Meeting Title: AI-Default-ABC Standup Date: 2025-11-19 Meeting participants: Casie Aviles, Samuel Roberts, Mustafa Raja, Rico Rejoso, Gabriel Lam, Uttam Kumaran
WEBVTT
1 00:00:20.330 ⇒ 00:00:21.460 Samuel Roberts: Hello.
2 00:00:21.670 ⇒ 00:00:22.980 Mustafa Raja: Hey, how are you?
3 00:00:23.930 ⇒ 00:00:25.780 Samuel Roberts: Doing alright. How about yourself?
4 00:00:26.370 ⇒ 00:00:27.499 Mustafa Raja: Yeah, doing good.
5 00:00:28.690 ⇒ 00:00:29.500 Samuel Roberts: Good.
6 00:00:40.990 ⇒ 00:00:42.890 Mustafa Raja: It’s pretty cold here now.
7 00:00:43.630 ⇒ 00:00:44.979 Mustafa Raja: wearing a jacket.
8 00:00:45.320 ⇒ 00:00:45.910 Mustafa Raja: Today.
9 00:00:45.910 ⇒ 00:00:46.480 Samuel Roberts: Oh.
10 00:00:47.470 ⇒ 00:00:50.910 Mustafa Raja: Couldn’t… couldn’t sit without anything warmer, so…
11 00:00:50.910 ⇒ 00:00:51.860 Samuel Roberts: Yeah.
12 00:00:54.070 ⇒ 00:00:56.129 Samuel Roberts: Yeah, it’s been getting cold here, too.
13 00:00:57.060 ⇒ 00:00:57.690 Mustafa Raja: Nice.
14 00:00:58.390 ⇒ 00:01:00.210 Mustafa Raja: Did you like this, or someone’s…
15 00:01:02.210 ⇒ 00:01:03.120 Samuel Roberts: Sorry?
16 00:01:03.520 ⇒ 00:01:06.470 Mustafa Raja: Do you like this weather, or do you like summers?
17 00:01:06.950 ⇒ 00:01:13.009 Samuel Roberts: Oh, yeah, no, I like all the different seasons, to be honest. Like, I like going through the whole year.
18 00:01:13.010 ⇒ 00:01:13.880 Mustafa Raja: Oh, nice.
19 00:01:14.210 ⇒ 00:01:17.380 Samuel Roberts: Yeah, I don’t mind it. We get them all here.
20 00:01:18.090 ⇒ 00:01:20.630 Samuel Roberts: Everything from heat to snow, so…
21 00:01:21.260 ⇒ 00:01:24.080 Mustafa Raja: We don’t get snow, but it gets pretty cold.
22 00:01:24.660 ⇒ 00:01:25.670 Samuel Roberts: Yeah…
23 00:01:28.370 ⇒ 00:01:29.070 Uttam Kumaran: Yes.
24 00:01:30.360 ⇒ 00:01:31.190 Samuel Roberts: Morning.
25 00:01:32.620 ⇒ 00:01:33.540 Uttam Kumaran: Morning.
26 00:01:33.970 ⇒ 00:01:39.160 Uttam Kumaran: Cool, maybe let’s talk about, AI stuff…
27 00:01:39.960 ⇒ 00:01:42.960 Uttam Kumaran: First, I think, Gabe, I have not…
28 00:01:43.680 ⇒ 00:01:48.470 Uttam Kumaran: given this any attention, so let’s do that in this meeting.
29 00:01:49.210 ⇒ 00:01:57.570 Uttam Kumaran: it’s been positive, though. The CTA stuff that I missed this meeting for yesterday was really.
30 00:01:57.570 ⇒ 00:01:58.150 Samuel Roberts: Yeah.
31 00:01:58.150 ⇒ 00:02:01.890 Uttam Kumaran: So… You know, it happens,
32 00:02:04.030 ⇒ 00:02:08.289 Uttam Kumaran: Okay, so… maybe, Gabe, what we can do is…
33 00:02:08.740 ⇒ 00:02:16.260 Uttam Kumaran: Should we just, you think a good use of time for, like, 5-10 minutes is just for all of us to review…
34 00:02:17.480 ⇒ 00:02:19.990 Uttam Kumaran: the PRD and leave comments, or what do you think?
35 00:02:20.870 ⇒ 00:02:29.019 Gabriel Lam: I think… I mean, that’s fine. I think maybe there’s a different question that I have, and maybe this is better to have an open discussion.
36 00:02:29.160 ⇒ 00:02:34.370 Gabriel Lam: Which is, I… Since there’s two parts of this…
37 00:02:35.330 ⇒ 00:02:39.319 Gabriel Lam: PRD, really, the first part being
38 00:02:39.450 ⇒ 00:02:48.230 Gabriel Lam: the sort of prompt refinement, and the second part being the sort of user flow, and I wanted to talk about the whole master migration.
39 00:02:48.420 ⇒ 00:02:54.529 Gabriel Lam: Because that seems to be upwalker, and I wanted to get that out of the way before maybe we…
40 00:02:55.560 ⇒ 00:02:57.969 Gabriel Lam: interrogate the PRD,
41 00:02:57.970 ⇒ 00:02:58.600 Uttam Kumaran: Okay.
42 00:02:59.840 ⇒ 00:03:05.580 Gabriel Lam: And so maybe, Sam or Mustafa, you guys have a better idea of whether… how feasible it is, how long it would take.
43 00:03:09.490 ⇒ 00:03:14.000 Samuel Roberts: Yeah… Alright, so for the… Go ahead.
44 00:03:14.320 ⇒ 00:03:20.870 Mustafa Raja: Yeah, I think the only thing that, my hinders would be, would be getting…
45 00:03:21.000 ⇒ 00:03:30.520 Mustafa Raja: linear data in, but I think, we are already… we already have, have that source ready in… in our platform, so…
46 00:03:30.700 ⇒ 00:03:34.800 Mustafa Raja: the work… transferring the workflow, I don’t think that should be…
47 00:03:35.650 ⇒ 00:03:41.900 Mustafa Raja: That should, that should, be such, big of a task.
48 00:03:45.330 ⇒ 00:03:51.849 Uttam Kumaran: Okay, yeah, but that’s… but I think I would… what I would like to see is some organization, like, what… what needs to get migrated?
49 00:03:52.280 ⇒ 00:03:54.570 Uttam Kumaran: Like, what is the plan?
50 00:03:54.840 ⇒ 00:03:58.979 Uttam Kumaran: Like, you know what I’m gonna say, not big of a task doesn’t really mean anything.
51 00:03:58.980 ⇒ 00:03:59.580 Mustafa Raja: Yeah.
52 00:03:59.760 ⇒ 00:04:00.350 Mustafa Raja: Yeah.
53 00:04:00.350 ⇒ 00:04:09.379 Uttam Kumaran: So… can… I guess, is the monster migration just moving? Like, that just seems like there’s a lot in there to unpack, so…
54 00:04:09.860 ⇒ 00:04:12.049 Samuel Roberts: For the linear ticket generator?
55 00:04:12.550 ⇒ 00:04:13.290 Mustafa Raja: Yeah.
56 00:04:14.610 ⇒ 00:04:15.120 Samuel Roberts: I think it’s just.
57 00:04:15.120 ⇒ 00:04:15.580 Mustafa Raja: That one’s.
58 00:04:15.580 ⇒ 00:04:16.820 Samuel Roberts: low, right?
59 00:04:16.829 ⇒ 00:04:17.459 Mustafa Raja: Yeah.
60 00:04:19.930 ⇒ 00:04:31.630 Uttam Kumaran: So I guess, like, you heard what Gabe asked for, right? Which, Gabe was like, hey, there’s this thing that’s unblocking. It’s clear Gabe isn’t familiar with, sort of, what entails there, so I guess I’m expecting you guys to…
61 00:04:31.730 ⇒ 00:04:35.059 Uttam Kumaran: To answer him on, like, okay, what… what…
62 00:04:35.160 ⇒ 00:04:38.969 Uttam Kumaran: what does need to move to Maestra, and how long is that gonna take, and does that block…
63 00:04:39.410 ⇒ 00:04:42.539 Uttam Kumaran: This task, and then can we get that done this week, you know?
64 00:04:43.120 ⇒ 00:04:49.879 Mustafa Raja: Yeah, so we would need these two, APIs, in the platform.
65 00:04:50.090 ⇒ 00:04:54.120 Mustafa Raja: And really what they are getting is,
66 00:04:54.550 ⇒ 00:04:59.180 Mustafa Raja: Some source data, from the webhook itself, and then…
67 00:04:59.340 ⇒ 00:05:07.600 Mustafa Raja: some data from linear. So, we have that… we have the linear connector already in the platform.
68 00:05:10.690 ⇒ 00:05:18.360 Mustafa Raja: So, yeah, this should be… this should be easily doable this week. Only thing that the platform wouldn’t have is this part.
69 00:05:18.610 ⇒ 00:05:22.039 Mustafa Raja: That is, client hubs.
70 00:05:22.530 ⇒ 00:05:24.379 Mustafa Raja: Client hub integration.
71 00:05:24.590 ⇒ 00:05:30.199 Casie Aviles: Yeah, it would be… more, like, that’ll take some more time, I think.
72 00:05:30.200 ⇒ 00:05:32.730 Mustafa Raja: Yeah, this part… this part will take time.
73 00:05:33.220 ⇒ 00:05:38.259 Samuel Roberts: But, okay, so could we do the prompt refinements in here, then, for now? And that would unblock.
74 00:05:38.260 ⇒ 00:05:39.180 Mustafa Raja: Like…
75 00:05:40.950 ⇒ 00:05:45.839 Samuel Roberts: But that might not allow is the kind of UX UI stuff, though.
76 00:05:47.400 ⇒ 00:05:48.190 Mustafa Raja: Yeah.
77 00:05:48.900 ⇒ 00:05:59.249 Mustafa Raja: There’s one other problem, since now, stand-ups are… stand-ups are consolidated, this agent takes more than 30 seconds to generate.
78 00:05:59.250 ⇒ 00:06:00.010 Samuel Roberts: Oh, right.
79 00:06:00.010 ⇒ 00:06:06.180 Mustafa Raja: But, NN does now support streaming responses, so… Oh.
80 00:06:06.180 ⇒ 00:06:07.000 Samuel Roberts: Oh, great.
81 00:06:07.000 ⇒ 00:06:13.630 Mustafa Raja: We’re gonna stream to, make the connection open for longer periods.
82 00:06:15.000 ⇒ 00:06:16.869 Mustafa Raja: So this is a good solution for that.
83 00:06:17.080 ⇒ 00:06:20.830 Mustafa Raja: If you want to, for now, if we want to stay in India, yeah.
84 00:06:23.860 ⇒ 00:06:27.559 Uttam Kumaran: Okay, but I’m still, I’m still, like, not really understanding anything.
85 00:06:28.060 ⇒ 00:06:34.189 Samuel Roberts: I don’t think it needs to go to Monster right now. If we can stream it out, then that’s fine. I think that avoids having to…
86 00:06:34.680 ⇒ 00:06:40.020 Samuel Roberts: take the time to do that right now, if we want to just get this in a better state with this PDR.
