Meeting Title: ABC working session Date: 2026-04-08 Meeting participants: Casie Aviles, Samuel Roberts, Pranav
WEBVTT
1 00:01:26.000 ⇒ 00:01:27.600 Samuel Roberts: Hello.
2 00:01:28.970 ⇒ 00:01:29.720 Casie Aviles: Hey, son.
3 00:01:31.830 ⇒ 00:01:33.099 Samuel Roberts: How’s it going today?
4 00:01:35.300 ⇒ 00:01:37.689 Casie Aviles: Yeah, doing okay. I’m just…
5 00:01:37.900 ⇒ 00:01:42.169 Casie Aviles: Working on the… yeah, the intake for… for…
6 00:01:42.170 ⇒ 00:01:42.870 Samuel Roberts: Okay.
7 00:01:42.870 ⇒ 00:01:44.220 Casie Aviles: the linear comments.
8 00:01:45.100 ⇒ 00:01:53.379 Samuel Roberts: Cool, cool. Yeah, I saw the video that you posted on the linear ticket, and it looked good. I think you’re right, the webhook is probably better than the polling.
9 00:01:53.520 ⇒ 00:01:55.360 Samuel Roberts: That was a good call.
10 00:01:56.710 ⇒ 00:01:59.369 Casie Aviles: So, yeah, right now it’s…
11 00:01:59.660 ⇒ 00:02:06.790 Casie Aviles: I was able to ask ops to set up the webhook, so I have been testing it.
12 00:02:07.820 ⇒ 00:02:08.469 Casie Aviles: Oh, good.
13 00:02:08.479 ⇒ 00:02:09.339 Samuel Roberts: Okay, great.
14 00:02:09.830 ⇒ 00:02:10.180 Casie Aviles: Yeah, good.
15 00:02:10.180 ⇒ 00:02:11.759 Samuel Roberts: So they got it done quickly?
16 00:02:12.520 ⇒ 00:02:18.659 Casie Aviles: Yeah. Yeah, I did just show… give them a Zoom clip as well, to walk them through. Okay, perfect.
17 00:02:20.300 ⇒ 00:02:23.960 Casie Aviles: Oh, that’s my linear. Oh, that’s my cursor.
18 00:02:25.870 ⇒ 00:02:33.899 Casie Aviles: Yeah, I was testing it, so I’ve updated, like, the fields that we needed to add.
19 00:02:34.310 ⇒ 00:02:37.439 Casie Aviles: So this is just another test, and this is for debugging.
20 00:02:37.850 ⇒ 00:02:38.890 Casie Aviles: So… Okay.
21 00:02:39.420 ⇒ 00:02:46.290 Casie Aviles: Now it should be able to pick up, although right now, I pushed an additional change, and it’s returning…
22 00:02:47.260 ⇒ 00:02:51.660 Casie Aviles: an error, so I’m just checking out why that’s the case.
23 00:02:52.840 ⇒ 00:02:53.190 Samuel Roberts: Okay.
24 00:02:53.190 ⇒ 00:02:59.640 Casie Aviles: because I added, like, a filter that, basically, it only picks up
25 00:03:00.280 ⇒ 00:03:06.649 Casie Aviles: on the comment, if it’s in the right team. So it’s if it’s in ABC Home.
26 00:03:07.110 ⇒ 00:03:08.050 Casie Aviles: Definitely.
27 00:03:08.050 ⇒ 00:03:08.690 Samuel Roberts: Yeah, cool.
28 00:03:10.160 ⇒ 00:03:18.729 Casie Aviles: Yeah, so I’ll just work on that, and… I think what I’m… I’m a little…
29 00:03:18.940 ⇒ 00:03:25.650 Casie Aviles: Like, I’m not sure yet what… what exactly to do here is, like, the state part.
30 00:03:25.980 ⇒ 00:03:28.739 Casie Aviles: I think that should be legalized, right? Yeah, let’s talk about that.
31 00:03:29.280 ⇒ 00:03:36.289 Samuel Roberts: Yeah, so I think those will be labels, and then that’s how we’ll keep track of where it is in that process. The state…
32 00:03:36.470 ⇒ 00:03:39.239 Samuel Roberts: machine got kind of big, I felt like.
33 00:03:39.550 ⇒ 00:03:40.570 Casie Aviles: Two states.
34 00:03:40.770 ⇒ 00:03:41.530 Samuel Roberts: But…
35 00:03:41.650 ⇒ 00:03:51.480 Samuel Roberts: We can maybe try to simplify it, if there’s a good way to do it. I guess there were… as I was going through it, I was like, this seems complicated, but then each state kind of made sense to me.
36 00:03:51.670 ⇒ 00:03:57.730 Samuel Roberts: So… We may, yeah, we can look at it if you want to talk it through.
37 00:04:00.680 ⇒ 00:04:04.330 Casie Aviles: Yes, it’s… let me see… I think it’s this one.
38 00:04:05.570 ⇒ 00:04:07.920 Samuel Roberts: Yes, is there…
39 00:04:09.090 ⇒ 00:04:16.959 Samuel Roberts: Applying… this should be a… I think this got updated. What… what are we looking at here? Which, branch are we on?
40 00:04:20.320 ⇒ 00:04:23.330 Casie Aviles: So, maybe 182370.
41 00:04:23.330 ⇒ 00:04:30.680 Samuel Roberts: Yeah, maybe just refresh? I thought I added the batch. Oh, there it is, never mind, I see it there now. Okay, sorry, I just didn’t see it at first.
42 00:04:30.790 ⇒ 00:04:31.980 Samuel Roberts: It’s there.
43 00:04:32.960 ⇒ 00:04:38.989 Samuel Roberts: Yeah, okay, good. Okay. Yeah, Pranav pointed out that we will need to batch them.
44 00:04:39.140 ⇒ 00:04:44.770 Samuel Roberts: And so we need… we need a state between… Approved and applying, basically.
45 00:04:47.410 ⇒ 00:04:49.870 Samuel Roberts: So, yeah.
46 00:04:52.980 ⇒ 00:04:56.380 Casie Aviles: Alright, so… I guess what?
47 00:04:57.030 ⇒ 00:05:02.550 Casie Aviles: So, do we have, like, these duplicate flags? But I don’t think we have, like, a…
48 00:05:03.160 ⇒ 00:05:12.020 Casie Aviles: A logic yet that… You know, that looks at whether something is a duplicate, so…
49 00:05:12.020 ⇒ 00:05:13.800 Samuel Roberts: No, I think that’s another ticket.
50 00:05:16.280 ⇒ 00:05:19.489 Samuel Roberts: Trying to see where the tickets are here. Let me get my hero.
51 00:05:20.190 ⇒ 00:05:22.999 Casie Aviles: Yeah, so far I have, like, two, three point…
52 00:05:23.450 ⇒ 00:05:29.669 Casie Aviles: Yeah, 3 tickets under, like, CDC3, which is 3.1 and 3.2.
53 00:05:30.220 ⇒ 00:05:30.940 Samuel Roberts: Okay.
54 00:05:37.440 ⇒ 00:05:39.470 Samuel Roberts: Alright, so that’s that one.
55 00:05:41.180 ⇒ 00:05:43.459 Samuel Roberts: Sorry, I’m just looking through them right now, trying to remember them all.
56 00:05:53.360 ⇒ 00:05:55.050 Samuel Roberts: Yeah, where’s the duplicate?
57 00:06:05.520 ⇒ 00:06:06.830 Pranav: Hey, what are we looking for?
58 00:06:08.140 ⇒ 00:06:15.439 Samuel Roberts: Just going through the tickets, because he was, wondering about the logic for the duplicate and the conflict stuff, and I think it should be in another ticket, I just want to see.
59 00:06:16.120 ⇒ 00:06:16.600 Pranav: Gotcha.
60 00:06:16.600 ⇒ 00:06:20.329 Samuel Roberts: point one. I think it’s the, yeah, the 5-point tickets.
61 00:06:21.600 ⇒ 00:06:24.620 Casie Aviles: Oh, okay, so it’s… Further down, then.
62 00:06:25.310 ⇒ 00:06:26.000 Samuel Roberts: Yeah.
63 00:06:27.360 ⇒ 00:06:29.520 Casie Aviles: Okay, so for these three point…
64 00:06:29.720 ⇒ 00:06:37.100 Casie Aviles: or, like, CDC3 tickets, do we have to set up, like, the state machine as well? Or just the…
65 00:06:39.260 ⇒ 00:06:41.269 Casie Aviles: You know, listening, and then…
66 00:06:43.400 ⇒ 00:06:48.450 Casie Aviles: Do, like, the analysis part, like this, these 3.1 and 3.2.
67 00:06:49.300 ⇒ 00:06:51.669 Samuel Roberts: 3.1 and 3.2 are…
68 00:07:08.570 ⇒ 00:07:15.139 Samuel Roberts: Yeah, I think this is just kind of laying out the… the structure of how everything will interact with linear, and…
69 00:07:17.060 ⇒ 00:07:22.370 Samuel Roberts: After that is when we’ll actually build in the logic for assigning those labels.
