Apr 12, 2026 Enterprise Brain Scrum - Transcript 00:00:00 Yeshwanth Reddy Yerraguntla: Hi bro. Rajashekar G: Hi and good Yeshwanth Reddy Yerraguntla: Good Rajashekar G: morning. Yeshwanth Reddy Yerraguntla: morning. Rajashekar G: Did you get any chance to go through the branch? Yeshwanth Reddy Yerraguntla: Sorry. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: Let me go through it right now. Rajashekar G: Okay, that's Yeshwanth Reddy Yerraguntla: Beta design is it or just Rajashekar G: beta. Yeshwanth Reddy Yerraguntla: beta? Okay. Rajashekar G: Okay. So other than that I'm sharing my screen also now. No, in the current branch we are having uh one issue like UI agent history prop UI agent we did some Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: changes one person will working on that next UI enhancements. down like Yeshwanth Reddy Yerraguntla: Okay. What do you mean? History is not propagating. Rajashekar G: yes I mean history means chat Yeshwanth Reddy Yerraguntla: Huh? What do you mean? Rajashekar G: session last five messages or six meages agent context that is not able to pass to the UI Yeshwanth Reddy Yerraguntla: Right. Rajashekar G: agent Yeshwanth Reddy Yerraguntla: So we are dealing with A2A protocol correct. 00:02:40 Rajashekar G: Yes, Yeshwanth Reddy Yerraguntla: So how are you planning to send the history to other agents? Rajashekar G: actually right now uh history pro RT agent will come pass it to that like whenever ID Yeshwanth Reddy Yerraguntla: Oh, I want to Oh, Rajashekar G: based last message session ID Yeshwanth Reddy Yerraguntla: you okay? Rajashekar G: based Yeshwanth Reddy Yerraguntla: So, if I understand it correctly, you're dumping all the history into radius and you're fetching it. Rajashekar G: right now not into radius it's it's into database like session create it will create a session in Yeshwanth Reddy Yerraguntla: Uh Rajashekar G: my database and it will add all the charts messages in that database continuously when a new question arise yes post Yeshwanth Reddy Yerraguntla: the post dating. Yeah, Rajashekar G: database when a new question arrives it will fetch the last Yeshwanth Reddy Yerraguntla: whichever it is fine. Okay. Rajashekar G: five uh questions and answers based on that particular ID and pass it as a Yeshwanth Reddy Yerraguntla: Okay. Ideally speaking, Rajashekar G: context Yeshwanth Reddy Yerraguntla: uh, okay, fine. Fine. 00:03:47 Yeshwanth Reddy Yerraguntla: Uh, okay. This the architecture sounds fine to me. So, why do you think there is a bug? Rajashekar G: I have tested like previously when there is only core agent and data steel agent it was working fine we introduced UI agent into And on the first day it did work but later uh uh when we are doing with UI enhancements and all we did lot of changes like we tested it Yeshwanth Reddy Yerraguntla: H Rajashekar G: like first we'll ask a question it will generate a chart again I ask like Yeshwanth Reddy Yerraguntla: M Hm. Rajashekar G: uh convert this chart into another graph it is not changing and Yeshwanth Reddy Yerraguntla: H. Rajashekar G: whenever I go into log file and see I'm not able to see the previous uh history has a context there. That's where the issue we found it on Friday Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: evening. That can be fixed. Not yet. Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: And uh the cache implementation we kept hold in the last sprint. Yeshwanth Reddy Yerraguntla: Okay. 00:04:53 Rajashekar G: And first priority we are having is this U agent has to work well and uh Yeshwanth Reddy Yerraguntla: Oops. Rajashekar G: whatever the widgets they have given in those widgets I have few concerns like a line graph and whenever line graph comes like there will be more rows they asked us to horizontal scroll to the more data that is one work uh is happening and uh I have a concern to render like tables So whenever we are rendering a table right now we are going and hitting and getting all the data Yeshwanth Reddy Yerraguntla: H. Rajashekar G: right but representation it need to have a page nation and all that repentation they haven't provided us. So that also we need to take care. Yeshwanth Reddy Yerraguntla: Correct. Pagination will be important. Rajashekar G: So table widget they haven't given from the design team. So I asked them like discuss with them and we'll pick that also. Yeshwanth Reddy Yerraguntla: Yeah. Rajashekar G: So stabilizing UA uh UI agent with those widgets and deploying is the first task for today. And uh next one for this print I have planned all these 00:06:08 Yeshwanth Reddy Yerraguntla: Can you share your screen share Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: stopped? Rajashekar G: So these are the sprint points and there is still bandwidth to add few on the priority. I have kept it like this. Caching mechanism implementation. Web soockets implementation is needed when we are dealing with the interactive ths and training life cycle Yeshwanth Reddy Yerraguntla: Perfect. Rajashekar G: v1 is not completed yet. Uh what pan is doing paw added some guidelines and those guidelines are pushing into the system from whenever uh new question comes. Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: So that's one thing and uh train life v2 is user should be able to upload a document that can be considered as a new instructions. Yeshwanth Reddy Yerraguntla: I see where Rajashekar G: Next uh the layout that may come from the design Yeshwanth Reddy Yerraguntla: will where will the document upload happen in the chat interface Rajashekar G: table in the charge Yeshwanth Reddy Yerraguntla: only Rajashekar G: interface or we will give a trainer role and give him another tab or something we need to figure. Yeshwanth Reddy Yerraguntla: better you give only the dedicated tab that like Right now we never 00:07:27 Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: spoke about uh document uploading chat interface. Rajashekar G: Right. Yeshwanth Reddy Yerraguntla: So don't add that. I have a dedicated uh trainer page that that I'm fine Rajashekar G: Okay. Sure. Yeshwanth Reddy Yerraguntla: with. Rajashekar G: And uh this part like in the last discussion we had with Navincer he discussed on few things like UI layout intelligence should come. Yeshwanth Reddy Yerraguntla: Correct. Oh, correct. On Friday, we had this. Rajashekar G: Yes that one also shall be prioritized right now or later this and interactive charts go hand in hand. Yeshwanth Reddy Yerraguntla: We should we should spend some time first uh dedicating ourselves to architecture, right? What is that interchange uh exchange format? Rajashekar G: Right. Yeshwanth Reddy Yerraguntla: How will AI tell the user interface? What plots to put and how which order to plot them in? how to put them via interactions. These three are critical and we need to understand how to come up with the right JSON structure. Once we know the JSON structure, we'll be in a position to tell whether we can do it this uh sprint or next sprint. 00:08:42 Yeshwanth Reddy Yerraguntla: Without that, we'll not uh we don't have the clarity. Rajashekar G: Okay, Yeshwanth Reddy Yerraguntla: So, Rajashekar G: then. Yeshwanth Reddy Yerraguntla: we need that meeting. Ideally let's do it today only before uh uh if I will sleep at 11:00 Indian time before that we should have the Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: meeting Rajashekar G: Okay. Shall we discuss at 10:30? Okay. Yeshwanth Reddy Yerraguntla: Yeah. Rajashekar G: Next. So I kept these as another to stories moving all the brown files to structure DB bringing in other uh things into our new react architecture and right now there is no error handling for the LML like any AP getting failed or something it is again recali but that's Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: not in a proper architectured manner or something that you can handle the story Yeshwanth Reddy Yerraguntla: Okay. Yeah, Rajashekar G: and multi data source agent orchestration I kept it but I don't know like whether we will be picking this up or not Yeshwanth Reddy Yerraguntla: definitely. Rajashekar G: right now it is there but there will be new data source comes how we need to handle it right now we are here and few things will be spill over from here from the previous print one is training life cycle and new agent there are few 00:10:13 Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: fixes things will come into the right current screen right now and one more concern I have is u we'll Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: be having Rama in our team so she's u how can we plan her to like uh get things learned faster and uh enjoyed in our Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: tasks Yeshwanth Reddy Yerraguntla: Like what is her expertise? What does she what is she interested in? Rajashekar G: actually she is with the she joined with this team like Abilash Amulya's team only she joined in office at the same time she under like uh react front end back end skills she have I think she's expertise in react and uh