87 00:06:40.880 ⇒ 00:06:47.040 Samuel Roberts: We can change the prompts here. If we can stream, that means that it shouldn’t hang up anymore.
88 00:06:48.910 ⇒ 00:06:51.519 Samuel Roberts: And then, depending on what the,
89 00:06:51.700 ⇒ 00:06:56.599 Samuel Roberts: UI, UX kind of stuff, and the flows there, I think it should be okay.
90 00:06:56.870 ⇒ 00:06:57.720 Mustafa Raja: Yeah.
91 00:06:57.720 ⇒ 00:07:01.279 Samuel Roberts: So that… that… it might not be as much of a blocker as I initially thought.
92 00:07:02.080 ⇒ 00:07:02.680 Gabriel Lam: Okay.
93 00:07:06.250 ⇒ 00:07:15.430 Gabriel Lam: Yeah, because it seems to me that a lot of what we were asking or were thinking about in terms of, like, getting context already exists. It really is just the output.
94 00:07:15.430 ⇒ 00:07:16.390 Samuel Roberts: is not…
95 00:07:16.570 ⇒ 00:07:22.389 Gabriel Lam: up to standard based on what information needs to come across, right? Like, if we look at
96 00:07:22.750 ⇒ 00:07:26.619 Gabriel Lam: You know, client hubs, or even for individual
97 00:07:27.020 ⇒ 00:07:29.730 Gabriel Lam: Videos when they’re recorded and we have transcripts.
98 00:07:30.240 ⇒ 00:07:37.590 Gabriel Lam: Like, technically we already have all the context there, it’s just whatever output it gives us is…
99 00:07:38.940 ⇒ 00:07:46.270 Gabriel Lam: Doesn’t communicate what it really needs to communicate, which is, like, what are the next steps, what are the action items, and what are the right tickets?
100 00:07:46.670 ⇒ 00:07:48.030 Gabriel Lam: I’m feeling that.
101 00:07:50.420 ⇒ 00:07:52.530 Gabriel Lam: The ideal is to, sort of.
102 00:07:53.100 ⇒ 00:07:58.339 Gabriel Lam: minimize the additional mental load of, like, oh, I need to navigate to this place. And…
103 00:07:58.340 ⇒ 00:07:58.970 Samuel Roberts: Yeah.
104 00:07:59.850 ⇒ 00:08:02.100 Gabriel Lam: If we want to do an exercise into…
105 00:08:02.830 ⇒ 00:08:06.419 Gabriel Lam: Updating the client hub to have, you know, a running
106 00:08:08.490 ⇒ 00:08:18.719 Gabriel Lam: summary, or running a sort of live set of action items, or a live summary updated based on every new video. I remember on Monday, we were talking about whether that
107 00:08:18.880 ⇒ 00:08:20.690 Gabriel Lam: The whole exercise would be.
108 00:08:21.440 ⇒ 00:08:23.290 Gabriel Lam: Feasible, or make sense for this week.
109 00:08:23.490 ⇒ 00:08:27.200 Gabriel Lam: So I think for now, I was just thinking, that…
110 00:08:28.610 ⇒ 00:08:34.429 Gabriel Lam: to really refine the summary and the linear ticket generation. And I saw, Mustafa, you had also…
111 00:08:34.630 ⇒ 00:08:41.009 Gabriel Lam: debugged some of the… the ability to generate tickets after each video? Is that correct?
112 00:08:43.490 ⇒ 00:08:46.049 Mustafa Raja: After each, after each stand-up.
113 00:08:46.660 ⇒ 00:08:48.229 Gabriel Lam: Yeah, there were some stand-ups where…
114 00:08:48.230 ⇒ 00:08:49.959 Mustafa Raja: No, no, no, I haven’t looked into that.
115 00:08:49.960 ⇒ 00:08:50.720 Gabriel Lam: Forge.
116 00:08:51.700 ⇒ 00:08:52.290 Gabriel Lam: Okay.
117 00:08:52.290 ⇒ 00:08:56.329 Mustafa Raja: Yeah, for that, I haven’t looked into it. I looked into this issue that was…
118 00:08:56.920 ⇒ 00:09:03.330 Mustafa Raja: That was, this was happening because, it’s just taking longer, longer.
119 00:09:03.330 ⇒ 00:09:06.340 Samuel Roberts: Just this, yeah, we make transcripts now, right? Okay.
120 00:09:06.340 ⇒ 00:09:07.040 Mustafa Raja: Yeah.
121 00:09:08.140 ⇒ 00:09:19.679 Samuel Roberts: Yeah, I would say let’s forget about migrating this then for now. That’s an additional piece we don’t need to worry about right now if this is working, and we want to just update… mostly just updating prompts, and then getting…
122 00:09:19.980 ⇒ 00:09:24.360 Samuel Roberts: The streaming responses to work, so that it… does return.
123 00:09:24.760 ⇒ 00:09:25.370 Mustafa Raja: Yeah.
124 00:09:26.610 ⇒ 00:09:30.900 Uttam Kumaran: I mean, what is the lift on moving all of this to… code?
125 00:09:32.070 ⇒ 00:09:41.330 Samuel Roberts: My initial thought was that it wouldn’t be a ton, because it’s just a couple agents, but I didn’t… I wasn’t thinking about the client hub stuff, so what… what we’d have to figure out…
126 00:09:41.680 ⇒ 00:09:43.430 Samuel Roberts: To understand that is.
127 00:09:43.680 ⇒ 00:09:47.630 Samuel Roberts: do we need to get all the Client Hub stuff as well migrated, or can we…
128 00:09:47.630 ⇒ 00:09:48.010 Uttam Kumaran: Okay.
129 00:09:48.010 ⇒ 00:09:52.319 Samuel Roberts: We might be able to hook into these through N8N if we set it up right, but that’s still…
130 00:09:52.740 ⇒ 00:09:54.810 Samuel Roberts: Not a… not a great solution.
131 00:09:55.260 ⇒ 00:10:00.599 Mustafa Raja: We can still communicate these as tools with, webhooks.
132 00:10:00.600 ⇒ 00:10:01.370 Casie Aviles: Yeah.
133 00:10:01.470 ⇒ 00:10:06.139 Samuel Roberts: How slow are those now, though? Is that gonna be a bunch of stuff to worry about anyway?
134 00:10:10.240 ⇒ 00:10:12.879 Uttam Kumaran: Sorry, I didn’t get… Like, how slow is it?
135 00:10:13.220 ⇒ 00:10:13.680 Uttam Kumaran: How’s the response?
136 00:10:13.680 ⇒ 00:10:17.860 Samuel Roberts: What is… what is, yeah, what is taking the time here for that whole interview?
137 00:10:17.860 ⇒ 00:10:21.030 Mustafa Raja: Oh, hmm… slip.
138 00:10:21.030 ⇒ 00:10:22.719 Samuel Roberts: Client Hub’s doing more, they’re…
139 00:10:22.720 ⇒ 00:10:25.199 Mustafa Raja: Yeah, yeah, the client hubs are taking time.
140 00:10:25.310 ⇒ 00:10:27.099 Mustafa Raja: In generating the response.
141 00:10:28.130 ⇒ 00:10:35.359 Samuel Roberts: Yeah, so I think… We’d still have to do the work around doing the streaming for that,
142 00:10:36.510 ⇒ 00:10:42.100 Samuel Roberts: I mean, the other side of it is getting these into these generic client hubs that just take a client as an input.
143 00:10:42.340 ⇒ 00:10:50.389 Samuel Roberts: And then we don’t, you know… but that’s a… that’s… that’s a bigger lift, so that’s why I didn’t want to add that to this… this PRD at this point.
144 00:10:50.810 ⇒ 00:11:01.550 Samuel Roberts: But… I mean, we could, I suppose, move, you know, the vast majority of, like, everything here… Probably…
145 00:11:02.750 ⇒ 00:11:08.139 Samuel Roberts: to Mastra, and then just do this as a separate, like, N8N webhook?
146 00:11:08.770 ⇒ 00:11:09.639 Samuel Roberts: For now?
147 00:11:12.510 ⇒ 00:11:14.489 Samuel Roberts: Does that make sense, Mustafa?
148 00:11:14.870 ⇒ 00:11:15.650 Mustafa Raja: Yeah.
149 00:11:19.030 ⇒ 00:11:22.960 Samuel Roberts: Yeah, so I think… If that’s what we’re interested in…
150 00:11:23.680 ⇒ 00:11:27.640 Uttam Kumaran: For the product objective itself, Is it, like…
151 00:11:27.890 ⇒ 00:11:34.699 Uttam Kumaran: I guess I’m still… are there… I guess that’s why… as part of the PRD game, are there, like, UI changes?
152 00:11:35.550 ⇒ 00:11:36.100 Uttam Kumaran: To me?
153 00:11:36.100 ⇒ 00:11:36.520 Gabriel Lam: Alright.
154 00:11:36.520 ⇒ 00:11:37.470 Uttam Kumaran: Or, not really.
155 00:11:37.470 ⇒ 00:11:42.899 Gabriel Lam: I initially thought of UI changes, I think, at this point, it makes…
156 00:11:43.320 ⇒ 00:11:47.450 Gabriel Lam: more sense to have UI changes in the client hub? Hmm…
157 00:11:48.020 ⇒ 00:11:48.670 Uttam Kumaran: Okay.
158 00:11:49.560 ⇒ 00:11:57.120 Gabriel Lam: Because people are already used to going through to the client hub and then doing that. I’m just trying to minimize the number of, like, oh, here’s a new…
159 00:11:57.120 ⇒ 00:11:58.409 Samuel Roberts: Oh, yeah, yeah.
160 00:11:58.410 ⇒ 00:12:03.450 Gabriel Lam: You have to go to, here’s a new thing, you have to go to, and, like, what is the existing flow?
161 00:12:03.790 ⇒ 00:12:05.920 Gabriel Lam: And I think I just noticed that, like, people are just…
162 00:12:06.260 ⇒ 00:12:11.770 Gabriel Lam: You know, either going to the dashboard and filtering by your project, or just going to the client help and looking through.
163 00:12:12.480 ⇒ 00:12:15.289 Gabriel Lam: past videos, and I think that ends up
164 00:12:15.940 ⇒ 00:12:17.819 Gabriel Lam: If… if the whole client hub
165 00:12:18.080 ⇒ 00:12:20.940 Gabriel Lam: Update is a pretty heavy lift, then…
166 00:12:21.680 ⇒ 00:12:25.709 Gabriel Lam: I think getting the prompts out is… the next best thing.
167 00:12:26.160 ⇒ 00:12:28.870 Gabriel Lam: And my ideal vision for this is…
168 00:12:29.060 ⇒ 00:12:33.650 Gabriel Lam: You know, you click into each client, and you’re able to have a live
169 00:12:33.900 ⇒ 00:12:36.269 Gabriel Lam: Almost like a sort of stand-up.
170 00:12:36.270 ⇒ 00:12:37.329 Samuel Roberts: I see, yes, yes.