70 00:07:23.480 ⇒ 00:07:25.230 Casie Aviles: Okay, okay, yeah, thanks.
71 00:07:25.760 ⇒ 00:07:26.510 Casie Aviles: Cool, because I…
72 00:07:26.510 ⇒ 00:07:27.049 Pranav: Oh, sure.
73 00:07:28.250 ⇒ 00:07:28.650 Samuel Roberts: Yeah.
74 00:07:28.650 ⇒ 00:07:29.940 Pranav: for…
75 00:07:30.640 ⇒ 00:07:49.559 Pranav: So, I guess, are we kind of aligned on milestones, and we’re just, like, wondering, like, okay, which tickets are defining which, which, I guess, task to complete? And making sure that there are tickets for each task, or are we also kind of a little bit confused on the milestone itself?
76 00:07:51.470 ⇒ 00:07:52.650 Pranav: A little bit to that.
77 00:07:52.650 ⇒ 00:07:54.130 Samuel Roberts: I’ll check back, please.
78 00:07:54.460 ⇒ 00:08:03.379 Samuel Roberts: Yeah, because Milestone 3… is… placement automation, right? At least in linear.
79 00:08:06.100 ⇒ 00:08:09.159 Samuel Roberts: And then milestone 4 is Duplicate conflict engine.
80 00:08:10.260 ⇒ 00:08:11.290 Samuel Roberts: Is that right?
81 00:08:12.580 ⇒ 00:08:17.230 Pranav: Yes, I’m forgetting what the numbers, but I’d be…
82 00:08:17.230 ⇒ 00:08:20.629 Samuel Roberts: Yeah, M3… M3 is what it looks like here, is…
83 00:08:20.920 ⇒ 00:08:25.340 Samuel Roberts: what UYB put in linear through here? So yeah, that… alright, thank you, Casey, yeah.
84 00:08:25.670 ⇒ 00:08:26.770 Pranav: Oh, perfect.
85 00:08:27.230 ⇒ 00:08:39.619 Pranav: Yeah. Yeah, so I guess, yeah, so M3, M4, perfect, yeah, those are just on the same day, those milestones. How are we feeling about that? So I think…
86 00:08:39.620 ⇒ 00:08:41.260 Samuel Roberts: Oh, yeah.
87 00:08:43.150 ⇒ 00:08:46.839 Pranav: I mean… It’s okay if we don’t get it.
88 00:08:47.160 ⇒ 00:08:48.430 Samuel Roberts: Yeah, I think we’re…
89 00:08:50.780 ⇒ 00:08:52.689 Pranav: Are we able to get one of them done?
90 00:08:52.810 ⇒ 00:08:53.920 Pranav: Like, yeah, that’s…
91 00:08:53.920 ⇒ 00:08:56.430 Samuel Roberts: We’re shooting for M3, was what I was thinking.
92 00:08:57.380 ⇒ 00:09:00.910 Samuel Roberts: Those are the ones in… in the cycle now, I think.
93 00:09:01.590 ⇒ 00:09:02.440 Pranav: Okay, perfect.
94 00:09:02.440 ⇒ 00:09:02.880 Samuel Roberts: Is there a good.
95 00:09:02.880 ⇒ 00:09:06.930 Pranav: And also, your vet isn’t gonna even be there on…
96 00:09:07.090 ⇒ 00:09:13.849 Pranav: On Friday, but, I mean, Janiece is gonna be more interested in this stuff anyway, so… yeah. Yeah. That’s… so she’s gonna be there.
97 00:09:13.850 ⇒ 00:09:19.079 Samuel Roberts: I’m also traveling on Friday, so I may or may not be available.
98 00:09:19.400 ⇒ 00:09:27.450 Samuel Roberts: I have out of office on, but if you need me for something, I’ll be accessible, but I’ll be flying early, so.
99 00:09:29.500 ⇒ 00:09:31.070 Pranav: Yeah, that’s totally fine.
100 00:09:31.070 ⇒ 00:09:31.600 Samuel Roberts: Yeah.
101 00:09:32.040 ⇒ 00:09:35.240 Samuel Roberts: I just wanted to let you guys know, I forgot to mention that earlier, but yeah.
102 00:09:35.240 ⇒ 00:09:49.080 Pranav: Yeah, I appreciate that. Yeah, let’s, let’s try to go for M3, and have that in a state where I can demo that with, Janiece on Friday. What we’re going to mostly be talking about is…
103 00:09:49.380 ⇒ 00:09:56.089 Pranav: we’re gonna be talking about just, like, the new workflow, and so I will be kind of…
104 00:09:56.430 ⇒ 00:10:10.779 Pranav: fully just, like, talking to them about, hey, this is kind of, like, the new system we’re putting in place, going forward with all the triage circuits. And, yeah, Sam, kind of… I read through that… that whole document, and it looked great. Just, like, that one missing state.
105 00:10:11.010 ⇒ 00:10:18.900 Samuel Roberts: Yeah, no, that was a good catch, that was a good catch. We were just talking about it, like, how complex the state machine kind of got, but I feel like every state
106 00:10:19.090 ⇒ 00:10:24.440 Samuel Roberts: makes sense, and I wasn’t really sure there was much we could, kind of condense anyway, so…
107 00:10:25.450 ⇒ 00:10:27.010 Pranav: Yeah, and…
108 00:10:27.010 ⇒ 00:10:30.160 Samuel Roberts: But I think, I think it’s still, like, I think it’s good.
109 00:10:30.950 ⇒ 00:10:44.429 Pranav: Yeah, yeah. And we can talk about… and I will… I’ll read through it again, and I can see if… how we can simplify it, but I think all the… as currently constructed, like, it’s all-encompassing for all the edge cases that…
110 00:10:44.830 ⇒ 00:10:46.200 Samuel Roberts: Yeah, exactly, exactly.
111 00:10:46.200 ⇒ 00:10:52.949 Pranav: Right? So, from there, I will figure out, hey, is this too complex? Like, I will… I will kind of go through…
112 00:10:53.250 ⇒ 00:10:58.790 Pranav: a couple of triage tickets to just see, like, okay, how would this actually work, hypothetically?
113 00:10:58.790 ⇒ 00:11:12.730 Samuel Roberts: That’s really a really good point, because I was thinking of it from, like, the technical side more, but if there’s too much for them… I think a lot of it is stuff that we’re handling anyway, like, a lot of those states. I think there’s only a few states where they’re really, like, gated to them.
114 00:11:13.550 ⇒ 00:11:14.669 Pranav: Right. Yeah.
115 00:11:14.670 ⇒ 00:11:21.730 Samuel Roberts: So I think, from the client side, ideally, it doesn’t seem quite as complicated, it’s just a bunch of steps we have to kind of go through to, like.
116 00:11:22.290 ⇒ 00:11:23.909 Samuel Roberts: Process it, you know.
117 00:11:25.080 ⇒ 00:11:32.820 Samuel Roberts: find the document locations, you know, I think that’s really where the complexity lies. I don’t think it’s really much surface to them.
118 00:11:33.120 ⇒ 00:11:37.269 Samuel Roberts: Fortunately, so… But definitely let me know if there’s anything we need to tweak there, but…
119 00:11:38.380 ⇒ 00:11:41.420 Pranav: Okay. Yep, I definitely will.
120 00:11:42.000 ⇒ 00:11:44.659 Pranav: Yeah, let me know when that’s in a state of…
121 00:11:44.790 ⇒ 00:11:48.869 Pranav: like, me to look at it again. I’m not sure if you already added that, additional state.
122 00:11:48.870 ⇒ 00:11:56.919 Samuel Roberts: Yeah, it’s in there, I just… I merged it into, the previous branch, so there’s only one PR there, I’ll…
123 00:11:57.060 ⇒ 00:12:01.460 Samuel Roberts: I can send you that. I think it was 70… actually, I don’t remember now, they’re both 70-something.
124 00:12:01.600 ⇒ 00:12:09.250 Samuel Roberts: But… Yeah, it’s in there, I just added that, that state, effectively.
125 00:12:09.620 ⇒ 00:12:10.980 Samuel Roberts: Okay, awesome.
126 00:12:10.980 ⇒ 00:12:13.670 Pranav: So, that ticket’s still assigned to me?
127 00:12:14.290 ⇒ 00:12:19.380 Samuel Roberts: I think I closed them both now, but feel free to… They were assigned.
128 00:12:19.380 ⇒ 00:12:20.360 Pranav: Okay.
129 00:12:22.740 ⇒ 00:12:27.049 Pranav: Okay, let me find that real quick, and let me just reopen it and assign it to me.
130 00:12:28.670 ⇒ 00:12:32.190 Pranav: Oh, actually, no, there’s a ticket for me to train on triage workflow.
131 00:12:32.950 ⇒ 00:12:35.730 Samuel Roberts: Yeah, that’s two, I think, right?
132 00:12:35.730 ⇒ 00:12:38.919 Pranav: Yep, CDC-2. So…
133 00:12:44.580 ⇒ 00:12:48.059 Samuel Roberts: But I think that was just, like, once we had the system in place.
