Arshad-Talpur

This is the third article in the series, the first covered Causal Lineage, why an agent did something. The second covered Session Replay, what actually happened, in order. This one covers Behavioral Drift, whether your agent is still behaving the way it did last week, last month, or on the day you shipped it

Debugging in multi-agentic systems is notoriously hard. ZizkaDB aims to solve that problem with step by step session replays.

Sometimes a very small journey leads to bigger discovery, I hope they keep it up and enlighten us more

I have genuine installations, but I agree with the other part. People are still in the deployment phase of AI, so auditability seems to be a future concern, and that’s something some users have told me as well.

Thank you so much for your words, and yes, English isn’t my first language. However, we are already going into design partnerships and trying our best to get things moving. We have already reached 300 active installations in different countries, but I honestly wanted to understand the science behind getting noticed on GitHub, because attention does bring that initial kick-off.

Though we are building a product for enterprises, so stars don’t necessarily convert. But in any case, thank you for your suggestion.

The list looks great but if i was the one to curate it, i would place Musk instead of Sam Altman

I have quite an story to tell, multiple time migrant, failures and success, I am usually not open to discuss but this 100 Github stars made me realize you cant get something unless you ask for, so here simple snippet of current milestone. I come from a small, poor village in Sindh, and I was probably the first person from it to go to university. I started in the insurance sector, then moved to Dubai and became an award-winning seller. After that came nine years in the startup world: one success, two grand failures, and four migrations and ended Up at Station F Paris (World’s Largest Startup Campus). I don’t complain about any of it, and I rarely even talk about it, because all I have is an amazing journey.

But today isn’t about me, It’s about my first 100 GitHub stars.

The Product Behind the Stars

I’m building ZizkaDB, an operational database that answers the why behind every agentic action. When an AI agent does something, most systems can tell you what happened. ZizkaDB is built to tell you why.

That is a notoriously hard problem, and it needs a specialized stack. Finding people who understand that stack, and who have the patience to build it with me, was harder still

Any comments or feedback from community here? I would love to hear

We are now at a crossroads, We are trying to work against the very science of probability that gave rise to LLMs and with them the whole AI phenomenon, It is a bit like asking the ocean to behave like a swimming pool.

Opensource database for Auditing Ai agents and session replay with behavioral drift measurment

Thanks for compliment, its just learned experience and believe on science not hype

Not only saying but building zizkadb for single purpose of auditability

We just released livekit sdk for native integration with zizkadb, would love to have community feedback specially those building voice agents.

Thanks in advance

The article for teams shipping AI agents into production

Causal Lineage and Session replay are two of the core components of AI Agents auditability, this blog explains how its being accomplished at ZizkaDB

though still needs some tweeks may be on design and function both sides but idea seems to be fantastic !

Historically developers no matter to which part of world they belong carried a a distinct Perhaps proud identity of being devevelopers, software engineers or tech nerds etc but the wave of AI is actually blurred the space, nowadays everyone is developer or atleast try to be one, and most importantly anyone who can sell themselves as developers are winning against those who really are, offcourse some of this true competitive advantage and some being promoted by LLM companies to justify their absurd vaulations!

I partially agree with that and that is what I argued, businesses may tolerate legal liability but they can not even sell to enterprises in eu with EU AI ACT compliance, second thing is recently we've seen anthropic is releasing so many new features basically killing top layers, so in other words vertical ai companies are just proving market exists once they do, they are replaceable

Hello community! There are many observability tools and databases so i tried to make this auditability platform and eu ai act compliant, looking forward to feedback and if anyone wants to contribute, go ahead.

Thanks

Hello Community, We are building this opensource operational database that makes ai agents and agentic systems auditable, reliable and EU AI ACT compliant. The core features are Causal Lineage db.why() , and Session Replay db.at(), we are getting initial good response but i would love to have some power testers here and give geniune feedback on the project

I dont know exactly but dont you think 90% of Vertical AI Companies are doing same?

reminds me of good old days of opensource, when we use to search everything on github, but good try, keep building

it looks super but I am more concerned about agentic determinism , because still they do change behavior, so curious how you are controlling it

The idea looks good, however I am of the opinion that LLMs cant be deterministic because of its inherent nature based on probablisitic mathematics, that is why we are building something to make AI agents auditable, if you want i can share it here its opensource

Looks cool, definitely worth a try

Is there any meetup in Paris?

What a tribute, what a way of writing, honestly i just felt i was living it

Why AI Agents Need the Right Goal, Not an Impossible One

I cant have an overall opinion but in my personal experience i have analysed that LLMs do reward concreteness

The role of developer is becoming more of a system thinker than a syntax writer, previously load balance was about what to do in a given timeline, now its more of what not to do , doing more is actually increasing technical debt.