171 00:12:37.330 ⇒ 00:12:38.529 Gabriel Lam: Assistant kind of view.
172 00:12:38.530 ⇒ 00:12:39.300 Samuel Roberts: dogs.
173 00:12:39.800 ⇒ 00:12:43.269 Gabriel Lam: you know, I think Utam had sent some examples of…
174 00:12:46.020 ⇒ 00:12:51.690 Gabriel Lam: like, a summary of what we need to do for the week, based on the calls, and sometimes stand-ups don’t capture that.
175 00:12:54.730 ⇒ 00:13:05.890 Samuel Roberts: Yeah, I think that’s… that actually sounds like a… a decent PDR in and of itself for the client hubs. That would also, you know, mean moving the client hubs into Monster, probably.
176 00:13:05.990 ⇒ 00:13:12.249 Samuel Roberts: Anyway, so I think I… yeah, that actually makes a ton of sense for the linear ticket, like.
177 00:13:12.370 ⇒ 00:13:20.219 Samuel Roberts: improvement. I think all we have to do is work on those prompts at this point, and make sure that it streams back the responses so that it doesn’t time out.
178 00:13:20.870 ⇒ 00:13:24.000 Samuel Roberts: And then it’s just refining those prompts to get it to be…
179 00:13:25.370 ⇒ 00:13:30.990 Uttam Kumaran: But the prompts are not in, like, LangFuse at this moment, right? So is that part…
180 00:13:30.990 ⇒ 00:13:32.250 Samuel Roberts: Right here, I think.
181 00:13:32.250 ⇒ 00:13:33.139 Uttam Kumaran: Oh, okay.
182 00:13:33.240 ⇒ 00:13:34.120 Uttam Kumaran: So…
183 00:13:34.310 ⇒ 00:13:40.189 Uttam Kumaran: I guess my question is more of, like, Gabe, are you able, at this moment, to, like, go edit prompts and see the outputs, and, like.
184 00:13:40.610 ⇒ 00:13:43.769 Uttam Kumaran: refine those? Or is, like, that still an issue?
185 00:13:44.250 ⇒ 00:13:48.090 Gabriel Lam: That’s an issue that I would like to get fixed today, because I… And…
186 00:13:48.090 ⇒ 00:13:48.860 Samuel Roberts: Okay, do you…
187 00:13:48.860 ⇒ 00:13:50.940 Gabriel Lam: I love to learn that time.
188 00:13:51.270 ⇒ 00:13:52.869 Gabriel Lam: Not yet, which is…
189 00:13:52.870 ⇒ 00:13:53.820 Samuel Roberts: Okay, I’ll go.
190 00:13:53.820 ⇒ 00:13:58.399 Gabriel Lam: grab time with one of you guys after, and just, like, I would appreciate a quick deep dive.
191 00:14:00.020 ⇒ 00:14:10.910 Uttam Kumaran: So I think one thing to solve for everybody here is we, again, we want as many people to be able to start to edit and refine and test new prompts. That’s not something that, like.
192 00:14:11.180 ⇒ 00:14:15.099 Uttam Kumaran: our AI team should… we should get the first version out, and then hand it to the user.
193 00:14:15.100 ⇒ 00:14:15.430 Samuel Roberts: Yeah.
194 00:14:15.430 ⇒ 00:14:18.759 Uttam Kumaran: Because everybody in this company is smart enough to edit prompts.
195 00:14:19.200 ⇒ 00:14:23.560 Uttam Kumaran: And then, so, I think, like, let’s see how far we can get.
196 00:14:24.010 ⇒ 00:14:25.950 Uttam Kumaran: With that, and then…
197 00:14:26.810 ⇒ 00:14:34.349 Uttam Kumaran: if we need to bite off a larger migration, then… then we… let’s… we should… we could bite it off. I… I don’t mind.
198 00:14:34.600 ⇒ 00:14:39.780 Uttam Kumaran: Because I’m… yeah, I mean, this is clearly blocking, like, almost every step of the way, so…
199 00:14:40.850 ⇒ 00:14:47.029 Uttam Kumaran: we can instead do a PR… we can instead do, like, a technical design around migrating a bunch of stuff
200 00:14:48.030 ⇒ 00:14:50.960 Uttam Kumaran: And that could be the sprint, you know, and if we need to do that…
201 00:14:50.960 ⇒ 00:14:51.560 Samuel Roberts: Yeah.
202 00:14:51.560 ⇒ 00:14:52.160 Uttam Kumaran: stuff.
203 00:14:53.040 ⇒ 00:14:55.280 Samuel Roberts: Yeah, I think that would be a good idea.
204 00:14:56.240 ⇒ 00:14:56.950 Samuel Roberts: Getting all this.
205 00:14:56.950 ⇒ 00:14:58.419 Gabriel Lam: My last question is…
206 00:14:58.420 ⇒ 00:14:59.580 Samuel Roberts: Great, yeah.
207 00:14:59.950 ⇒ 00:15:06.599 Gabriel Lam: Do you… How… I’m thinking that the linear ticket
208 00:15:06.970 ⇒ 00:15:14.690 Gabriel Lam: Would, be able to sort of semantically look for… look for semantically similar tickets and be able to update them.
209 00:15:15.910 ⇒ 00:15:17.959 Gabriel Lam: Do you guys have any idea what that…
210 00:15:18.430 ⇒ 00:15:20.680 Gabriel Lam: That would fit into here. Is that something that…
211 00:15:20.880 ⇒ 00:15:23.920 Gabriel Lam: Is that a feature that we’re able to… put in.
212 00:15:26.030 ⇒ 00:15:28.669 Samuel Roberts: Yeah, go ahead, Casey.
213 00:15:29.600 ⇒ 00:15:36.700 Casie Aviles: Yeah, I think that’s… I mean, for me, I think that’s going to be a… A little shaky, but…
214 00:15:36.920 ⇒ 00:15:44.180 Casie Aviles: Because we’ll have to, like… For example, like, in the meeting, since we have merged meetings,
215 00:15:44.540 ⇒ 00:15:50.200 Casie Aviles: We’re gonna have to look through… multiple… linear, teams.
216 00:15:50.950 ⇒ 00:15:55.720 Casie Aviles: And… We’re gonna have to match…
217 00:15:56.200 ⇒ 00:16:00.249 Casie Aviles: So, I think, I mean, it’s going to be rough at first, like, the output.
218 00:16:01.110 ⇒ 00:16:07.460 Casie Aviles: We’re gonna have to, like, match, like, the transcript and the tickets, and there can be a lot of tickets, but…
219 00:16:08.180 ⇒ 00:16:12.329 Casie Aviles: Okay. I’m just thinking of an LLM step that will probably do that, but…
220 00:16:13.040 ⇒ 00:16:15.990 Casie Aviles: I mean, yeah, I think that’s… that’s…
221 00:16:16.900 ⇒ 00:16:18.789 Casie Aviles: what I have in mind for that.
222 00:16:19.780 ⇒ 00:16:30.400 Samuel Roberts: Yeah, I think Casey’s right. I think right now, this is… this piece is… is just generating new tickets from transcripts, really. So it… to add that, kind of, pulling in more stuff.
223 00:16:30.560 ⇒ 00:16:35.219 Samuel Roberts: From linear as well, adding that to the context to know what to update is…
224 00:16:35.520 ⇒ 00:16:43.000 Samuel Roberts: is a bigger lift here than just the creation, but I think once we… if we take, like, that week to get stuff
225 00:16:43.180 ⇒ 00:16:47.280 Samuel Roberts: You know, hardened a little bit, then that opens up, maybe.
226 00:16:48.390 ⇒ 00:16:53.940 Samuel Roberts: you know, adding that as a feature in a way that’s more robust in Mastra instead of…
227 00:16:54.260 ⇒ 00:16:59.130 Samuel Roberts: Here in N8in, and trying to build that into this flow that we’re gonna move anyway.
228 00:17:05.069 ⇒ 00:17:10.219 Uttam Kumaran: Okay, I’m still not hearing, like, a defined yes or no in what direction. So, for me, when I.
229 00:17:10.220 ⇒ 00:17:12.550 Samuel Roberts: My thought is no for now, yeah, no.
230 00:17:12.550 ⇒ 00:17:17.479 Uttam Kumaran: Yeah, for me, when I see this anything, I’m gonna throw… I’m gonna throw up, like, because there’s no way…
231 00:17:17.640 ⇒ 00:17:20.040 Uttam Kumaran: Like, this is just way too complicated.
232 00:17:20.140 ⇒ 00:17:23.310 Samuel Roberts: Yeah. And it’s not, like, a reliable system, but…
233 00:17:23.430 ⇒ 00:17:28.709 Uttam Kumaran: I kind of just… I think one is that if we can get Gabe into Lank Fuse, and he can…
234 00:17:28.710 ⇒ 00:17:29.280 Samuel Roberts: Yep.
235 00:17:29.280 ⇒ 00:17:43.789 Uttam Kumaran: start data prompts, and he’s comfortable there, like, that would be one thing. Second piece here is, I would love, maybe, Gabe, one for, like, one of the next weeks. We do a more robust migration of, like, several of the core
236 00:17:43.910 ⇒ 00:17:50.280 Uttam Kumaran: applications from N8N into… The codebase, and, like, that itself is a project.
237 00:17:50.690 ⇒ 00:17:51.220 Gabriel Lam: Yeah.
238 00:17:51.220 ⇒ 00:17:55.240 Uttam Kumaran: Because I don’t think we can do a lot of these things very easily, like…
239 00:17:55.400 ⇒ 00:18:00.609 Uttam Kumaran: I think team will figure it out, but it’s gonna just… just gonna make this way messier.
240 00:18:01.090 ⇒ 00:18:01.600 Uttam Kumaran: Yeah.
241 00:18:01.600 ⇒ 00:18:02.280 Samuel Roberts: Agreed.
242 00:18:02.510 ⇒ 00:18:03.220 Uttam Kumaran: you know.
243 00:18:07.810 ⇒ 00:18:08.430 Gabriel Lam: Okay.
244 00:18:09.760 ⇒ 00:18:13.219 Samuel Roberts: Yeah, so we’ll get you, we’ll get you into LangFuse today.
245 00:18:13.220 ⇒ 00:18:13.950 Gabriel Lam: Yeah.
246 00:18:13.950 ⇒ 00:18:16.369 Samuel Roberts: And you can start working on those prompt slides.
247 00:18:16.370 ⇒ 00:18:17.190 Gabriel Lam: the problem.
248 00:18:17.190 ⇒ 00:18:19.049 Samuel Roberts: A few of them here, yeah.
249 00:18:21.500 ⇒ 00:18:25.849 Samuel Roberts: And see if that can get us, you know, most of the way there for this… this week.
250 00:18:26.440 ⇒ 00:18:28.780 Samuel Roberts: For the, the tickets.
251 00:18:29.000 ⇒ 00:18:32.559 Samuel Roberts: And then, yeah, prioritize a lot of that stuff in another PRD for another week.
252 00:18:32.560 ⇒ 00:18:33.140 Gabriel Lam: Okay.