134 00:12:49.380 ⇒ 00:12:53.110 Samuel Roberts: get it to… A good state for…
135 00:12:53.290 ⇒ 00:12:58.720 Samuel Roberts: Training them on how to use it, but that might be a little premature on training them at this point, but…
136 00:13:00.760 ⇒ 00:13:01.450 Samuel Roberts: I don’t think that happens.
137 00:13:01.450 ⇒ 00:13:15.440 Pranav: I think it’s because this, how we have it presently constructed, it could also be a manual process, right? It doesn’t necessarily depend on, the automations. So that was kind of by design, like…
138 00:13:16.250 ⇒ 00:13:17.390 Samuel Roberts: Yeah, good point.
139 00:13:17.390 ⇒ 00:13:35.570 Pranav: We can show them this new process, and then we can just also say, like, hey, going forward, like, there will be these automations, but, you know, given, you know, if there’s automations that go down, or just, like, in the interim, where the automations are still in progress, like, this process should still be kind of what we use going forward.
140 00:13:38.810 ⇒ 00:13:41.250 Samuel Roberts: Yeah, that’s a good point. I was trying to think.
141 00:13:43.140 ⇒ 00:13:49.980 Samuel Roberts: So we may have to keep that, a lot of transitions kind of thing, and the state diagram in mind.
142 00:13:50.710 ⇒ 00:13:55.809 Samuel Roberts: So maybe simplifying that is a good idea, but I don’t know where that… you know, if we’re doing anything manually, I guess, but…
143 00:13:59.120 ⇒ 00:14:05.389 Pranav: Yeah, there’s a few in there where Janiece won’t be touching, there’s a few that we won’t be touching.
144 00:14:05.390 ⇒ 00:14:09.739 Samuel Roberts: I felt like it was split, yeah, like… I’m not too.
145 00:14:09.740 ⇒ 00:14:10.370 Pranav: Yeah.
146 00:14:10.930 ⇒ 00:14:14.620 Pranav: Like, actually, let’s maybe… since it’s a working session, we might as well…
147 00:14:15.250 ⇒ 00:14:20.370 Pranav: talk through it. Do you want to pull up that, that, graph again?
148 00:14:20.480 ⇒ 00:14:22.219 Pranav: Casey, or whoever’s sharing?
149 00:14:22.740 ⇒ 00:14:24.549 Samuel Roberts: Yeah, Casey, if you could just.
150 00:14:25.530 ⇒ 00:14:26.080 Pranav: Perfect, thanks.
151 00:14:26.080 ⇒ 00:14:29.539 Samuel Roberts: So I tried to… I tried to mess with mermaids so it wouldn’t look quite as…
152 00:14:29.820 ⇒ 00:14:31.629 Samuel Roberts: Like, I don’t know why this arrow…
153 00:14:32.160 ⇒ 00:14:35.769 Samuel Roberts: does this, but who knows? Mermaid’s a little weird.
154 00:14:35.770 ⇒ 00:14:37.820 Pranav: I think it looks great, actually.
155 00:14:37.820 ⇒ 00:14:42.650 Samuel Roberts: Yeah, I was just… I feel like it could have been cleaned up, but there’s certain things about, like, the order you define things in.
156 00:14:42.650 ⇒ 00:14:43.060 Pranav: You know what I mean?
157 00:14:43.060 ⇒ 00:14:46.940 Samuel Roberts: And there’s not a good way to, like, rearrange it, it’s all kind of deterministic that way.
158 00:14:47.180 ⇒ 00:14:49.309 Pranav: Yeah, that one arrow is pretty funny.
159 00:14:49.310 ⇒ 00:14:54.400 Samuel Roberts: Yeah, I was just like, everything else was like, okay, I see there’s overlaps, and there has to be, blah blah blah, but I was just like, what?
160 00:14:54.910 ⇒ 00:15:00.879 Samuel Roberts: Anyway, yeah, so… The idea here was that it would, it would… A ticket would come in.
161 00:15:01.720 ⇒ 00:15:07.849 Samuel Roberts: That would be, obviously, like, this here. Needs info, I think, is where it would be,
162 00:15:08.400 ⇒ 00:15:15.649 Samuel Roberts: the first pass if it doesn’t have a document structure, but I suppose it looks like it could go straight to triage if they have
163 00:15:16.490 ⇒ 00:15:17.620 Samuel Roberts: Like, pretty quickly.
164 00:15:17.620 ⇒ 00:15:18.020 Pranav: So…
165 00:15:18.020 ⇒ 00:15:20.130 Samuel Roberts: document structure? The change?
166 00:15:20.490 ⇒ 00:15:27.949 Pranav: Yeah, I guess that’s my… As it currently stands, right, when a…
167 00:15:28.200 ⇒ 00:15:34.450 Pranav: thumbs-down feedback is given in Andy. That’ll automatically go into triage.
168 00:15:34.590 ⇒ 00:15:35.110 Pranav: Right?
169 00:15:35.110 ⇒ 00:15:46.690 Samuel Roberts: Right, right, okay, okay, yeah, so here’s the thing. Triage as a state in linear, I think, is probably different than what I meant here as a label of where we are in the process with that ticket.
170 00:15:47.090 ⇒ 00:15:47.550 Pranav: Okay.
171 00:15:47.550 ⇒ 00:15:48.140 Samuel Roberts: But maybe that was…
172 00:15:48.140 ⇒ 00:15:49.509 Pranav: So… I probably shouldn’t.
173 00:15:49.650 ⇒ 00:15:53.800 Samuel Roberts: Change that word, maybe. Like, pre-process or something.
174 00:15:54.730 ⇒ 00:16:01.959 Samuel Roberts: Cause you’re right, you’re right, there’s a… there’s a linear triage state, which is different than what I was thinking here, I think.
175 00:16:02.630 ⇒ 00:16:08.419 Pranav: Okay, gotcha. So, yeah, so basically it’ll automatically go into…
176 00:16:09.860 ⇒ 00:16:14.780 Pranav: Yeah, it’ll automatically go into linear,
177 00:16:15.000 ⇒ 00:16:21.140 Pranav: And then we’ll just say, by default, it’ll be in… needs info.
178 00:16:22.770 ⇒ 00:16:28.940 Pranav: Actually, you know, yeah, by default, we’ll say, actually, that it just goes straight to,
179 00:16:30.210 ⇒ 00:16:36.020 Pranav: Do we even need the… I’m trying to think, do we need the triage?
180 00:16:36.300 ⇒ 00:16:36.859 Pranav: Right now.
181 00:16:36.860 ⇒ 00:16:43.720 Samuel Roberts: Probably… not. I think it was just kind of a… A state between…
182 00:16:44.590 ⇒ 00:16:50.779 Samuel Roberts: You know, getting that info, that documented structure, and then running these loops.
183 00:16:51.000 ⇒ 00:16:51.990 Samuel Roberts: Right.
184 00:16:53.200 ⇒ 00:16:53.600 Pranav: Yeah.
185 00:16:53.600 ⇒ 00:16:58.770 Samuel Roberts: Potentially. So, like, there is a state that, like, if a conflict gets flagged or not, it can go back…
186 00:17:00.710 ⇒ 00:17:02.960 Samuel Roberts: Or just be straight up projected, you know?
187 00:17:04.880 ⇒ 00:17:05.540 Pranav: Right.
188 00:17:05.540 ⇒ 00:17:08.299 Samuel Roberts: That is identified, we follow this path.
189 00:17:10.569 ⇒ 00:17:13.009 Samuel Roberts: Or… this path.
190 00:17:15.280 ⇒ 00:17:20.469 Samuel Roberts: So maybe… maybe I didn’t… hmm, right, so I… so you were thinking of batching them, like, once a day, right?
191 00:17:21.150 ⇒ 00:17:23.960 Pranav: Yeah, because they want that 24-hour SLA.
192 00:17:24.500 ⇒ 00:17:30.509 Samuel Roberts: Yeah, so… Approval pending can still be rejected.
193 00:17:31.390 ⇒ 00:17:35.060 Samuel Roberts: But it goes to batch… approved to batch to applying. Yeah, that’s right, okay.
194 00:17:35.850 ⇒ 00:17:43.959 Samuel Roberts: probably… it could go back to triage, it looks like, here, I don’t know why that’s… There, necessarily, but…
195 00:17:44.250 ⇒ 00:17:47.429 Pranav: Yeah, that probably shouldn’t be there. Once it gets bad…
196 00:17:47.430 ⇒ 00:17:49.959 Samuel Roberts: It’s approved, it’s batched, it should be ready to go.
197 00:17:50.130 ⇒ 00:17:51.330 Samuel Roberts: We can, we can…
198 00:17:52.820 ⇒ 00:17:58.549 Pranav: Yeah. Yeah, once it gets rejected, then it can go… then it goes back to triage, which is right.
199 00:17:59.110 ⇒ 00:17:59.700 Samuel Roberts: Yeah.
200 00:18:00.460 ⇒ 00:18:01.070 Pranav: Yeah.
201 00:18:01.260 ⇒ 00:18:06.650 Pranav: So it says it’s going through… let’s see… Where is that? Actually.
