions issue with shadcn/tailwind-4 and React 18
I'm unable to find much info about this maybe unusual setup(?)
Starting new (vite) project, so I naturally want to use latest shadcn/tailwind.
But also trying to stay on React 18 to avoid (yet unknown) possible issues, because I plan to use this project as remote in with module-federation that has other react 18 apps.
...
Are bangs in T3 Chat a good idea?
Solve a knapsack problem in Python !r1l
What color is the banana? !g2m
Write me a proposal !o1
and so on...
...T3 Chat History Deletion vs Privacy Policy
Delete Chat History says: This will permanently delete all your chat history from both your device and our servers (if sync is enabled). This action cannot be undone.
However, their privacy policy claims they retain message history "User Data". Is the User Data deleted after? Or is it only removed from your view while they retain a copy of all your information....
switching between models in the same chat
After switching between models in a chat, does the new model "know" everything that was said previously, or is it hallucinating the information? Also, could this be used to trick Claude into bypassing its token limit per session?
is it ok to store to use a user unique id as a Session_id in LocalStorage ?
I am working on a chat application I wanted to send online friends to the current user what i plan to do is to fetch the user friends from the database I have a socket server everytime the user logs in an object will be added to the map the object contains user info such as the username , email, etc.. I am thinking of having the UniqueId of each user object in the map as the key the reason is i believe it will be easier and faster to map online friends to the friends array returned to the...
Updating a different table based on changes from a settings table
How does one implement this in a way that won't crash a db?
For example, a different table (T1) references a central settings(T2) table, and when that setting changes, ALL rows in T1 must update with that new setting
To make it more complicated, T1 has rows which use the setting for calculations, so those values will have to be recalculated as well
----...
Running multiple projects from one database (Drizzle)
I'm pretty new to backend concepts and drizzle. I have created one project already using T3 stack + drizzle and have the database on Supabase.
It's deployed and working as expected but I am now looking at starting another project.
I am aware that the following bit of table creation code effectively allows me to run multiple projects on one database as all tables for that project are prefixed:...
hello bro i'm perusing this boot-camp https://fullstackopen.com/en/ i've just intermediate education
hello bro i'm perusing this boot-camp https://fullstackopen.com/en/ i've just intermediate education but i'm learning web dev from last two plus years as part time from various online courses but i failed to get even single interview call i decided to get certification in know basics so i made a plan to get certs like https://fullstackopen.com/en/ and from cs50 https://fullstackopen.com/en/ from and for learning core cs concepts i choose https://www.scaler.com/topics/courses/ to learn dbms oo...
Is 1500 messages enough to get me through studying for 5 midterms
I usually get chatgpt pro when i have exams to help me study and thought if i can save the money and pay 8$ here but i am not sure if the 1500 messages would be enough for me (Engineering Student) and i would gladly get another subscription if i hit the limit just thought i would ask first
Modal Getting Cut Off on iOS Safari but not on small desktop screen sizes
I'm working on a modal component for the project card's on a website I'm making for a friend. The modal pops up and looks fine on all browsers on desktop, even when simulating mobile screen sizes. However on iOS I'm noticing the top and bottom of the popup get cutoff, and I can't scroll. Not really sure if there's some webkit specific styling I need to add?
"url" parameter is valid but upstream response is invalid
I keep getting this when I use the Image from next js. However when I pass loader={(src) => {src}} it works, for whatever reason I can't seem to figure out what is going on, anyone have any ideas????
I get:
500 Internal Server Error
...
Advice please
Hi guys,
I'm a software engineer from Morocco, and I recently started my first job at a startup that leverages AI to build a large SaaS platform aimed at helping companies with their recruitment process. The software is not live yet, so I can't share a demo with you.
I joined this company as a broke student with barely a penny, but now I'm doing a bit better for my age. Since it's a startup, they don't pay me very well, and the product is still in development. We expect to launch in 1 to 2 months. (By the way, I still have one year left before I graduate.)...
Local LLM for Code
Hey all,
I am planning on getting a 5090 for gaming but would love if this could double as a tool for running LLMs locally.
In this fast changing LLM world, can anybody recommend some local models that would pair nicely with Roo Code or similar VS Code extensions?...