253 00:18:34.190 ⇒ 00:18:35.879 Uttam Kumaran: Yeah, so I think probably…
254 00:18:36.260 ⇒ 00:18:49.349 Uttam Kumaran: Ideally, and we’re talking about a migration, we need sort of, like, okay, which of the core AND inflows? Like, what are the migration steps? And, like, that’s probably in a Google Sheet or something, but yeah, I think that’s, like…
255 00:18:49.700 ⇒ 00:18:52.990 Uttam Kumaran: Especially next week, we’re gonna have some downtime.
256 00:18:53.540 ⇒ 00:18:57.689 Uttam Kumaran: So, that would be… Great, if possible.
257 00:18:58.650 ⇒ 00:18:59.190 Gabriel Lam: Yep.
258 00:18:59.190 ⇒ 00:18:59.750 Samuel Roberts: Yeah.
259 00:18:59.920 ⇒ 00:19:12.180 Samuel Roberts: Yeah, there’s some stuff together for, like, which flows need to go into, and trying to make them more generic so that we’re not recreating them every time for new clients and stuff that I have in the client hubSOP, but we can…
260 00:19:12.360 ⇒ 00:19:14.729 Samuel Roberts: Get that into an actual plan now.
261 00:19:15.200 ⇒ 00:19:16.720 Samuel Roberts: Of how to go about it.
262 00:19:17.380 ⇒ 00:19:18.090 Uttam Kumaran: Okay, cool.
263 00:19:18.090 ⇒ 00:19:18.940 Samuel Roberts: Sounds good.
264 00:19:19.270 ⇒ 00:19:23.420 Uttam Kumaran: Give me one second, someone’s just at my door, hold on, I’ll be right back.
265 00:19:25.920 ⇒ 00:19:32.220 Uttam Kumaran: Also, maybe, Sam, you can also talk about,
266 00:19:33.580 ⇒ 00:19:36.320 Uttam Kumaran: the, like, if we want to migrate ABC stuff.
267 00:19:36.520 ⇒ 00:19:42.319 Uttam Kumaran: Maybe we could also talk through that now, if you want to own that. Like, how do we migrate ABC stuff from N8N?
268 00:19:43.790 ⇒ 00:19:45.120 Samuel Roberts: Yeah, okay.
269 00:19:45.280 ⇒ 00:19:52.110 Samuel Roberts: So that’s a conversation we’ve been having about, making Andy a little more… reliable.
270 00:19:52.470 ⇒ 00:19:56.770 Samuel Roberts: And I know that that flow… can we actually get that out? Who’s sharing right now, Mustafa?
271 00:19:57.560 ⇒ 00:20:02.020 Samuel Roberts: Are you able to pull up the, like, the anti-flow.
272 00:20:09.090 ⇒ 00:20:17.510 Samuel Roberts: Yeah. So, I’m wondering, and you guys know this better than I know it, because I’ve not been so super involved in this, and it predates me, but like…
273 00:20:17.880 ⇒ 00:20:23.710 Samuel Roberts: Where… like, there’s several different flows here, I imagine, based on what I’m seeing.
274 00:20:23.980 ⇒ 00:20:29.170 Samuel Roberts: So, like, what is the key Andy flow, for the most part, here? And, like.
275 00:20:29.650 ⇒ 00:20:31.640 Samuel Roberts: What do you think it would take
276 00:20:31.850 ⇒ 00:20:41.949 Samuel Roberts: in a very broad strokes, to, like, maybe move this into a platform, because if they’re going to want to be hosting it on their own infra and everything, too, I don’t know if NAN makes sense for all of it anymore.
277 00:20:46.280 ⇒ 00:20:54.470 Mustafa Raja: Yeah, so this, this web book is where, where we are really, really, querying Andy, Andy from.
278 00:20:55.110 ⇒ 00:20:55.440 Samuel Roberts: Okay.
279 00:20:56.270 ⇒ 00:21:07.819 Mustafa Raja: And then, this agent, what it does is, decides the department. And here, what this sheet does is.
280 00:21:08.370 ⇒ 00:21:13.519 Mustafa Raja: It has all of the emails, for all of the CSRs.
281 00:21:13.710 ⇒ 00:21:15.029 Mustafa Raja: So, Andy?
282 00:21:16.020 ⇒ 00:21:17.210 Mustafa Raja: And all of.
283 00:21:17.210 ⇒ 00:21:17.670 Samuel Roberts: Okay.
284 00:21:17.670 ⇒ 00:21:30.370 Mustafa Raja: that they are associated with, so Andy can then decide, okay, this person is from this department, there I should route the question, and then if they are from multiple ones, it’ll ask.
285 00:21:31.620 ⇒ 00:21:33.409 Samuel Roberts: Interesting, okay, okay, cool.
286 00:21:33.410 ⇒ 00:21:35.249 Mustafa Raja: Yeah, I’ll ask the department.
287 00:21:38.730 ⇒ 00:21:47.029 Mustafa Raja: And then… the quest… Yeah, and then the question just goes to, the agent, and then we have…
288 00:21:47.030 ⇒ 00:21:47.380 Samuel Roberts: Yep.
289 00:21:47.380 ⇒ 00:21:48.810 Mustafa Raja: Multiple 10 years.
290 00:21:49.280 ⇒ 00:21:50.050 Mustafa Raja: dodged.
291 00:21:51.120 ⇒ 00:21:51.999 Samuel Roberts: So this is…
292 00:21:52.000 ⇒ 00:21:52.380 Mustafa Raja: Yeah.
293 00:21:55.480 ⇒ 00:21:56.950 Mustafa Raja: These ones.
294 00:21:56.950 ⇒ 00:22:01.880 Samuel Roberts: Yeah, so this is all breaking out to, like, different… Answer different questions, effectively.
295 00:22:02.350 ⇒ 00:22:06.200 Mustafa Raja: Yes, so we have… QueryDB tool.
296 00:22:07.480 ⇒ 00:22:10.730 Mustafa Raja: I think the main thing, and then we have…
297 00:22:11.240 ⇒ 00:22:24.989 Mustafa Raja: Yeah, so for sheets, we are now relying… I guess, we are now relying properly on this, QueryDB tool. And then for the RAG, we have, two stores. One is the coverage one, and then…
298 00:22:25.100 ⇒ 00:22:27.230 Mustafa Raja: The other one is the central dock.
299 00:22:28.540 ⇒ 00:22:29.200 Samuel Roberts: Okay.
300 00:22:29.700 ⇒ 00:22:30.160 Mustafa Raja: Yeah.
301 00:22:30.160 ⇒ 00:22:35.810 Samuel Roberts: Okay, so I think what is probably a good first step here is to… like…
302 00:22:35.930 ⇒ 00:22:40.910 Samuel Roberts: Get this into some kind of, like…
303 00:22:41.400 ⇒ 00:22:45.490 Samuel Roberts: like a README documentation kind of thing.
304 00:22:45.830 ⇒ 00:22:46.590 Mustafa Raja: Yes.
305 00:22:47.200 ⇒ 00:22:55.199 Samuel Roberts: Like, kind of exactly what you were just doing right here, and so, like, walking through, like, what are the key steps that…
306 00:22:55.430 ⇒ 00:22:58.669 Samuel Roberts: Are happening? What are the key tools that are being called?
307 00:22:58.830 ⇒ 00:23:09.729 Samuel Roberts: Data sources, because… If there’s things like, you know, if we have… Like, this spreadsheet, for example.
308 00:23:11.410 ⇒ 00:23:22.970 Samuel Roberts: we, like, we can still make use of that, you know, because that’s on their infra, right? And then, like, the DB tool, like, this is… could be a master agent, you know? So I think my thought is to, like.
309 00:23:23.710 ⇒ 00:23:30.189 Samuel Roberts: basically… walk through this in a way… like, obviously, like.
310 00:23:30.840 ⇒ 00:23:33.479 Samuel Roberts: You know, what matters here is, like, that data…
311 00:23:33.780 ⇒ 00:23:47.299 Samuel Roberts: the agents, you know, little bits of the code, or not little bits of the code, but, like, the, the flow is, you know, the stuff that can be put into just regular code, but the actual agents, the responses, the prompts.
312 00:23:48.790 ⇒ 00:23:55.459 Samuel Roberts: I think that’s a… you know, we wanna make sure we don’t miss anything, I guess, if we’re gonna do a big, you know, move, but…
313 00:23:56.490 ⇒ 00:24:01.160 Samuel Roberts: it’s hard for me to even parse some of this right now when I’m looking at it, because I’m not as familiar with it.
314 00:24:01.400 ⇒ 00:24:02.460 Mustafa Raja: Yeah. So…
315 00:24:02.830 ⇒ 00:24:06.400 Samuel Roberts: Are there things here that are extraneous now, or is this all…
316 00:24:07.100 ⇒ 00:24:12.620 Mustafa Raja: Still, like… I guess this… this we wouldn’t be using now, because.
317 00:24:12.620 ⇒ 00:24:13.070 Samuel Roberts: Okay.
318 00:24:13.070 ⇒ 00:24:15.420 Mustafa Raja: of DBQuery to… query DB2.
319 00:24:15.420 ⇒ 00:24:16.130 Samuel Roberts: Got it, okay.
320 00:24:16.130 ⇒ 00:24:17.520 Mustafa Raja: me if I’m wrong, Casey.
321 00:24:17.520 ⇒ 00:24:19.020 Casie Aviles: Yeah, yeah, that’s some of the…
322 00:24:19.020 ⇒ 00:24:19.570 Mustafa Raja: similar.
323 00:24:19.570 ⇒ 00:24:19.960 Casie Aviles: things here.
324 00:24:19.960 ⇒ 00:24:21.210 Mustafa Raja: Similar with this.
325 00:24:22.350 ⇒ 00:24:22.840 Casie Aviles: Yeah.
326 00:24:22.840 ⇒ 00:24:29.490 Mustafa Raja: these are deactivated, so I guess we are disregarding these ones also.
327 00:24:30.310 ⇒ 00:24:31.339 Mustafa Raja: Cool, okay.
328 00:24:31.340 ⇒ 00:24:32.050 Samuel Roberts: Okay.
329 00:24:32.350 ⇒ 00:24:33.759 Mustafa Raja: Yeah, similar… So it’s not quite as.
330 00:24:33.760 ⇒ 00:24:34.610 Samuel Roberts: As crazy as it.
331 00:24:34.610 ⇒ 00:24:36.050 Mustafa Raja: These ones also.
332 00:24:36.350 ⇒ 00:24:37.490 Samuel Roberts: Okay. Huh.
333 00:24:38.270 ⇒ 00:24:40.819 Mustafa Raja: Yeah, so I think, either…
334 00:24:41.300 ⇒ 00:24:42.289 Samuel Roberts: Yeah, go ahead.
335 00:24:42.720 ⇒ 00:24:46.550 Mustafa Raja: Yeah, and then what this is, let me…
336 00:24:49.770 ⇒ 00:24:50.410 Mustafa Raja: Oh.
337 00:24:50.410 ⇒ 00:24:50.910 Casie Aviles: cleanup.