202 00:18:07.260 ⇒ 00:18:08.090 Samuel Roberts: Which one?
203 00:18:08.740 ⇒ 00:18:15.510 Pranav: If it gets rejected… oh, actually, if it gets rejected… no, it actually shouldn’t go back to triage.
204 00:18:15.880 ⇒ 00:18:18.050 Samuel Roberts: It goes to closed, I think, only.
205 00:18:18.050 ⇒ 00:18:23.500 Pranav: Yeah, it should just go to close, that makes sense. Now…
206 00:18:23.500 ⇒ 00:18:26.480 Samuel Roberts: We could have a point in the rejected state where maybe we…
207 00:18:26.620 ⇒ 00:18:29.300 Samuel Roberts: Know that it’s gonna go from closed to…
208 00:18:29.510 ⇒ 00:18:33.170 Samuel Roberts: You know, re-redo something, and follow this flow again.
209 00:18:33.820 ⇒ 00:18:37.919 Pranav: Yeah, there could be another… there needs to be…
210 00:18:39.380 ⇒ 00:18:45.509 Pranav: I think that’s where the needs info actually comes in. The needs info doesn’t come in… Yeah.
211 00:18:46.920 ⇒ 00:18:49.030 Pranav: So it can be… So, are we assuming that the…
212 00:18:50.090 ⇒ 00:18:50.900 Samuel Roberts: Sorry, what?
213 00:18:51.340 ⇒ 00:18:56.770 Pranav: It can be rejected for multiple reasons, right. Or…
214 00:18:57.730 ⇒ 00:19:01.319 Pranav: I think rejection, how we should define rejection, right?
215 00:19:01.620 ⇒ 00:19:04.310 Pranav: And actually, let’s just look at what we’ve defined it above.
216 00:19:04.310 ⇒ 00:19:06.549 Samuel Roberts: I was just pulling up the doc now, yeah.
217 00:19:06.920 ⇒ 00:19:08.909 Pranav: So right here,
218 00:19:08.910 ⇒ 00:19:10.130 Samuel Roberts: I could scroll a little bit.
219 00:19:11.550 ⇒ 00:19:17.180 Pranav: Actually, is rejection… yeah, rejected proposal, reviewer, or approve or reject action?
220 00:19:24.870 ⇒ 00:19:26.270 Pranav: Okay…
221 00:19:29.490 ⇒ 00:19:31.300 Pranav: Purpose.
222 00:19:31.300 ⇒ 00:19:32.800 Samuel Roberts: So maybe, maybe it just didn’t land.
223 00:19:32.800 ⇒ 00:19:42.670 Pranav: proposal is right, I think what we should probably define it as, though, is…
224 00:19:46.290 ⇒ 00:19:51.099 Pranav: Things should only be rejected if… Yeah, I mean…
225 00:19:52.680 ⇒ 00:19:56.750 Pranav: Because we don’t want the automations to be the one rejecting, because at the end of the day, if they want to… Right.
226 00:19:56.750 ⇒ 00:19:57.520 Samuel Roberts: Right, right, so…
227 00:19:57.520 ⇒ 00:20:07.370 Pranav: If the AI says it’s duplicate, and then they confirm that it’s not duplicate, then we’re going to add it in there. Well, we shouldn’t, actually, because then Andy’s not going to be able to…
228 00:20:07.570 ⇒ 00:20:10.019 Pranav: Retrieve that information properly.
229 00:20:11.980 ⇒ 00:20:21.719 Samuel Roberts: So maybe we need another… so maybe rejected, or, like, so here’s, like, apply failed, where, like, it’s another form of, like, rejection, theoretically. But, like, maybe we need another…
230 00:20:22.130 ⇒ 00:20:28.659 Pranav: Oh, I thought applied field just meant, like, if we’re using, like, the GWS CLI.
231 00:20:28.660 ⇒ 00:20:33.009 Samuel Roberts: Yeah, yeah, yeah, I’m just saying, like, I’m trying to think of states like failure states, basically, right?
232 00:20:33.010 ⇒ 00:20:34.000 Pranav: Okay. Yeah.
233 00:20:34.000 ⇒ 00:20:39.000 Samuel Roberts: like, an applied failed is, like, this didn’t go through, we need to look at it. Rejected could just be…
234 00:20:39.190 ⇒ 00:20:47.549 Samuel Roberts: it was not properly formatted, it wasn’t… it got flagged for one of these reasons.
235 00:20:47.960 ⇒ 00:20:49.120 Samuel Roberts: Yeah.
236 00:20:51.230 ⇒ 00:20:53.240 Samuel Roberts: It, like, didn’t make it to approval.
237 00:20:54.980 ⇒ 00:20:57.869 Samuel Roberts: Casey, can you scroll back down to the doc… or the, diagram real quick?
238 00:20:58.640 ⇒ 00:20:59.809 Samuel Roberts: Yeah, cause, like.
239 00:21:00.010 ⇒ 00:21:05.480 Samuel Roberts: If it flags a duplicate, it goes automatically to rejected, but you’re thinking we should just have it go to…
240 00:21:05.810 ⇒ 00:21:08.699 Samuel Roberts: Approval pending, maybe, and then approval pending could have rejected.
241 00:21:08.900 ⇒ 00:21:09.989 Samuel Roberts: Like, these don’t…
242 00:21:10.470 ⇒ 00:21:14.240 Samuel Roberts: You just flag what’s going on, and then we have a human in the loop to do it?
243 00:21:15.340 ⇒ 00:21:30.050 Pranav: Yeah, yeah, I think, okay. No, no, actually, I don’t think I agree with that, because, yeah, I’m kind of thinking about this in real time, because we should not be allowing information into the central dock that isn’t going to be retrieved properly.
244 00:21:30.710 ⇒ 00:21:39.510 Samuel Roberts: I agree, I agree. I think it’s fine to go to this rejected, and then maybe we just don’t go straight to closed, we have another, like…
245 00:21:39.930 ⇒ 00:21:51.920 Samuel Roberts: another… Pass, basically, where it can be finally closed, Or it can be… Like, reopened, effectively?
246 00:21:53.410 ⇒ 00:21:58.679 Pranav: Yeah, there’s probably, like, a word for that, right? Like, if it gets rejected and you want to, like, appeal it, or something?
247 00:21:58.780 ⇒ 00:22:04.239 Samuel Roberts: Right, yeah, exactly, yeah, we could do something like that. I don’t think that was really considered in here, but I think the.
248 00:22:04.240 ⇒ 00:22:04.640 Pranav: Yeah, positive.
249 00:22:04.640 ⇒ 00:22:06.450 Samuel Roberts: was, if you scroll back up, Casey…
250 00:22:07.780 ⇒ 00:22:08.610 Casie Aviles: Sure.
251 00:22:08.610 ⇒ 00:22:16.029 Samuel Roberts: I think the process was, yeah, if it’s rejected, the exit condition is reopen only, or here it is, with a new structured triage comment revision.
252 00:22:17.900 ⇒ 00:22:19.259 Pranav: Yeah, yeah.
253 00:22:19.610 ⇒ 00:22:23.080 Samuel Roberts: I thought that might not be quite captured by the document, but I think that was the idea.
254 00:22:23.980 ⇒ 00:22:24.500 Samuel Roberts: Or the…
255 00:22:24.500 ⇒ 00:22:24.880 Pranav: The diagram.
256 00:22:24.880 ⇒ 00:22:25.699 Samuel Roberts: Instagram, I should say.
257 00:22:26.960 ⇒ 00:22:30.909 Samuel Roberts: Or it’s just like, yeah, no, this is a bad one, we’ll get rid of it, you know?
258 00:22:31.960 ⇒ 00:22:42.269 Pranav: Yeah, I like that, and at the end of the day, what is gonna come from this whole process is, once we make the changes every single day, we’re going to then…
259 00:22:42.440 ⇒ 00:22:47.780 Pranav: create a memo, which then shows, hey, these have been applied, these have been rejected.
260 00:22:47.780 ⇒ 00:22:49.170 Samuel Roberts: Yeah. Yeah.
261 00:22:49.170 ⇒ 00:22:53.580 Pranav: And so, with that rejection memo, they’ll get to know, hey, like.
262 00:22:53.690 ⇒ 00:23:01.759 Pranav: these changes which we wanted in there got rejected, so then they can go through the appeal process, like, I’ll get a message from Janiece, probably, or one of the trainers.
263 00:23:02.170 ⇒ 00:23:06.060 Pranav: And then we can… We can take care of that.
264 00:23:06.410 ⇒ 00:23:08.870 Samuel Roberts: Yeah, okay, so we want to add, like, an appeal…
265 00:23:09.080 ⇒ 00:23:12.559 Samuel Roberts: state that might be, kind of, essentially
266 00:23:12.880 ⇒ 00:23:14.870 Samuel Roberts: Back up to, like, this level.
267 00:23:17.140 ⇒ 00:23:17.660 Samuel Roberts: Yay.