the pentic course also Yeshwanth Reddy Yerraguntla: Yes. Rajashekar G: like whenever you train right few people in our office again. So those sessions he has attended. So right now we want to take her into our team and uh use it for our works also. You have one task in UI or something like that or shall we give her some Yeshwanth Reddy Yerraguntla: Okay. 00:11:37 Rajashekar G: videos? Yeshwanth Reddy Yerraguntla: You have like two days to understand what is going on. Let her come back. Uh if you still if you think no no we don't have that much time put her in UA Rajashekar G: But to understand also like what shall we give Har like the documentation part like videos whatever like Yeshwanth Reddy Yerraguntla: mainly she has to come back with uh a report of what is enterprise brain what are all the features uh um Rajashekar G: All Yeshwanth Reddy Yerraguntla: grouped by uh somehow she has to know that right she has to know there are there is art agent there are data storage data source agents UI agent what is the role of each agent all these things she has to know it's not that I want Rajashekar G: right. Yeshwanth Reddy Yerraguntla: you or one of us to sit her and explain to her she has to come up with that information by herself. We we are you collecting all the meetings that are happening on a day-to-day basis in minutes of meeting. Rajashekar G: last two days I missed. 00:12:59 Yeshwanth Reddy Yerraguntla: Okay. But before that everything is there. Rajashekar G: Yes. Yeshwanth Reddy Yerraguntla: Okay. Ahu. Yeah. That that is fine. the.com. It should have a lot of text files also. Rajashekar G: Yeah, texture is nice. Yeshwanth Reddy Yerraguntla: The same folder has all the raw matings also. Correct. Rajashekar G: Uh the skill you have given right meeting yeah Yeshwanth Reddy Yerraguntla: Uh that's all. Yeah. Rajashekar G: that's only I using Yeshwanth Reddy Yerraguntla: So it will o why Rajashekar G: that Yeshwanth Reddy Yerraguntla: we I don't see any July sorry April meetings in conversations Rajashekar G: difficult And Yeshwanth Reddy Yerraguntla: in chanaka. Rajashekar G: that's Yeshwanth Reddy Yerraguntla: Sorry vasa. We.com Rajashekar G: slash all for it. Yeshwanth Reddy Yerraguntla: Mhm. Rajashekar G: Dogs phase two conversations Yeshwanth Reddy Yerraguntla: H see I don't see a single 20264 Rajashekar G: 04. I have pushed it. Ma check. Yeshwanth Reddy Yerraguntla: I don't see it. So my point is uh with every m there is also associated txt file that uh uh that skill is putting. 00:15:21 Yeshwanth Reddy Yerraguntla: Basically, she just has to talk with uh uh her co-pilot and get back with understanding of what is good. You just give her a bunch of questions like these are the things you need to know very Rajashekar G: So, Yeshwanth Reddy Yerraguntla: thoroughly and let her come back with the Rajashekar G: okay. Yeshwanth Reddy Yerraguntla: answers. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: That's it. So once she's ready, maybe after lunch you can then assign talk to her for 15 minutes, see what she's interested in. if she has no preference. Uh um I was thinking UI but you are saying something right. Rajashekar G: UI or uh our enterprise back end also. We can take Yeshwanth Reddy Yerraguntla: Whichever it is fine Rajashekar G: 10 to 11 you have a meeting internal review meeting with Yanshall Yeshwanth Reddy Yerraguntla: that is with uh this guy's um genrx Rajashekar G: I keep it at 10:30 to Yeshwanth Reddy Yerraguntla: Sir P. Rajashekar G: Honey. Yeshwanth Reddy Yerraguntla: Oh my god. Okay. I don't have F when I have continuous work. 00:16:45 Yeshwanth Reddy Yerraguntla: Sure. Put it from 10:30 to 11. That synthetic data meeting. Uh, I Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: noted Rajashekar G: I'm keeping V3 like you and Okay. Don't know. Yeshwanth Reddy Yerraguntla: Sir, Rajashekar G: Yeah. Yeshwanth Reddy Yerraguntla: I wanted to tell you something before Rajashekar G: Yeah. Yeshwanth Reddy Yerraguntla: Pent came up with a new feature feature. Rajashekar G: Mhm. Yeshwanth Reddy Yerraguntla: Um, it's called capabilities. I'm sharing my screen. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: uh base sorry can see my screen right. Rajashekar G: Yes. Yeshwanth Reddy Yerraguntla: So a capability is uh nothing but a mix of tools hook uh hooks for uh agent when it is running before LLM call what to do after LLM call what to do before tool call what to do it's a collect hooks and a collection of tools so Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: basically MLX agent if you give the something like this uh capabilities What it does is it is directly attaching all the tools and behaviors to the LLM. Rajashekar G: Okay. 00:19:44 Yeshwanth Reddy Yerraguntla: what that I was checking here one such thing I was trying was to give a shared state capability that between uh agents of different uh microservices right if I have something like a shared uh uh postgress visibility I don't need to keep repeating my uh Rajashekar G: Yes. Yeshwanth Reddy Yerraguntla: tool calls again and again I'll have this one guy in some place common place and uh uh basically every agent will have the ability to fetch uh conversation history from Postgress and dump conversation to Postgress that kind of a thing. So just as an example if you see here after run is a thing where after every uh agent call along with giving the response it is also appending into postgress here I was appending to some fake radius. This way you don't need to your the the kind of code that you'll have to use will come down. Uh so can you explore this just so that our architecture will be a little more Rajashekar G: I don't know. Yeshwanth Reddy Yerraguntla: cleaner? Rajashekar G: Sure. 00:20:57 Yeshwanth Reddy Yerraguntla: Yeah. Rajashekar G: Yeah. Yeshwanth Reddy Yerraguntla: Uh if you want you can Rajashekar G: Sure. Yeshwanth Reddy Yerraguntla: connect uh in the same call and discuss this also. If you read like read for five 10 minutes and come back. Okay. Rajashekar G: And now one more thing right now we are using like uh Gemini only for everything like our internal and this Yeshwanth Reddy Yerraguntla: Right. Rajashekar G: deploymented uh and press also. So so far our usage is $124. Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: So there is a tracking who's using what how much Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: but uh that MLX model we haven't kept yet that issue and this is increasing continuously this is this Yeshwanth Reddy Yerraguntla: All right. Rajashekar G: is Yeshwanth Reddy Yerraguntla: Yeah, we should explore that MLX model already. Rajashekar G: okay Yeshwanth Reddy Yerraguntla: Like I don't have anything additional to add. Rajashekar G: but phone number. Yeshwanth Reddy Yerraguntla: You guys can also try llama CPP. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: Uh that is also supposed to be extremely cost CP Rajashekar G: Okay. 00:22:20 Yeshwanth Reddy Yerraguntla: Mac OS. Uh yeah with identic closed. So this is this is done actually completed. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: So uh try to see explore this also. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: Yeah, this is also fast. If this also is not working, uh we'll think of something. Rajashekar G: Sure. Yeshwanth Reddy Yerraguntla: Uh let me know if this is also not working. Let me know. And how fast was that? MLX this Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: thing. Okay. Rajashekar G: Common system Yeshwanth Reddy Yerraguntla: Yeah, I know. I know. Rajashekar G: is Yeshwanth Reddy Yerraguntla: You can bring this up with uh Joshua then that we are burning a lot of tokens. If you can give us internal deployment uh that will save a lot of money. Rajashekar G: okay. Yeshwanth Reddy Yerraguntla: Yeah. Rajashekar G: Shut up. Thank you. Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: Yeah. Vara Kumar Jagarapu: Yes. Rajashekar G: anything from. Vara Kumar Jagarapu: Nothing. No. Rajashekar G: Okay. Vara Kumar Jagarapu: Thank you. Bye. Yeshwanth Reddy Yerraguntla: Is Vara Kumar Jagarapu: Yeah. Yeah. Rajashekar G: Okay. Vara Kumar Jagarapu: Sure. Rajashekar G: Hello Satyasri Prabhakar Mantripragada: meeting. Rajashekar G: sir. Satyasri Prabhakar Mantripragada: Okay, Rajashekar G: Designs are coming. New design. Satyasri Prabhakar Mantripragada: whatever that we have done so far. This is what it is. Yes, we can Rajashekar G: Okay. Satyasri Prabhakar Mantripragada: client Rajashekar G: Sure. Sure. Yeah. Satyasri Prabhakar Mantripragada: power should continue on that Rajashekar G: Uh Satyasri Prabhakar Mantripragada: database Rajashekar G: changes. Satyasri Prabhakar Mantripragada: feedback button. Rajashekar G: Okay. Sir, Satyasri Prabhakar Mantripragada: Okay. Rajashekar G: sure. Satyasri Prabhakar Mantripragada: Enter. Rajashekar G: Yeah. Satyasri Prabhakar Mantripragada: Bye. Rajashekar G: Bye. Transcription ended after 00:25:27 This editable transcript was computer generated and might contain errors. People can also change the text after it was created.