338 00:24:50.910 ⇒ 00:24:55.729 Mustafa Raja: I believe we… Cedric in, please?
339 00:24:59.080 ⇒ 00:24:59.800 Samuel Roberts: Sorry?
340 00:25:00.590 ⇒ 00:25:02.480 Mustafa Raja: Yeah, can see your shining something?
341 00:25:02.750 ⇒ 00:25:08.970 Casie Aviles: Oh, no, no, I was just saying that, there are a lot of parts there that are… that we can clean up, man.
342 00:25:09.500 ⇒ 00:25:27.069 Samuel Roberts: Okay, yeah, I would say then, yeah, task one might be, like, sorting through that and flow, and just getting, like, moving things that are not as relevant down, or… I don’t know if we want to delete them or not, but I would just say, like, separate them so we know what the important flow still is, and then documenting…
343 00:25:27.700 ⇒ 00:25:34.989 Samuel Roberts: Like, data sources, agents, and we can work on a plan to then migrate that into code.
344 00:25:35.380 ⇒ 00:25:40.219 Samuel Roberts: That can hopefully live on their infra in some way, once we sort out that stuff as well.
345 00:25:40.220 ⇒ 00:25:48.730 Mustafa Raja: Yeah, yeah. Yeah, for this trigger, what this is, it, we are tracking, we are tracking, responses from Andy, I believe in this.
346 00:25:48.730 ⇒ 00:25:49.140 Samuel Roberts: Yep.
347 00:25:49.550 ⇒ 00:26:03.859 Mustafa Raja: Okay. And then, I believe if CSRs make any changes in the central docs, they can trigger a test using this button, and then the output that Andy generates ends up here.
348 00:26:04.890 ⇒ 00:26:05.210 Samuel Roberts: Okay.
349 00:26:05.210 ⇒ 00:26:14.449 Mustafa Raja: And if they need the references that Andy is hitting, they can trigger this button to get those.
350 00:26:14.850 ⇒ 00:26:16.260 Mustafa Raja: Okay.
351 00:26:16.510 ⇒ 00:26:21.390 Mustafa Raja: And I just don’t know if they are still using it or not, but… Yeah.
352 00:26:21.390 ⇒ 00:26:22.390 Samuel Roberts: No, definitely.
353 00:26:22.700 ⇒ 00:26:32.819 Samuel Roberts: definitely, like, logging is another key aspect here for us, in general. So, I would say, as part of this exercise, to document everything.
354 00:26:32.950 ⇒ 00:26:40.560 Samuel Roberts: Like, linking to those, and then knowing that if things are potentially triggering from them, or if we’re just logging to them,
355 00:26:41.190 ⇒ 00:26:44.500 Samuel Roberts: would be… clarifying.
356 00:26:46.570 ⇒ 00:26:47.280 Mustafa Raja: Yeah.
357 00:26:49.160 ⇒ 00:26:59.640 Samuel Roberts: And then, once that, like, once we have that kind of documentation together, we can all get together and make a plan for, you know, how we move those pieces into code.
358 00:27:00.510 ⇒ 00:27:07.109 Samuel Roberts: How exactly we’re gonna… you know, keep that in a… probably a new repo for Andy that would…
359 00:27:07.470 ⇒ 00:27:13.859 Samuel Roberts: be something we can test and run, but also then deploy to their Google Cloud platform eventually.
360 00:27:14.080 ⇒ 00:27:14.920 Samuel Roberts: So…
361 00:27:16.010 ⇒ 00:27:30.990 Samuel Roberts: I think that’s a worthwhile endeavor at this point. I’ve been kind of thinking about it for a little while, but I didn’t want to just say, like, yeah, let’s redo a bunch of this, but I think at this point that would help us,
362 00:27:32.110 ⇒ 00:27:37.269 Samuel Roberts: Get things kind of reorganized a little bit. Probably track down some of the triage issues cleaner.
363 00:27:37.450 ⇒ 00:27:41.200 Samuel Roberts: Understand where… Maybe add some…
364 00:27:41.320 ⇒ 00:27:45.279 Samuel Roberts: other logging that might have been harder to do in NIDN as well.
365 00:27:47.450 ⇒ 00:27:48.599 Samuel Roberts: Any thoughts on that?
366 00:27:54.950 ⇒ 00:27:57.460 Mustafa Raja: Yeah, I like this, this way to move ahead with this.
367 00:27:58.530 ⇒ 00:27:59.140 Samuel Roberts: Okay.
368 00:28:02.090 ⇒ 00:28:07.649 Samuel Roberts: Yeah, I don’t know who… I mean, do you guys want to work together on that documentation, or would…
369 00:28:07.980 ⇒ 00:28:08.679 Mustafa Raja: Yeah, I…
370 00:28:08.680 ⇒ 00:28:09.130 Casie Aviles: Yeah, yeah.
371 00:28:09.130 ⇒ 00:28:14.390 Mustafa Raja: Like, like, yeah, and, like, to do a working session in case you said.
372 00:28:15.600 ⇒ 00:28:23.480 Samuel Roberts: Okay, yeah, so, yeah, if you guys can get together, and to start to, like I said, just put together, like, a…
373 00:28:24.970 ⇒ 00:28:28.809 Samuel Roberts: just, like, a README kind of explanation of all this.
374 00:28:29.420 ⇒ 00:28:35.789 Samuel Roberts: Even if it means, like, renaming nodes and stuff, so that, like, we can know which ones we’re talking about at different times and stuff.
375 00:28:35.900 ⇒ 00:28:45.770 Samuel Roberts: You know, using the names and whatnot, just so that someone who’s less familiar, like myself, can get in here and understand it, and then we can figure out the plan to, like.
376 00:28:45.920 ⇒ 00:28:50.770 Samuel Roberts: which agents we need, which agents… yeah, I think that sounds good. Okay.
377 00:28:50.770 ⇒ 00:28:51.400 Mustafa Raja: Yeah.
378 00:28:51.400 ⇒ 00:28:52.020 Samuel Roberts: Great.
379 00:28:55.040 ⇒ 00:28:56.099 Samuel Roberts: Excuse me.
380 00:28:59.900 ⇒ 00:29:00.700 Samuel Roberts: Alright.
381 00:29:03.800 ⇒ 00:29:07.539 Mustafa Raja: And this is the 4-month, PM team.
382 00:29:08.100 ⇒ 00:29:11.600 Mustafa Raja: So it would be nice to have for the tickets to PRD.
383 00:29:13.590 ⇒ 00:29:15.040 Samuel Roberts: Oh, yes, yes.
384 00:29:19.920 ⇒ 00:29:21.990 Samuel Roberts: So is that gonna be… yeah.
385 00:29:22.390 ⇒ 00:29:23.500 Samuel Roberts: Gabe, have you seen that?
386 00:29:23.500 ⇒ 00:29:26.160 Mustafa Raja: this, I believe this could be…
387 00:29:26.300 ⇒ 00:29:29.559 Mustafa Raja: This could be done with LangFuse.
388 00:29:30.170 ⇒ 00:29:30.670 Samuel Roberts: Yeah.
389 00:29:43.940 ⇒ 00:29:47.130 Samuel Roberts: I’m trying to get into LangFuse right now, and I’m having a…
390 00:29:48.020 ⇒ 00:29:51.499 Samuel Roberts: Weird, go to project… there we go, I’m in the wrong project.
391 00:29:53.470 ⇒ 00:29:55.750 Samuel Roberts: Why am I getting an unauthorized…
392 00:29:59.080 ⇒ 00:30:01.069 Mustafa Raja: I’m doing any downtime yesterday.
393 00:30:04.470 ⇒ 00:30:06.650 Mustafa Raja: Hmm.
394 00:30:10.270 ⇒ 00:30:11.349 Mustafa Raja: Walks for me.
395 00:30:12.010 ⇒ 00:30:17.809 Samuel Roberts: Okay, yeah, I’m getting that loading, but I’m also getting a little unauthorized thing in the top corner. Maybe I’m just gonna log out and log back in.
396 00:30:18.180 ⇒ 00:30:18.960 Mustafa Raja: Yeah.
397 00:30:18.960 ⇒ 00:30:20.480 Samuel Roberts: Make sure everything’s good.
398 00:30:27.600 ⇒ 00:30:29.999 Samuel Roberts: Okay, and then I go to BringForge Internal…
399 00:30:30.680 ⇒ 00:30:36.089 Samuel Roberts: Yeah, alright, that cleared it up. We’re good. Okay, let me see about…
400 00:30:40.250 ⇒ 00:30:47.579 Samuel Roberts: Managing… Members… Alright, Gabe, I’m gonna invite you right now.
401 00:31:17.530 ⇒ 00:31:20.270 Samuel Roberts: Alright, Gabe, you should have an invite to Langviews.
402 00:31:20.270 ⇒ 00:31:20.920 Gabriel Lam: Okay.
403 00:31:25.940 ⇒ 00:31:27.000 Samuel Roberts: And then…
404 00:31:41.110 ⇒ 00:31:46.949 Samuel Roberts: And then I guess while we have this call still, maybe we can give you a little tour of Langviews, I suppose?
405 00:31:48.720 ⇒ 00:31:50.369 Gabriel Lam: Yes, that would be great.
406 00:31:52.640 ⇒ 00:31:57.910 Samuel Roberts: Excuse me, alright, let me… Unless Mustafa.
407 00:31:57.910 ⇒ 00:31:58.620 Mustafa Raja: Do you understand?
408 00:31:58.620 ⇒ 00:32:02.599 Samuel Roberts: jump over to LangFuse. Yeah, I can share, that’s fine. I just… I… you…
409 00:32:02.600 ⇒ 00:32:06.090 Mustafa Raja: You’re the one that put most of these things in here, I think, right? So, I figured.
410 00:32:06.090 ⇒ 00:32:14.150 Samuel Roberts: But I can… I can share… So.
411 00:32:14.150 ⇒ 00:32:14.850 Uttam Kumaran: Is, is lame.
412 00:32:14.850 ⇒ 00:32:15.950 Samuel Roberts: I don’t know exactly…
413 00:32:15.950 ⇒ 00:32:18.140 Uttam Kumaran: User pricing?
414 00:32:18.910 ⇒ 00:32:19.760 Uttam Kumaran: Or no?
415 00:32:19.760 ⇒ 00:32:21.970 Samuel Roberts: That’s a good question.
416 00:32:22.650 ⇒ 00:32:24.880 Samuel Roberts: It just let me invite someone, so I don’t…
417 00:32:41.340 ⇒ 00:32:43.050 Samuel Roberts: I don’t think so, it doesn’t look like it.
418 00:32:45.600 ⇒ 00:32:46.200 Uttam Kumaran: Okay.
419 00:32:46.200 ⇒ 00:32:53.010 Samuel Roberts: I think… I think if you don’t have a subscription, you only get a limited number of members, is what it was before.
420 00:32:54.650 ⇒ 00:32:55.020 Uttam Kumaran: Okay.
421 00:32:55.030 ⇒ 00:32:59.860 Samuel Roberts: Because when we set it up for… yeah, when we set it up for free, I could only invite, like.