268 00:23:17.660 ⇒ 00:23:20.179 Pranav: Yeah, yeah, I think that would be great, and then…
269 00:23:20.310 ⇒ 00:23:25.220 Pranav: Yeah, if… if you can add that, that would be… that would be good.
270 00:23:25.450 ⇒ 00:23:31.079 Pranav: And then, yeah, I will also… I know, like, the last couple days, I’ve just been, like, heads down on Eden. Today, I have a lot.
271 00:23:31.080 ⇒ 00:23:31.489 Samuel Roberts: All good.
272 00:23:32.110 ⇒ 00:23:32.650 Pranav: To cut.
273 00:23:32.650 ⇒ 00:23:33.530 Samuel Roberts: Yeah, okay, cool.
274 00:23:34.030 ⇒ 00:23:44.819 Pranav: ABC stuff, so I will, I’ll put some more eyes on this. Yeah, Sam, after you make that update, like, I will just sit down for, like, at least 30 minutes and just kinda…
275 00:23:45.130 ⇒ 00:23:48.800 Pranav: Just, like, really hack at this, and then that should be an.
276 00:23:48.800 ⇒ 00:23:49.220 Samuel Roberts: That’d.
277 00:23:49.220 ⇒ 00:23:50.820 Pranav: To finalize this.
278 00:23:51.630 ⇒ 00:23:52.300 Samuel Roberts: Okay.
279 00:23:53.030 ⇒ 00:23:56.670 Samuel Roberts: Yeah, so I’ll add a… Yep.
280 00:23:56.670 ⇒ 00:23:57.270 Pranav: Go ahead.
281 00:23:57.750 ⇒ 00:24:04.440 Samuel Roberts: I was gonna add the appeal, and then I’ll kind of flow it the right way, and I’ll send you the diagram and stuff before I, like, really commit stuff, maybe.
282 00:24:04.920 ⇒ 00:24:05.940 Pranav: Perfect, yeah.
283 00:24:06.350 ⇒ 00:24:17.950 Pranav: Yeah, I need to hop soon, because I have to finish some last, like, last few things for Eden, actually, before my call with him. But yeah, Casey, I just want to make sure, like, you feel good about, Friday, because I know, like.
284 00:24:18.300 ⇒ 00:24:26.099 Pranav: there is quite a few things on your plate, but I feel like it’s pretty straightforward, specifically with the scripts, I think…
285 00:24:26.300 ⇒ 00:24:30.530 Pranav: They’re…
286 00:24:31.200 ⇒ 00:24:41.710 Pranav: Yeah, with the scripts, I think, you know, Cursor’s gonna be able to help a lot with, detecting where to add these things into the document.
287 00:24:42.350 ⇒ 00:24:43.410 Pranav: Sam.
288 00:24:43.680 ⇒ 00:24:54.890 Pranav: Well, as I say that too, like, there’s a little bit of complexity here, because we’re essentially using the embeddings from Andy, and then just retrieving them with a different system prompt, I’m assuming.
289 00:24:56.850 ⇒ 00:24:57.820 Pranav: Right?
290 00:24:58.920 ⇒ 00:25:01.919 Samuel Roberts: Well, okay, so there’s, there’s a few options there.
291 00:25:04.380 ⇒ 00:25:07.270 Samuel Roberts: Part of the reason… so there’s… there’s a lot more…
292 00:25:07.970 ⇒ 00:25:14.690 Samuel Roberts: chatter, I guess, overall, about, like, agentic retrieval now. I don’t know if you’ve been following, like…
293 00:25:14.910 ⇒ 00:25:21.850 Samuel Roberts: RAG versus, like, the way Cursor does it with, like, bash, kind of, you know, LS, cat, grep.
294 00:25:21.980 ⇒ 00:25:22.580 Samuel Roberts: All that sort.
295 00:25:22.580 ⇒ 00:25:24.090 Pranav: Yeah, right.
296 00:25:24.090 ⇒ 00:25:36.749 Samuel Roberts: So, I was… I was debating for a while if we wanted to move Andy to that sort of thing, because we have this document, there’s a lot of context and text that, you know, might be either overlooked by the vectorization, or…
297 00:25:36.750 ⇒ 00:25:37.250 Pranav: Yeah.
298 00:25:37.250 ⇒ 00:25:43.840 Samuel Roberts: you know, confused by the vectorization, but the real benefit of the RAG is the speed, and that’s huge.
299 00:25:44.420 ⇒ 00:25:47.519 Samuel Roberts: For this, the speed is less critical, so I was…
300 00:25:47.520 ⇒ 00:25:48.270 Pranav: That’s not like that.
301 00:25:48.590 ⇒ 00:25:52.090 Samuel Roberts: Mulling over if we want to try to do something where we just, like.
302 00:25:52.430 ⇒ 00:25:55.899 Samuel Roberts: Have these things represented as some, like, markdown?
303 00:25:56.040 ⇒ 00:26:03.559 Samuel Roberts: Or however we want to access that, and we can have a little bit of a maybe slower, more agentic kind of search, kind of the way Cursor would operate.
304 00:26:06.070 ⇒ 00:26:07.530 Pranav: Yeah, so it would still be able.
305 00:26:07.530 ⇒ 00:26:09.900 Samuel Roberts: I don’t really know which one’s better, but… sorry, go ahead.
306 00:26:10.230 ⇒ 00:26:13.949 Pranav: You would still be able to search, like, semantically, not just on keyword.
307 00:26:14.720 ⇒ 00:26:27.279 Samuel Roberts: That’s sort of my thought. So we could either… a combination of the rag to identify things and then search broader, or just kind of imagine cursor being like, I know these files, I need to find where this type of thing, like.
308 00:26:27.280 ⇒ 00:26:37.359 Samuel Roberts: semantically exists, keyword exists, whatever, in the document without the embeddings. So there’s kind of two different ways we could go about it. This is just something very recently that I’ve been thinking about after I watched somewhere.
309 00:26:37.370 ⇒ 00:26:38.790 Samuel Roberts: Talks about it.
310 00:26:38.940 ⇒ 00:26:40.480 Pranav: I mean, I love that idea.
311 00:26:41.190 ⇒ 00:26:51.070 Samuel Roberts: Yeah, I’m not sure if it’s performance-wise, like, where that gets us, if it’s better or worse. I just know people have been talking a lot more about it, especially when it comes to code stuff, which is… this is not, but…
312 00:26:51.070 ⇒ 00:26:51.920 Pranav: Yeah.
313 00:26:52.420 ⇒ 00:26:59.299 Samuel Roberts: You know, code stuff like you miss imports, and you miss exports, and things like that if you’re just using embedded chunks.
314 00:26:59.740 ⇒ 00:27:00.260 Samuel Roberts: Got it.
315 00:27:00.260 ⇒ 00:27:00.700 Pranav: Gotcha.
316 00:27:00.700 ⇒ 00:27:06.709 Samuel Roberts: we don’t have a lot of, like, cross-linking that way, I feel like, so I’m less worried about that, but I think it’s an interesting path to go down.
317 00:27:06.900 ⇒ 00:27:15.129 Samuel Roberts: But we already have the rag set up, so, like, there’s also that, you know what I mean? We don’t have to just rebuild something. But if we see performance issues, that’s another option, maybe?
318 00:27:15.340 ⇒ 00:27:20.029 Samuel Roberts: The speed is less critical here, like, this can just run in the background and do its thing.
319 00:27:20.140 ⇒ 00:27:23.060 Samuel Roberts: We don’t need 5-second, you know, responses.
320 00:27:23.060 ⇒ 00:27:31.679 Pranav: I think it’s cool, like, if you’re noticing, like, that, you know, I mean, cursor, obviously, like, we use this every single day, and we notice that it does a pretty good job of…
321 00:27:31.790 ⇒ 00:27:42.050 Pranav: figuring out what is relevant, what is not. And so that’s kind of the system that Kurtzer’s using. For something at our scale, which is, like, you know, very specific type of information, you know.
322 00:27:42.400 ⇒ 00:27:43.530 Pranav: Less fraud.
323 00:27:43.530 ⇒ 00:27:46.829 Samuel Roberts: specific set of docs, you know, like, it’s not…
324 00:27:46.830 ⇒ 00:27:47.520 Pranav: Yeah.
325 00:27:47.840 ⇒ 00:27:49.380 Pranav: So, I’m…
326 00:27:49.380 ⇒ 00:27:49.760 Samuel Roberts: What’s up?
327 00:27:49.760 ⇒ 00:27:51.849 Pranav: I’m super happy to, like…
328 00:27:52.070 ⇒ 00:27:57.370 Pranav: And if you’re also saying, yeah, it could actually make things a little bit lighter weight.
329 00:27:57.370 ⇒ 00:28:03.279 Samuel Roberts: Possibly, yeah. This, though, I think is for, like, the next milestone, though, when we do the detection kind of logic.
330 00:28:03.390 ⇒ 00:28:08.070 Samuel Roberts: So I’m not… we have a little bit of time to sort out which way we want to handle that logic, I think.
331 00:28:08.460 ⇒ 00:28:12.020 Pranav: Would it not also work similarly, though, for figuring out
332 00:28:12.320 ⇒ 00:28:14.679 Pranav: Where we want to place it within the document.
