Yashraj

AI Engineer & Full-Stack Dev

Building:
SupportGPT → AI support widget for SaaS
Cliy → component registry for AI agents

Open for retainer work ·

[Yashraj @yrjdev]

Most RAG systems work like this:
→ User asks question
→ Retrieve similar chunks
→ Send to LLM
→ Generate answer

The problem: question-to-chunk similarity is often weak.

Example: “How do I reset my password?”
That may not semantically match a documentation paragraph talking about authentication flows, account settings, or credential recovery.

So I’m experimenting with a different retrieval approach:
At ingest time: generate possible questions each page can answer.

At query time: match question-to-question instead of question-to-chunk.

Early results:
→ significantly better retrieval relevance
→ fewer unnecessary tokens sent to the LLM
→ much cleaner context windows

Calling this approach: QIndex-RAG.

Still testing and refining it, but the retrieval quality improvement is already noticeable.

[Yashraj @yrjdev]

@codeswithroh that's why I always keep my local and production keys separately and most of the time used the locally hosted keys only.

[Yashraj @yrjdev]

Secret Keys need to be Secret

[Rohit Purkait @codeswithroh]

never store private keys in plain .env files

@PatrickAlphaC keeps saying this for a reason

but people still do it because the safer workflow is usually clunky

so i built vaultenv

a small npm package to store secrets encrypted locally, then load them into your shell only when needed

npm install -g codeswithroh/vaultenv

[Yashraj @yrjdev]

Check your knowledge

[Yashraj @yrjdev]

@soham_nayak04

[Yashraj @yrjdev]

@codeswithroh back to work

[Yashraj @yrjdev]

@dayonefoundry Yes, Focus is the key

[Yashraj @yrjdev]

When you work with LLM APIs.

Before setting up the Response structure
Setup for request data, so LLM can get only the information that needed.

Otherwise
Garbage In, Garbage Out

[Yashraj @yrjdev]

Generating a simple tweet was costing us >>>>>>> 18,000+ tokens.

The prompt: ~400 tokens.
The tweet: 70 tokens.

So where did the other 17,500 go?

// GPT-5 Nano is a reasoning model. max_output_tokens doesn't just cap the output.

It caps reasoning + output combined.
The model was spending 17,000 tokens thinking before writing 70.

The fix wasn't a bigger cap.

It was a dynamic one:
>> inputTokens = ceil(promptChars / 4.5)
>> reasoning = max(4000, inputTokens × 5)
>> cap = outputBudget + reasoning

Why ×5?
Measured in production:
1,737 input → 7,273 reasoning tokens
That's x4.2, We use 5 for safety margin.

Short videos = small cap.
Long videos = large cap.

No waste. No truncation.

Reasoning models need reasoning budgets.
Not output limits.

[Yashraj @yrjdev]

@gamerhelo7 @O_Anu_O yes I know, good connections matter, but I also don't have connections to get a job.

[Yashraj @yrjdev]

Python Workout

[Yashraj @yrjdev]

Print the value:

If you understand slicing in Python,
Then you can answer this.

[Yashraj @yrjdev]

Print the value:

If you understand slicing in Python,
Then you can answer this.

[Yashraj @yrjdev]

@dayonefoundry waiting and too excited to join

[Yashraj @yrjdev]

@tdinh_me when the Next will happen?

[Yashraj @yrjdev]

@codeswithroh I'm iron man

[Yashraj @yrjdev]

@BatsouElef right, and I'm working on the same projects that become a layer for the coding agents

[Yashraj @yrjdev]

@codeswithroh same, i WANT THIS ONE

[Yashraj @yrjdev]

@kirat_tw Yes, I'd faced same in this apis, when working in normal is fine, but when try to optimize to the best outcome, token limit, dynamic capping, prompt template - all are really a battle not all the things works like described in docs

[Yashraj @yrjdev]

@codeswithroh Amazing bro, video looks really good.
keep enjoying