422 00:33:00.020 ⇒ 00:33:01.200 Samuel Roberts: two people.
423 00:33:01.490 ⇒ 00:33:05.030 Samuel Roberts: But I think once you start on a plan.
424 00:33:05.280 ⇒ 00:33:09.320 Casie Aviles: Yeah, there’s unlimited users for core, but… Yeah.
425 00:33:09.320 ⇒ 00:33:13.680 Samuel Roberts: Initially, that’s what I… yep, then we’re good. We’re fine with user design. Okay, cool.
426 00:33:14.570 ⇒ 00:33:22.460 Samuel Roberts: So, the prompts, this is under the prompt management,
427 00:33:22.920 ⇒ 00:33:27.689 Samuel Roberts: We probably wanna… can we organize these into folders? Yeah, okay, there’s a folder, so…
428 00:33:28.300 ⇒ 00:33:33.280 Mustafa Raja: The folders doesn’t work with, the… Soon.
429 00:33:33.280 ⇒ 00:33:33.900 Samuel Roberts: Oh, within AM.
430 00:33:33.900 ⇒ 00:33:35.520 Mustafa Raja: Please note, yeah.
431 00:33:35.520 ⇒ 00:33:39.190 Samuel Roberts: That’s right, okay, that’s why I would like this. Okay, so maybe we can add some tags, then, for now.
432 00:33:39.560 ⇒ 00:33:43.980 Samuel Roberts: Just to keep things a little more organized, because right now, I mean, it’s easier, at least by the name.
433 00:33:44.700 ⇒ 00:33:50.620 Samuel Roberts: to figure out which ones these are. So these are… Gonna match those…
434 00:33:51.160 ⇒ 00:33:53.510 Samuel Roberts: a few nodes in the N8N.
435 00:33:53.770 ⇒ 00:33:58.550 Samuel Roberts: So there’s… I don’t know which one.
436 00:34:03.220 ⇒ 00:34:07.090 Samuel Roberts: Yeah, so this just has the prompt we can try to…
437 00:34:07.970 ⇒ 00:34:11.519 Samuel Roberts: you can work on. I believe there’s versioning here as well.
438 00:34:12.460 ⇒ 00:34:12.900 Mustafa Raja: Oh, yeah.
439 00:34:12.909 ⇒ 00:34:16.269 Samuel Roberts: It should be able to track and roll back and all that stuff,
440 00:34:17.859 ⇒ 00:34:22.919 Samuel Roberts: Okay. We can’t really run experiments as well, because it’s an end, isn’t that right?
441 00:34:22.920 ⇒ 00:34:23.649 Mustafa Raja: Yeah, yeah.
442 00:34:23.650 ⇒ 00:34:24.330 Samuel Roberts: Yeah.
443 00:34:24.469 ⇒ 00:34:28.679 Samuel Roberts: So at some point, we might be able to actually, like, Test from here more easily.
444 00:34:28.810 ⇒ 00:34:31.650 Samuel Roberts: But because of NATM, we’re not really able to do that.
445 00:34:31.800 ⇒ 00:34:34.790 Samuel Roberts: But… .
446 00:34:34.790 ⇒ 00:34:40.679 Casie Aviles: Yeah, the experiments there, I believe, are just more on testing it with different models.
447 00:34:41.159 ⇒ 00:34:43.659 Samuel Roberts: Is that all it is? Alright, so you can’t even do that within,
448 00:34:44.569 ⇒ 00:34:49.269 Samuel Roberts: It’s just pure… okay, so it’s not really part of another flow or anything in Monstra or anything?
449 00:34:50.659 ⇒ 00:34:51.639 Casie Aviles: Yeah, like…
450 00:34:51.639 ⇒ 00:34:52.989 Samuel Roberts: I suppose.
451 00:34:53.849 ⇒ 00:34:57.699 Samuel Roberts: I suppose we could build some test datasets from the NAN, maybe.
452 00:34:57.959 ⇒ 00:35:01.469 Samuel Roberts: Like, poppy some of that output, maybe?
453 00:35:02.289 ⇒ 00:35:07.039 Samuel Roberts: Like, the way we can pin things in N8N.
454 00:35:07.239 ⇒ 00:35:12.729 Samuel Roberts: Did we, like, give it… a few… Example runs, maybe?
455 00:35:13.029 ⇒ 00:35:14.739 Samuel Roberts: Would that be worth doing, so that we can…
456 00:35:14.740 ⇒ 00:35:15.210 Mustafa Raja: I should just run.
457 00:35:15.210 ⇒ 00:35:15.529 Samuel Roberts: this time.
458 00:35:15.850 ⇒ 00:35:16.780 Samuel Roberts: Confused?
459 00:35:17.320 ⇒ 00:35:20.580 Mustafa Raja: Yeah, I guess we can, we can work, work with it that way.
460 00:35:21.530 ⇒ 00:35:22.590 Samuel Roberts: Yeah…
461 00:35:22.890 ⇒ 00:35:28.589 Samuel Roberts: But I… they might not be worth it at this point, it might just be worth running the, you know, making these changes and then running it again on the platform.
462 00:35:28.840 ⇒ 00:35:29.920 Samuel Roberts: on the forge.
463 00:35:29.920 ⇒ 00:35:32.930 Gabriel Lam: I think that’s maybe a better use of time at this point.
464 00:35:32.930 ⇒ 00:35:35.160 Samuel Roberts: Yeah, exactly. Okay.
465 00:35:35.550 ⇒ 00:35:40.779 Samuel Roberts: Yeah, so that’s them, that’s them there. Where am I? Yeah, there’s a few here.
466 00:35:41.400 ⇒ 00:35:43.260 Samuel Roberts: Linear, linear, linear, linear.
467 00:35:43.770 ⇒ 00:35:51.989 Samuel Roberts: And I would defer to Mustafa and Casey about which ones
468 00:35:52.130 ⇒ 00:35:55.780 Samuel Roberts: Are the most, highest leverage ones to tweak.
469 00:35:55.780 ⇒ 00:35:56.330 Gabriel Lam: Okay.
470 00:35:56.330 ⇒ 00:35:56.990 Mustafa Raja: Okay. Yeah.
471 00:35:57.470 ⇒ 00:35:58.629 Samuel Roberts: They are on the flow.
472 00:35:59.310 ⇒ 00:36:04.510 Mustafa Raja: Yeah, so the ticket generator and generated tickets groomer would be the one.
473 00:36:04.690 ⇒ 00:36:07.059 Mustafa Raja: To work with.
474 00:36:07.400 ⇒ 00:36:17.890 Mustafa Raja: Okay, the first one. The groomer one only triggers… yeah, yeah, the groomer one only triggers for, client, client hub agents, so,
475 00:36:18.060 ⇒ 00:36:30.370 Mustafa Raja: If it’s… if the ticket is related to a client team, then, this, this prompt will be used, so if we can… if we can ask it to, give us output in the required format.
476 00:36:30.550 ⇒ 00:36:36.150 Mustafa Raja: That would… that would be nice, and similar thing for the tickets generator also.
477 00:36:36.870 ⇒ 00:36:37.570 Gabriel Lam: Okay.
478 00:36:40.520 ⇒ 00:36:41.060 Samuel Roberts: Yeah.
479 00:36:42.590 ⇒ 00:36:50.810 Mustafa Raja: Yeah, so the description here, we need, we need to, give it more clarity on what the description should look like.
480 00:36:51.210 ⇒ 00:36:52.160 Samuel Roberts: Yeah, yeah.
481 00:36:52.390 ⇒ 00:36:54.590 Uttam Kumaran: So then, how do you test, like, one of these?
482 00:36:55.970 ⇒ 00:37:00.950 Uttam Kumaran: like, how can I test this in, like, isolation just here? Is that possible, or no?
483 00:37:01.240 ⇒ 00:37:18.020 Samuel Roberts: That’s what we were just talking about. I think the best way to do it… right now, this experiment feature looks like you can give it a dataset, and so what we would have to do is set up, like, a bunch of… like, take a bunch of inputs that have already been run in N8N and build a dataset here to run it against.
484 00:37:18.460 ⇒ 00:37:20.829 Samuel Roberts: It can’t really…
485 00:37:20.990 ⇒ 00:37:33.059 Samuel Roberts: all it does is just this step, right? So we’d have to get that data, build that data set, and get the outputs. So for now, we think it’s probably just worth changing this, and then testing it on the forge and seeing how it goes.
486 00:37:33.360 ⇒ 00:37:34.519 Samuel Roberts: Because it will update them.
487 00:37:34.520 ⇒ 00:37:34.910 Mustafa Raja: Anyway.
488 00:37:40.710 ⇒ 00:37:46.529 Samuel Roberts: I think there might be a way to make that a little nicer with Monster, because Monstra has a whole little,
489 00:37:47.610 ⇒ 00:37:49.090 Samuel Roberts: dev environment.
490 00:37:49.400 ⇒ 00:37:57.220 Samuel Roberts: That you can isolate pieces, which I don’t know if I can show, because I don’t think I have anything running with that, but,
491 00:37:57.680 ⇒ 00:38:03.729 Samuel Roberts: the… Dev server… let me see if I can find a…
492 00:38:04.800 ⇒ 00:38:10.660 Samuel Roberts: This is not gonna have any screenshots of anything, but yeah, there’s a really nice… this is one of the things,
493 00:38:12.530 ⇒ 00:38:14.209 Samuel Roberts: Yeah, I’m not gonna have it here.
494 00:38:15.200 ⇒ 00:38:21.879 Uttam Kumaran: Yeah, I guess I just want to make it easy for folks to be able to pass in an input, test a prompt, see the output, and, like, kind of iterate, right?
495 00:38:22.050 ⇒ 00:38:26.010 Samuel Roberts: Yeah, that’s… Where is there, a screenshot here?
496 00:38:27.240 ⇒ 00:38:30.290 Samuel Roberts: That’s Langview Screenshot, but Monster Dev Server…
497 00:38:34.050 ⇒ 00:38:36.159 Samuel Roberts: That’s length view, that’s lengthy, yes.
498 00:38:37.320 ⇒ 00:38:39.920 Samuel Roberts: That’s gonna be Sigmas, yeah, okay.
499 00:38:40.330 ⇒ 00:38:41.899 Samuel Roberts: I don’t know if you’ve seen the…
500 00:38:42.990 ⇒ 00:38:45.419 Samuel Roberts: when you run the Mazda 8, is this it?
501 00:38:46.650 ⇒ 00:38:48.170 Samuel Roberts: It gives you this nice…
502 00:38:52.100 ⇒ 00:38:57.580 Samuel Roberts: So this might be a little more technical, because you’d have to actually get the master agent running, but I can’t even…
503 00:38:58.560 ⇒ 00:39:03.669 Samuel Roberts: So small, never mind. But if you can see here, like, you can isolate down to the agents, the workflows, the tools.
504 00:39:05.480 ⇒ 00:39:16.520 Samuel Roberts: So this is where, like, you’d be able to actually test and iterate more quickly on something. Lang views might be another good place to do it, if we could get the right
505 00:39:16.970 ⇒ 00:39:19.869 Samuel Roberts: datasets set up.