333 00:28:15.940 ⇒ 00:28:22.210 Samuel Roberts: You’re right, you’re right. I was thinking about, like, duplicate detection, which is also part of the same thing, but yes, where in the document this would even be.
334 00:28:22.670 ⇒ 00:28:23.110 Pranav: Yeah.
335 00:28:23.110 ⇒ 00:28:25.069 Samuel Roberts: on the input we have, you’re absolutely right.
336 00:28:25.070 ⇒ 00:28:27.679 Pranav: That’s for this… that’s for this milestone, that’s for this…
337 00:28:27.680 ⇒ 00:28:31.800 Samuel Roberts: Yeah, you’re right, you’re right. I was thinking duplicate conflict stuff, which is also.
338 00:28:31.800 ⇒ 00:28:32.240 Pranav: Oh.
339 00:28:32.240 ⇒ 00:28:36.789 Samuel Roberts: search, but you’re right, you’re right, I’m thinking, yeah, for the where this could go.
340 00:28:36.970 ⇒ 00:28:41.449 Samuel Roberts: you know, if… so I think… what I’m thinking is this milestone is just, like.
341 00:28:41.880 ⇒ 00:28:45.729 Samuel Roberts: We’re just adding things, and then the next ton is we will…
342 00:28:46.040 ⇒ 00:28:50.280 Samuel Roberts: Add if, you know, then we will do the duplicate conflict kind of check.
343 00:28:51.380 ⇒ 00:28:54.579 Pranav: Yeah, they both are actually, like, once you do one.
344 00:28:54.910 ⇒ 00:28:59.309 Pranav: The duplicate conflict one is gonna be really simple to just follow up with.
345 00:28:59.310 ⇒ 00:29:05.970 Samuel Roberts: That’s what I mean. I think… I was thinking just those ones, I kind of wasn’t thinking the first pass when I was just talking, but you’re absolutely right. So I think…
346 00:29:05.970 ⇒ 00:29:13.869 Pranav: Okay, so then I think what would be good to do today is let’s, let’s give Casey some clarity on which approach he should take.
347 00:29:14.000 ⇒ 00:29:18.550 Pranav: And so maybe, Sam, that means, like, you kind of…
348 00:29:19.100 ⇒ 00:29:32.280 Pranav: take one of the MD… like, turn the… let’s say, like, that Lawn Central doc into an MD file, or whatever formatting that you need, and then just do, like, kind of just, like, a POC version of, like, what you just described.
349 00:29:32.280 ⇒ 00:29:36.880 Samuel Roberts: Yeah, we might not even need the MD if we can use the CLI, the GWS CLI for that. Like, that might be.
350 00:29:36.880 ⇒ 00:29:38.269 Pranav: That would be a little more detail.
351 00:29:38.880 ⇒ 00:29:47.229 Samuel Roberts: So I can play with that a little bit, and see if that’s, like, functionally working as well as I anticipate it might.
352 00:29:47.230 ⇒ 00:29:48.580 Pranav: Yeah. Try to do it.
353 00:29:48.580 ⇒ 00:29:49.570 Samuel Roberts: drag to fall back on.
354 00:29:49.570 ⇒ 00:29:52.240 Pranav: for today? Like, for today, maybe if…
355 00:29:52.240 ⇒ 00:29:52.710 Samuel Roberts: Yeah, yeah.
356 00:29:53.030 ⇒ 00:29:53.580 Samuel Roberts: Yeah.
357 00:29:53.580 ⇒ 00:30:00.699 Pranav: even use the CLI, like, because you can just, like, export it as an MD. Right now, we just want to make sure that, hey, this
358 00:30:00.770 ⇒ 00:30:16.529 Pranav: this, this technical approach is going to work, and it’s going to meet our needs. Yeah. And then from there, Casey can work on, okay, how is this going to work in practice? Because, yeah, with this batch system, we’re gonna need to get the latest version of each document.
359 00:30:16.530 ⇒ 00:30:35.900 Pranav: So, at 2PM, close the batch. At that point, that should then trigger, I guess, like, an export of the file, or, you know, at that point, Casey can also look into, hey, maybe I can just use the GWS CLI to pull the document. But that’s for Casey… that’s for you and Casey to kind of map out after you.
360 00:30:35.900 ⇒ 00:30:37.759 Samuel Roberts: Definitely, yeah, we can coordinate on that.
361 00:30:38.690 ⇒ 00:30:39.140 Pranav: Yeah.
362 00:30:39.140 ⇒ 00:30:41.430 Samuel Roberts: There was something else I was gonna say about that.
363 00:30:41.430 ⇒ 00:30:43.080 Pranav: No, that sounds… that sounds awesome.
364 00:30:43.560 ⇒ 00:30:49.879 Samuel Roberts: Yeah, my… I mean, the other side of it is, like, we have the rag, so if that… if this becomes too technically, like.
365 00:30:50.510 ⇒ 00:30:53.390 Samuel Roberts: Different, and not worth the effort right now.
366 00:30:53.550 ⇒ 00:31:01.929 Samuel Roberts: Falling back to that is a good alternative, and we can always enhance later if it’s, you know, not quite up to what we want.
367 00:31:02.050 ⇒ 00:31:09.789 Samuel Roberts: I was just thinking, because I realized we don’t need the speed here from the rag, which is really the main benefit there. But we already have it, so that’s a benefit too, you know what I mean?
368 00:31:10.930 ⇒ 00:31:11.960 Pranav: Yeah.
369 00:31:12.200 ⇒ 00:31:16.730 Pranav: I do like what you’re saying, though.
370 00:31:17.270 ⇒ 00:31:33.319 Pranav: Especially since that, like, we know that Cursor does a good job of, like, not necessarily retrieval in the same way RAG does retrieval, but retrieving the information that we need. Yeah. POC that today, just to confirm that it works well for this.
371 00:31:33.540 ⇒ 00:31:37.130 Samuel Roberts: Yeah, no, that’s exactly what I’m thinking. I’ll try to get a quick, like…
372 00:31:37.530 ⇒ 00:31:41.510 Samuel Roberts: Like, agent running to see if it can handle that, and then,
373 00:31:41.620 ⇒ 00:31:47.419 Samuel Roberts: That can inform how we move forward with the, which one is that? Probably the…
374 00:31:49.150 ⇒ 00:31:51.130 Pranav: Build heading resolution.
375 00:31:51.470 ⇒ 00:31:58.060 Samuel Roberts: Yeah, why don’t want my drawing thing on? But yeah, okay, I’ll take a look at that then. Casey, can you click into 7.1 real quick?
376 00:31:59.860 ⇒ 00:32:04.150 Pranav: Yeah, and then if that’s the case, then Casey doesn’t need to work on 4.2.
377 00:32:13.190 ⇒ 00:32:16.040 Samuel Roberts: Yeah, Casey, can you just assign this to me for now?
378 00:32:16.770 ⇒ 00:32:21.710 Samuel Roberts: So I can… I’ll make some comments on it, and then we can toss it back over.
379 00:32:22.820 ⇒ 00:32:25.930 Pranav: Perfect. And I think 4.2, that’s, like, creating the.
380 00:32:25.930 ⇒ 00:32:26.400 Samuel Roberts: Yeah, I was gonna say…
381 00:32:26.400 ⇒ 00:32:26.950 Pranav: Yeah.
382 00:32:27.540 ⇒ 00:32:28.630 Samuel Roberts: Click on that one, too.
383 00:32:31.110 ⇒ 00:32:38.460 Pranav: Casey, I don’t know if you’re already going to be working on this today, or what your plan was for today, and, like, what you would get to.
384 00:32:39.150 ⇒ 00:32:40.279 Pranav: So, yeah, just…
385 00:32:40.650 ⇒ 00:32:45.569 Pranav: I think, just given, kind of, like, the tight deadline, let’s just make sure that, like, nobody’s blocked.
386 00:32:45.780 ⇒ 00:32:46.580 Pranav: Yeah.
387 00:32:46.580 ⇒ 00:32:47.660 Samuel Roberts: Yeah, no, I hear you.
388 00:32:47.660 ⇒ 00:32:59.659 Pranav: Yeah, and so, Casey, if that means, like, you work more closely with Sam on the POC, just so you can assist him in getting the, you know, the technical approach defined quicker, then…
389 00:33:00.330 ⇒ 00:33:03.900 Samuel Roberts: Yeah, actually, sign this one to me too, Casey, because this is definitely, like.
390 00:33:04.790 ⇒ 00:33:05.310 Casie Aviles: Unique.
391 00:33:05.310 ⇒ 00:33:09.369 Samuel Roberts: Might… might change if I, if they’re PSC’d as well.
392 00:33:09.560 ⇒ 00:33:14.500 Casie Aviles: Yeah, since I was wondering, like, these sound very similar to, like, what we already.
393 00:33:14.500 ⇒ 00:33:22.270 Samuel Roberts: Yeah, well, these were… so what happened here was they were one ticket that was just CDC4, and then it was, like, too many points, basically.
394 00:33:22.390 ⇒ 00:33:33.850 Samuel Roberts: And so, I had it break it down, and then they… so they’re related, but I just didn’t want them as… I wanted separate tickets for them, basically. So, what… what’s been going on with this one? Can you click into there?
395 00:33:35.820 ⇒ 00:33:42.709 Casie Aviles: Yeah, like, this was… I remember, like I mentioned, that we already have a fetcher, right, for…
396 00:33:42.710 ⇒ 00:33:48.930 Samuel Roberts: That’s right, we do have it, yeah. So I think this is where I actually want to test out if, which way we want to actually do it.
397 00:33:50.580 ⇒ 00:33:51.230 Casie Aviles: Okay.
398 00:33:53.690 ⇒ 00:33:54.240 Samuel Roberts: Okay.
399 00:33:57.840 ⇒ 00:34:00.850 Samuel Roberts: Yeah, I guess…
400 00:34:01.300 ⇒ 00:34:06.909 Casie Aviles: In terms of, like, implementation, I did not really implement anything for this one.
401 00:34:07.270 ⇒ 00:34:13.700 Samuel Roberts: That’s fine, I think this is where I need to see if the GWS CLI is good enough, or if we want to fall back to something else.
402 00:34:14.620 ⇒ 00:34:15.310 Casie Aviles: Okay.
403 00:34:15.880 ⇒ 00:34:24.109 Samuel Roberts: So yeah, maybe pop that one to me real quick, and then I’ll pop them back when we need, and then, that actually frees up to do more of the…
404 00:34:25.940 ⇒ 00:34:33.420 Samuel Roberts: Did we close out the other… wait, what is… Edge cases, build preview…
405 00:34:37.510 ⇒ 00:34:43.909 Samuel Roberts: What was… oh, that was right, that was 4, yes, okay. Once we have the state defined, okay, I get it now.
406 00:34:45.030 ⇒ 00:34:45.870 Samuel Roberts: Okay.
407 00:34:46.409 ⇒ 00:34:49.520 Samuel Roberts: We may need to reflow these a little bit, then.
408 00:34:49.980 ⇒ 00:34:52.059 Samuel Roberts: Pranav, when you have to run…
409 00:34:52.610 ⇒ 00:34:54.880 Pranav: Yeah, I think I need a hop right now.
410 00:34:54.880 ⇒ 00:35:03.369 Samuel Roberts: Okay, okay, I just wanted to make sure, okay, I didn’t want you missing whatever it was. Alright, cool. Yeah, Casey, let’s talk through how we can actually, like, do this.
411 00:35:03.870 ⇒ 00:35:05.940 Samuel Roberts: I have to go back to… yep.
412 00:35:06.890 ⇒ 00:35:12.249 Samuel Roberts: Scroll down here… I’m wondering if we can pull in…
413 00:35:12.790 ⇒ 00:35:15.110 Samuel Roberts: Can you click into, this one?
414 00:35:16.630 ⇒ 00:35:18.829 Samuel Roberts: Is there a way to… yeah, Susie, yeah, thank you.
415 00:35:19.460 ⇒ 00:35:23.300 Samuel Roberts: I just want to say, I’m just trying to remember what the order is, because the numbers are all…
416 00:35:23.830 ⇒ 00:35:26.959 Samuel Roberts: These numbers and these numbers aren’t necessarily, like, an order of…
417 00:35:27.880 ⇒ 00:35:28.380 Casie Aviles: Yeah.
418 00:35:28.380 ⇒ 00:35:33.960 Samuel Roberts: whatnot. And Linear does not have a great view of that, either. I want, like, a good Gantt flow, you know?
419 00:35:36.030 ⇒ 00:35:40.610 Casie Aviles: Yeah, that’s true. I was also kind of looking at blocking and…
420 00:35:41.110 ⇒ 00:35:47.880 Samuel Roberts: Yeah, well, that was the idea, was that we were getting off Instagant, because it didn’t… it was just very manual, and trying to.
421 00:35:47.880 ⇒ 00:35:48.270 Casie Aviles: to get something.
422 00:35:48.270 ⇒ 00:35:54.440 Samuel Roberts: out of linear, and that’s why all these blockers were set up and everything, but I don’t think linear just has a good way to view that, which is annoying.
423 00:35:54.620 ⇒ 00:36:01.499 Samuel Roberts: Okay, can we… let me actually… just… let’s jump back to,
424 00:36:02.000 ⇒ 00:36:04.809 Samuel Roberts: There, yeah, go back one.
425 00:36:05.380 ⇒ 00:36:07.620 Samuel Roberts: And then go to issues here.
426 00:36:10.800 ⇒ 00:36:16.029 Samuel Roberts: And then, I guess clear the filters real quick?
427 00:36:17.290 ⇒ 00:36:21.160 Samuel Roberts: I just want to see everything. Yeah, okay, I was trying to… I was missing the in-progress stuff here, that’s what I was thinking.
428 00:36:21.300 ⇒ 00:36:26.800 Samuel Roberts: This is what you’re working on now, that’s gonna be a big one, right? That blocks… what?
429 00:36:27.020 ⇒ 00:36:28.700 Samuel Roberts: 3.2…
430 00:36:29.660 ⇒ 00:36:30.819 Casie Aviles: Yeah. Okay.
431 00:36:30.820 ⇒ 00:36:34.519 Samuel Roberts: There’s plenty in CDC3 to do, I think, still, right?
432 00:36:35.490 ⇒ 00:36:36.330 Samuel Roberts: like…
433 00:36:36.690 ⇒ 00:36:38.340 Casie Aviles: CDC…
434 00:36:38.530 ⇒ 00:36:42.270 Samuel Roberts: 3.1 and 3. This blocks 3.2, which is the actual, like…
435 00:36:42.490 ⇒ 00:36:46.870 Casie Aviles: Yeah, yeah. I plan to work on this today as well.
436 00:36:46.870 ⇒ 00:36:48.189 Samuel Roberts: Okay, cool.
437 00:36:48.760 ⇒ 00:36:55.969 Samuel Roberts: That’s what I… I just wanted… I couldn’t… I couldn’t see that one in the other view, and I was like, I knew there was something else here for, doing all that. Okay.
438 00:36:56.220 ⇒ 00:37:03.910 Samuel Roberts: So I’m gonna say I’ll take a look at those other ones and get some clarity on which way we want to go, and hopefully have a little bit of progress there in the POC.
439 00:37:04.340 ⇒ 00:37:11.500 Samuel Roberts: Then, yeah, if you can keep building the, like, flow here.
440 00:37:11.710 ⇒ 00:37:14.189 Samuel Roberts: That we can then hook into with the…
441 00:37:14.340 ⇒ 00:37:18.740 Samuel Roberts: either the CLI, or the API, or just the already embedded stuff.
442 00:37:19.040 ⇒ 00:37:22.740 Samuel Roberts: I’ll have a little bit of clarity on that in a little bit.
443 00:37:23.120 ⇒ 00:37:23.670 Casie Aviles: Okay.
444 00:37:23.670 ⇒ 00:37:24.800 Samuel Roberts: Oh, shit.
445 00:37:24.800 ⇒ 00:37:26.049 Casie Aviles: What’s that? Sorry.
446 00:37:26.650 ⇒ 00:37:27.430 Samuel Roberts: Sorry, what?
447 00:37:27.430 ⇒ 00:37:30.409 Casie Aviles: Which tool was that that you were referring to again?
448 00:37:31.300 ⇒ 00:37:34.130 Samuel Roberts: the… the GWS CLI?
449 00:37:34.710 ⇒ 00:37:35.580 Casie Aviles: Oh, okay.
450 00:37:35.580 ⇒ 00:37:36.200 Samuel Roberts: You mean?
451 00:37:36.320 ⇒ 00:37:43.399 Samuel Roberts: So, yeah, so okay, let me back up a little bit. So, the thought was, how do we identify where these changes need to be made?
452 00:37:44.090 ⇒ 00:37:44.660 Casie Aviles: Hmm.
453 00:37:45.010 ⇒ 00:37:52.059 Samuel Roberts: Because… the previous tickets, like, the very first ticket was that option A or option B thing, right?
454 00:37:52.300 ⇒ 00:37:53.439 Casie Aviles: Yes, yes.
455 00:37:53.440 ⇒ 00:37:55.360 Samuel Roberts: And so that was, is it…
456 00:37:55.550 ⇒ 00:38:02.160 Samuel Roberts: you know, I… I had the idea that maybe they meet… they have… Suggestion… Access…
457 00:38:03.110 ⇒ 00:38:06.030 Samuel Roberts: And they make suggested changes in…
458 00:38:06.260 ⇒ 00:38:09.740 Samuel Roberts: The doc, and then we process that into a ticket.
459 00:38:09.950 ⇒ 00:38:14.590 Samuel Roberts: But Pranav doesn’t really want them touching the documents directly if they don’t need to, which I think.
460 00:38:14.590 ⇒ 00:38:15.100 Casie Aviles: Okay.