506 00:39:20.770 ⇒ 00:39:25.190 Samuel Roberts: But I think at this point.
507 00:39:25.600 ⇒ 00:39:30.279 Samuel Roberts: It’s probably, you know, easier to just make these changes and then run and see.
508 00:39:30.800 ⇒ 00:39:35.130 Samuel Roberts: How it does, because it’s… we’d have to pull all that data right now and get that set up, and…
509 00:39:35.900 ⇒ 00:39:40.510 Samuel Roberts: I think it’s probably a better use of time to just make the changes and see how it goes, because it should go immediately.
510 00:39:40.920 ⇒ 00:39:41.530 Uttam Kumaran: Okay.
511 00:39:42.660 ⇒ 00:39:46.130 Uttam Kumaran: I guess, Gabe, what else?
512 00:39:46.600 ⇒ 00:39:49.980 Uttam Kumaran: unique. I mean, for me, this is still too complicated, but…
513 00:39:49.980 ⇒ 00:39:51.090 Samuel Roberts: Yeah, yeah.
514 00:39:51.090 ⇒ 00:39:57.170 Uttam Kumaran: like… I guess I don’t know how else we can… Make that better right now.
515 00:39:58.000 ⇒ 00:39:58.500 Gabriel Lam: Yeah.
516 00:39:58.500 ⇒ 00:40:02.650 Samuel Roberts: Yeah, especially with N8N, I think this is the best thing we can do.
517 00:40:02.860 ⇒ 00:40:08.969 Samuel Roberts: Once it’s in… I’m trying to think if there’s a better way to do this on the forge, and…
518 00:40:09.650 ⇒ 00:40:16.389 Samuel Roberts: you know, if we could get that Mastra dev server somewhere that’s more accessible and not having to run it, you know…
519 00:40:16.890 ⇒ 00:40:20.060 Samuel Roberts: From the command line and stuff, but…
520 00:40:20.960 ⇒ 00:40:26.679 Samuel Roberts: once this is in code, I think we have a little bit more flexibility, I just don’t know where that… where exactly that gets us to.
521 00:40:27.060 ⇒ 00:40:29.240 Samuel Roberts: At this point. For testing.
522 00:40:32.030 ⇒ 00:40:34.050 Samuel Roberts: So I think it would be easier to just…
523 00:40:34.080 ⇒ 00:40:41.320 Gabriel Lam: test and, like, edit and deploy. I mean, sure, it’s gonna not be as isolated, or it’ll take a little longer, but…
524 00:40:43.170 ⇒ 00:40:45.919 Gabriel Lam: I think for the most part, it’ll get us through this week.
525 00:40:46.690 ⇒ 00:40:54.489 Samuel Roberts: Yeah. And if for some reason that is taking way too long, then that’s a good indication we can build one of these data sets from the previous data.
526 00:40:54.490 ⇒ 00:40:55.090 Gabriel Lam: Yeah.
527 00:40:59.420 ⇒ 00:41:05.670 Samuel Roberts: And then you can run it quicker, especially if, like, there’s, you know, the one that’s specifically the groomer, or the generator, whatever is, like…
528 00:41:05.800 ⇒ 00:41:08.120 Samuel Roberts: Where more of this work needs to happen.
529 00:41:08.520 ⇒ 00:41:09.270 Samuel Roberts: Huh.
530 00:41:10.230 ⇒ 00:41:15.579 Samuel Roberts: I mean, the other side of it is you could get into N8N and see the outputs and stuff,
531 00:41:16.210 ⇒ 00:41:18.169 Samuel Roberts: That’s probably helpful.
532 00:41:18.720 ⇒ 00:41:23.339 Samuel Roberts: I don’t know how comfortable you feel in there or not, but, we can definitely get you.
533 00:41:23.760 ⇒ 00:41:24.240 Gabriel Lam: Okay.
534 00:41:24.240 ⇒ 00:41:24.889 Samuel Roberts: you know.
535 00:41:25.080 ⇒ 00:41:28.760 Samuel Roberts: Into those nodes where you can pin things and see all the changes, and…
536 00:41:29.710 ⇒ 00:41:38.829 Samuel Roberts: make the changes in length use and rerun that step or something. That might actually be even. If you… if you want to get a little more into the weeds, I think that would be even,
537 00:41:39.540 ⇒ 00:41:42.929 Samuel Roberts: Nicer from, like, not having to rerun the whole step every time.
538 00:41:44.000 ⇒ 00:41:45.890 Samuel Roberts: In the whole, the whole process?
539 00:41:47.150 ⇒ 00:41:49.869 Samuel Roberts: So that could be worth it. I think it probably is.
540 00:41:51.870 ⇒ 00:41:54.069 Samuel Roberts: Are you in any den yet?
541 00:41:54.230 ⇒ 00:41:54.890 Gabriel Lam: Yes.
542 00:41:55.330 ⇒ 00:41:56.070 Gabriel Lam: Okay.
543 00:41:56.070 ⇒ 00:41:56.660 Samuel Roberts: Yeah, so…
544 00:41:56.660 ⇒ 00:42:02.060 Gabriel Lam: I only have visibility onto the flow, but I haven’t… been.
545 00:42:02.060 ⇒ 00:42:04.250 Samuel Roberts: Okay, but you have an account there, at least, so we’re…
546 00:42:04.290 ⇒ 00:42:05.569 Gabriel Lam: Yeah. We can add you.
547 00:42:06.460 ⇒ 00:42:08.909 Samuel Roberts: Yeah, I would say make sure,
548 00:42:09.990 ⇒ 00:42:16.960 Samuel Roberts: Actually, like, Mustafa, can you, after this, maybe show… or I should maybe share your screen, and we can show, like, Gabe, where these…
549 00:42:19.450 ⇒ 00:42:21.979 Samuel Roberts: Prompts are used, would that be helpful?
550 00:42:22.810 ⇒ 00:42:24.839 Samuel Roberts: Which parts of the flow you’re making?
551 00:42:27.200 ⇒ 00:42:32.440 Mustafa Raja: Yeah, so we want to know, which, which agents we’ll be updating, right, with the prompts.
552 00:42:34.860 ⇒ 00:42:40.019 Samuel Roberts: Yeah, so, like, when he makes a change to one of those prompts, where is that gonna get called?
553 00:42:40.020 ⇒ 00:42:40.519 Mustafa Raja: Yeah, yeah, yeah.
554 00:42:40.520 ⇒ 00:42:43.729 Samuel Roberts: My thought is we can pin some data there for him to test.
555 00:42:47.390 ⇒ 00:42:47.980 Mustafa Raja: Yeah, man.
556 00:42:47.980 ⇒ 00:42:51.990 Uttam Kumaran: Maybe I can have you guys do this off stand-up, I just want to get to…
557 00:42:51.990 ⇒ 00:42:52.450 Samuel Roberts: Sure, yeah.
558 00:42:52.450 ⇒ 00:42:53.630 Uttam Kumaran: Okay.
559 00:42:53.630 ⇒ 00:42:56.389 Samuel Roberts: Yeah, you two get together later and do that, yeah.
560 00:42:56.390 ⇒ 00:42:57.890 Mustafa Raja: Yeah, okay, perfect.
561 00:42:58.150 ⇒ 00:42:58.870 Mustafa Raja: Okay.
562 00:42:58.870 ⇒ 00:42:59.210 Samuel Roberts: Cool.
563 00:42:59.210 ⇒ 00:43:03.460 Uttam Kumaran: Maybe we can talk about, let’s talk about…
564 00:43:04.070 ⇒ 00:43:07.410 Uttam Kumaran: Maybe for Honey Stinger, I know Sam…
565 00:43:07.830 ⇒ 00:43:12.419 Uttam Kumaran: You tested Walmart. Are you gonna… you’re gonna be downloading some of that Walmart data today?
566 00:43:12.900 ⇒ 00:43:20.490 Samuel Roberts: Yeah, I… so I ran a quick one last night just to see how far back it went, and it only went back to, like, February of last year?
567 00:43:20.880 ⇒ 00:43:23.399 Samuel Roberts: Do we know when they started on there? Is that…
568 00:43:25.230 ⇒ 00:43:28.960 Uttam Kumaran: If you… I would just say take whatever it has, and…
569 00:43:28.960 ⇒ 00:43:29.630 Samuel Roberts: Okay.
570 00:43:30.130 ⇒ 00:43:33.849 Uttam Kumaran: And then I can confirm if whatever questions you have, I can send it over to him.
571 00:43:33.850 ⇒ 00:43:37.879 Samuel Roberts: Okay, yeah, I’m just gonna get that into Mother Duck later, after this.
572 00:43:38.380 ⇒ 00:43:39.000 Uttam Kumaran: Okay.
573 00:43:40.050 ⇒ 00:43:48.740 Samuel Roberts: And then I was trying to figure out if there were other… there were a few other dashboards that were, like, by product. I’m not sure if I can mass export that or not, but I’m gonna see what else I can pull out of there.
574 00:43:49.210 ⇒ 00:43:49.780 Uttam Kumaran: Okay.
575 00:43:50.160 ⇒ 00:43:55.219 Uttam Kumaran: Yeah, we can get… we can get as much… That would be great.
576 00:43:55.720 ⇒ 00:43:57.030 Samuel Roberts: Okay.
577 00:43:57.430 ⇒ 00:43:59.169 Uttam Kumaran: Okay, cool, and then…
578 00:43:59.560 ⇒ 00:44:08.189 Uttam Kumaran: Casey, is there… I know you’re working on text, the SQL stuff for Eden, is there other, like… are you so… are you doing, like, analysis work or other work for Eden right now?
579 00:44:08.900 ⇒ 00:44:11.530 Casie Aviles: I think they were…
580 00:44:11.890 ⇒ 00:44:17.839 Casie Aviles: What Henry asked me for help was something about Mixpanel, where I had to, like, set up
581 00:44:18.020 ⇒ 00:44:23.450 Casie Aviles: heatmaps, and he was asking if I need JavaScript, so I’m helping him for that.
582 00:44:24.050 ⇒ 00:44:27.480 Casie Aviles: But I think for, like, analysis in general,
583 00:44:28.000 ⇒ 00:44:33.210 Casie Aviles: Yeah, I don’t think we have any analysis yet for, like, Eden.
584 00:44:34.240 ⇒ 00:44:34.890 Uttam Kumaran: Okay.
585 00:44:35.600 ⇒ 00:44:42.360 Uttam Kumaran: Okay, great, and then you’re still just working on ABC… Stop.
586 00:44:43.380 ⇒ 00:44:49.310 Casie Aviles: Yeah, later we’ll have, like, a session with Amber to digest some of the triage.
587 00:44:50.820 ⇒ 00:44:51.390 Uttam Kumaran: Okay.
588 00:44:53.330 ⇒ 00:44:54.110 Uttam Kumaran: Okay.
589 00:44:56.390 ⇒ 00:45:03.599 Uttam Kumaran: Great, yeah, I think it would be great, like, I still haven’t seen anything about how we’re gonna start to get alerts into Slack, like…
590 00:45:03.940 ⇒ 00:45:13.880 Uttam Kumaran: Are you guys working on that? Like, every… almost every two weeks, there’s, like, an Andy outage, and, like, the client finds out faster than we do, so, like, what’s the plan there?
591 00:45:16.990 ⇒ 00:45:22.419 Mustafa Raja: Yeah, I believe Casey added, Slack support in the LangFuse node.
592 00:45:22.570 ⇒ 00:45:26.350 Mustafa Raja: So if LangFuse has any outage, we would know.
593 00:45:28.250 ⇒ 00:45:35.759 Casie Aviles: Yeah, it’s just mostly just the error, but I think we still have to work on, like, you know, getting general alerts on.
594 00:45:36.310 ⇒ 00:45:43.759 Casie Aviles: Outages, so… I think one of the suggestions raised was to have, like, an RSS subscription,
595 00:45:44.320 ⇒ 00:45:46.890 Casie Aviles: So that’s something that we’ve yet to set up.
596 00:45:47.490 ⇒ 00:45:50.649 Casie Aviles: What I added initially for now is to just…
597 00:45:51.090 ⇒ 00:45:54.319 Casie Aviles: I looked at the failing nodes, and I added more.
598 00:45:55.640 ⇒ 00:45:58.190 Casie Aviles: alerting, so it would send to Slack.
599 00:45:58.190 ⇒ 00:46:01.609 Uttam Kumaran: Did the alert come in… oh, okay, it comes into Slack. Did we test that out?
600 00:46:03.160 ⇒ 00:46:09.370 Casie Aviles: Yeah, it should be, like, in… it’s a different channel, it’s the… Client ABC logs.
601 00:46:09.500 ⇒ 00:46:14.479 Uttam Kumaran: No, no, put it all into the ABC… nobody, like, put it all into the normal ABC channel.
602 00:46:14.760 ⇒ 00:46:19.170 Uttam Kumaran: Yeah, I’m not… I don’t even know what the logs… I’m not checking logs at all.
603 00:46:19.930 ⇒ 00:46:20.630 Casie Aviles: Okay.
604 00:46:25.140 ⇒ 00:46:26.250 Casie Aviles: Yeah, yeah, I’ll move it.
605 00:46:26.250 ⇒ 00:46:32.940 Uttam Kumaran: These are all now systems that are getting into production, so, like, we just need to have some posture around alerting.
606 00:46:33.380 ⇒ 00:46:38.930 Uttam Kumaran: So, it just really, like…
607 00:46:39.840 ⇒ 00:46:44.370 Uttam Kumaran: just… yeah, I just… this is now the… probably the third or fourth time I’ve asked about
608 00:46:44.590 ⇒ 00:46:48.050 Uttam Kumaran: trying to catch these before the… before the ABC team does.
609 00:46:48.190 ⇒ 00:46:49.160 Uttam Kumaran: So…
610 00:46:49.490 ⇒ 00:46:58.730 Uttam Kumaran: Can you make sure that we’ve… there’s an example of an… like, if you can backtest it and show that the error that just happened would have got caught?
611 00:46:59.170 ⇒ 00:47:03.199 Uttam Kumaran: And it would have sent an alert to the client channel. Like, that’s what I want to see.
612 00:47:04.210 ⇒ 00:47:04.890 Casie Aviles: Okay.
613 00:47:06.180 ⇒ 00:47:13.660 Uttam Kumaran: like, I don’t… but I… again, like, and even, like, all these triage problems, like, I’m not as interested in that, like, we need to build…
614 00:47:13.810 ⇒ 00:47:21.670 Uttam Kumaran: like, I’m interested in two things. I want to move this to code, so I want to see a, like, I just need, like, a notion or a plan around that.
615 00:47:21.760 ⇒ 00:47:23.859 Samuel Roberts: I think if you can work with Sam…
616 00:47:23.860 ⇒ 00:47:27.989 Uttam Kumaran: Yeah, but I… but I just need to, like, we just need it written down somewhere.
617 00:47:28.210 ⇒ 00:47:28.999 Uttam Kumaran: So I need to know.
618 00:47:29.000 ⇒ 00:47:29.430 Samuel Roberts: Yeah, we didn’t.
619 00:47:29.430 ⇒ 00:47:36.849 Uttam Kumaran: what’s in scope, and then how long it’s gonna take, and then when we can get that done that you can work with Amber on, and then second is this alerting thing.
620 00:47:38.980 ⇒ 00:47:39.400 Casie Aviles: Oh, boy.
621 00:47:39.400 ⇒ 00:47:39.850 Uttam Kumaran: Yeah.
622 00:47:40.040 ⇒ 00:47:42.280 Casie Aviles: I’ll work on the notion, then, for this.
623 00:47:43.520 ⇒ 00:47:44.430 Uttam Kumaran: Okay, cool.
624 00:47:49.810 ⇒ 00:47:53.030 Uttam Kumaran: Okay, so that’s, like, ABC Eden stuff.
625 00:47:56.100 ⇒ 00:48:06.529 Uttam Kumaran: And then, Mustafa, you’re, we’re… we basically finished… we sent the message yesterday on default, we finished the captain data and other… the two vendors, right?
626 00:48:07.150 ⇒ 00:48:07.930 Mustafa Raja: Yeah.
627 00:48:10.360 ⇒ 00:48:15.439 Mustafa Raja: We still don’t have, anything for Catalyst.
628 00:48:19.320 ⇒ 00:48:22.520 Uttam Kumaran: Yeah, I will… I’ll send an… I’ll keep sending a note there, so…
629 00:48:23.350 ⇒ 00:48:23.970 Mustafa Raja: Yeah.
630 00:48:27.070 ⇒ 00:48:31.770 Mustafa Raja: Also, I added the dbt stuff that you sent.
631 00:48:32.070 ⇒ 00:48:46.279 Mustafa Raja: Yeah, it’s really helpful. I read the Notion doc, all of the Notion doc, and then I’m looking into the best practices that DBT Labs has, and yeah, really helpful. For which one? I should have…
632 00:48:46.610 ⇒ 00:48:49.480 Mustafa Raja: the dbt Labs, best practices.
633 00:48:49.770 ⇒ 00:48:52.570 Uttam Kumaran: I guess, but what are we setting up dbt for?
634 00:48:54.140 ⇒ 00:48:55.480 Mustafa Raja: The honey finger thing.
635 00:48:55.750 ⇒ 00:48:56.640 Uttam Kumaran: Oh, okay, okay.
636 00:48:56.640 ⇒ 00:48:59.339 Mustafa Raja: Intermediate one, yeah, intermediate models.
637 00:48:59.860 ⇒ 00:49:00.850 Uttam Kumaran: Okay, okay.
638 00:49:04.780 ⇒ 00:49:05.540 Uttam Kumaran: Okay.
639 00:49:06.450 ⇒ 00:49:12.359 Uttam Kumaran: Great, and then on CTA, yeah, Sam, I think there’s to-dos…
640 00:49:12.880 ⇒ 00:49:18.130 Uttam Kumaran: from yesterday. I think one thing, Rico, that could be helpful is…
641 00:49:18.600 ⇒ 00:49:23.589 Uttam Kumaran: And this is something that I think, Gabe, we should work on, but, like, at the end of every, like.
642 00:49:24.080 ⇒ 00:49:27.719 Uttam Kumaran: At the end of every client meeting, and we’ve sort of had this before, but
643 00:49:27.840 ⇒ 00:49:32.380 Uttam Kumaran: One thing that’s so manual right now is, like, someone has to go in, get the transcript.
644 00:49:32.740 ⇒ 00:49:47.000 Uttam Kumaran: and basically send an update into Slack, and, like, it’s just this whole process, we’re seeing how manual it is, and we’re gonna bring on another one or two PMs, but a lot of them are gonna be going through the same process, so…
645 00:49:47.270 ⇒ 00:49:50.430 Uttam Kumaran: But maybe in the meantime,
646 00:49:51.330 ⇒ 00:49:57.390 Uttam Kumaran: like, I don’t know, Sam, like, I don’t have the bandwidth right now to create tickets for everything, so what do you think is, like, best?
647 00:50:00.680 ⇒ 00:50:06.119 Samuel Roberts: Well, I was working on the, like, just the graph, but I wasn’t sure, like, starting to figure out
648 00:50:06.920 ⇒ 00:50:10.980 Samuel Roberts: where we’re… how we’re moving things and figuring out the ETL, is that the… like.
649 00:50:10.980 ⇒ 00:50:16.160 Uttam Kumaran: No, just, like, all… just kind of, like, all that… I just… just, like, all the tasks that are coming up.
650 00:50:16.310 ⇒ 00:50:20.409 Uttam Kumaran: Like, I just don’t… I just don’t have the bandwidth to, like, PM so closely right now.
651 00:50:21.890 ⇒ 00:50:30.239 Uttam Kumaran: So, wondering, like, what you think we can do… Like… Do you wanna…
652 00:50:30.700 ⇒ 00:50:36.280 Uttam Kumaran: create tickets, or do you want to work with Rico to create tickets, or should we just, like, kind of run it without tickets?
653 00:50:38.010 ⇒ 00:50:46.619 Samuel Roberts: I mean, probably having tickets is a good idea, as much as, you know, I would be like, oh, we can do it without it. I think we should make some tickets. I can work with Rico to ticket some stuff out.
654 00:50:47.020 ⇒ 00:50:48.410 Uttam Kumaran: Okay. Okay.
655 00:50:50.610 ⇒ 00:50:56.780 Uttam Kumaran: Okay, great. I think that’s all, kind of, for engineering work. Awash and Demilade are out today, so I’ll be on the next call.
656 00:50:56.990 ⇒ 00:50:59.470 Uttam Kumaran: As well, to support, so…
657 00:50:59.780 ⇒ 00:51:05.180 Uttam Kumaran: Yeah, and then I think, I’ll… I’m gonna have time… I’m gonna be working on engineering stuff.
658 00:51:05.480 ⇒ 00:51:07.520 Uttam Kumaran: Basically, the rest of the day, so…
659 00:51:07.640 ⇒ 00:51:11.960 Uttam Kumaran: Maybe I’ll, I’ll grab time with…
660 00:51:12.360 ⇒ 00:51:16.900 Uttam Kumaran: I’ll probably just send a note if I’m working on stuff, and we can all hop on and pair on things.
661 00:51:18.080 ⇒ 00:51:18.680 Gabriel Lam: Okay.
662 00:51:19.250 ⇒ 00:51:20.180 Uttam Kumaran: Okay, okay.
663 00:51:20.940 ⇒ 00:51:21.700 Uttam Kumaran: Okay.
664 00:51:21.820 ⇒ 00:51:23.369 Uttam Kumaran: Alright, thank you guys.
665 00:51:24.790 ⇒ 00:51:25.360 Samuel Roberts: Any…
666 00:51:25.360 ⇒ 00:51:26.150 Mustafa Raja: Thank you.
667 00:51:26.670 ⇒ 00:51:27.380 Uttam Kumaran: Bye.
668 00:51:30.340 ⇒ 00:51:31.809 Gabriel Lam: Is it okay if I…