461 00:38:15.100 ⇒ 00:38:25.199 Samuel Roberts: is valid. And so that’s why it’s gonna be a linear ticket. They’re gonna propose a change for, like, what information they see, what they think should be updated, and then we’re gonna take that
462 00:38:25.600 ⇒ 00:38:28.449 Samuel Roberts: Identify where in the document we think that belongs.
463 00:38:29.890 ⇒ 00:38:31.060 Casie Aviles: Oh, okay.
464 00:38:31.390 ⇒ 00:38:36.590 Samuel Roberts: And so the question is, do we use our embedding, like, vectors already?
465 00:38:36.610 ⇒ 00:38:51.419 Samuel Roberts: do we just use the API, or do we use the CLI, the GWS CLI? That’s where I was just talking about, like, the way Cursor does things is very file-based, so even if we give it a fake file system.
466 00:38:52.130 ⇒ 00:38:54.230 Samuel Roberts: That is just this doc… these documents.
467 00:38:54.410 ⇒ 00:39:06.350 Samuel Roberts: it might be able to figure out, like, oh, okay, this one’s labeled pest, you know, it has a lot more context as it’s going through, and that’s why I was saying it might be a little slower, the way cursor is not, you know, 5-second response time, but…
468 00:39:06.890 ⇒ 00:39:10.379 Samuel Roberts: for this process, I’m not sure it matters the way it does for Andy.
469 00:39:12.380 ⇒ 00:39:13.689 Casie Aviles: I see.
470 00:39:13.820 ⇒ 00:39:16.519 Samuel Roberts: And so, I don’t think we had… I had not really…
471 00:39:17.130 ⇒ 00:39:20.709 Samuel Roberts: figured out the best way to do that yet, so I’m gonna take a stab at that real quick.
472 00:39:21.280 ⇒ 00:39:26.709 Samuel Roberts: And then… With the work you’re doing on the actual, like, automation side?
473 00:39:27.340 ⇒ 00:39:33.130 Samuel Roberts: We should be able to get, like, We’ll have their… you know, information.
474 00:39:36.150 ⇒ 00:39:42.800 Samuel Roberts: And then I… I’m gonna do the testing on the other side of that, which is, like, if we have this information, how do I identify a spot where…
475 00:39:43.120 ⇒ 00:39:44.469 Samuel Roberts: We would want to put it.
476 00:39:45.210 ⇒ 00:39:46.859 Samuel Roberts: Or remove, or whatever.
477 00:39:47.080 ⇒ 00:39:48.510 Samuel Roberts: Who is that?
478 00:39:48.510 ⇒ 00:39:49.200 Casie Aviles: Okay.
479 00:39:49.200 ⇒ 00:39:56.900 Samuel Roberts: best with RAG, best with just the pure API endpoint, or best with the CLI, or some hybrid of that, maybe?
480 00:39:58.940 ⇒ 00:40:06.900 Samuel Roberts: So, like, think of it as, like, a piece where it’s, like, you’re gonna get to this point where we have all this information, and it’ll pass it off to an agent.
481 00:40:07.300 ⇒ 00:40:13.170 Samuel Roberts: That, like, either looks through the files, or queries the database, whatever that is, you know?
482 00:40:14.130 ⇒ 00:40:21.229 Casie Aviles: Okay, so it will be… so ideally, we’ll be looking through, like, these restructured docs, right, and then figure out, like, where…
483 00:40:21.480 ⇒ 00:40:22.010 Casie Aviles: Yeah.
484 00:40:22.010 ⇒ 00:40:35.419 Samuel Roberts: So, yeah, exactly. And these restructured docs are also embedded, and we have the API, and we have the CLI, and so there’s just a few options here that I want to kind of hash out now that we’re in the middle of it now, because it wasn’t totally clear what the best one would be yet.
485 00:40:38.800 ⇒ 00:40:39.510 Samuel Roberts: I think…
486 00:40:40.490 ⇒ 00:40:51.609 Samuel Roberts: you know, this is definitely, like, human-in-the-loop stuff, so we’re not gonna, like, just make the changes automatically, but the more we can sort out with the agent, the better. So I will take a stab at that.
487 00:40:52.170 ⇒ 00:40:55.170 Samuel Roberts: And then… Let you know…
488 00:40:55.850 ⇒ 00:41:03.070 Samuel Roberts: where I get with that, and I might even be able to just, like, tackle a few of those if I can get the POC working well, and if it’s not, we’ll have to reevaluate, but…
489 00:41:03.070 ⇒ 00:41:04.370 Casie Aviles: Okay. Clear.
490 00:41:05.140 ⇒ 00:41:09.239 Samuel Roberts: But yeah, I think for you… what is the… can you scroll up a little bit here?
491 00:41:09.630 ⇒ 00:41:12.350 Samuel Roberts: Just to the ticket itself.
492 00:41:12.970 ⇒ 00:41:23.469 Samuel Roberts: Scope, so I just want to say, like, yeah, build listener, that’s perfect. If doc native, auto-create corresponding linear ticket, so we don’t need to worry about that anymore, because we’re not doc-native. Perfect, yep.
493 00:41:23.630 ⇒ 00:41:30.640 Samuel Roberts: And then… emit normalized trigger payload. Yeah, so that this is basically where I’ll hook in the agent, right?
494 00:41:31.070 ⇒ 00:41:34.530 Samuel Roberts: So… Yeah, you’ve built the pipeline.
495 00:41:34.680 ⇒ 00:41:38.530 Samuel Roberts: That will take the ticket, take the structured comment.
496 00:41:39.040 ⇒ 00:41:41.529 Samuel Roberts: And then pass that into…
497 00:41:42.920 ⇒ 00:41:51.259 Samuel Roberts: the agent… I’ll probably put together a master agent that’s, like, cursor-esque, where it can have access to the files, or the CLI, or whatever.
498 00:41:52.440 ⇒ 00:41:53.780 Casie Aviles: Okay, okay.
499 00:41:54.010 ⇒ 00:41:57.940 Casie Aviles: So, the payload, or, like, once… what’s…
500 00:41:58.150 ⇒ 00:42:01.340 Casie Aviles: Linear… or the automation detects this.
501 00:42:01.900 ⇒ 00:42:07.840 Casie Aviles: I’ll pass it off to what I’ll be working on for this ticket, right? 3.2.
502 00:42:07.840 ⇒ 00:42:14.080 Samuel Roberts: Yeah, so this is proposal detection, hold on, let me just read this again, sorry, and analysis, execution, right back once reliable path.
503 00:42:17.640 ⇒ 00:42:19.070 Samuel Roberts: Okay, right, trigger now.
504 00:42:19.330 ⇒ 00:42:22.610 Samuel Roberts: Trigger analysis from detection events. So yeah, so these are… this is…
505 00:42:22.820 ⇒ 00:42:36.859 Samuel Roberts: again, I split up tickets into sub-tickets, which is why these are somewhat related. But yeah, so, like, trigger analysis from detection events, that’s what I will be… the analysis itself, I think, is what I’m working on, but the triggering is that.
506 00:42:39.030 ⇒ 00:42:46.389 Samuel Roberts: this is, I think, a little bit on hold, but we want to have, like, an… a hook for it, right?
507 00:42:46.750 ⇒ 00:42:47.510 Casie Aviles: Okay.
508 00:42:47.960 ⇒ 00:42:56.339 Samuel Roberts: Where we’ll be able to say, like, oh, we identified a duplicate, or oh, we identified a conflict, but we’re not actually doing that logic yet. But if something were to come back and say.
509 00:42:56.920 ⇒ 00:42:58.770 Samuel Roberts: It’s duplicate, where would that go?
510 00:43:00.330 ⇒ 00:43:05.470 Samuel Roberts: And then, if an error comes back, also worry about, like, retries or something.
511 00:43:08.690 ⇒ 00:43:09.670 Samuel Roberts: Okay.
512 00:43:10.500 ⇒ 00:43:11.330 Casie Aviles: Okay.
513 00:43:11.330 ⇒ 00:43:13.220 Samuel Roberts: Yeah, so it runs through automatically.
514 00:43:13.330 ⇒ 00:43:15.530 Samuel Roberts: Results are consistently formatted.
515 00:43:15.840 ⇒ 00:43:23.670 Samuel Roberts: Retry error, end-to-end writeback supports, less than 60 seconds parent expectation. This is where I’m going to try to figure out if I can actually do this fast enough or not.
516 00:43:25.690 ⇒ 00:43:32.729 Casie Aviles: I see. So for 3, at least the 3 CDC3 tickets, these are all from Linear’s side for now, right?
517 00:43:33.510 ⇒ 00:43:39.979 Samuel Roberts: Yeah, because we’re gonna be writing back to linear with the, like, identified, like, here’s the propo… so…
518 00:43:40.290 ⇒ 00:43:41.780 Samuel Roberts: If you think of it from, like, a…
519 00:43:42.130 ⇒ 00:43:45.279 Samuel Roberts: like, Janiece or your vet level, this ticket will come in.
520 00:43:45.430 ⇒ 00:43:51.299 Samuel Roberts: the CSR or the trainer, whoever, will put a proposal in the comment, right?
521 00:43:52.870 ⇒ 00:43:54.360 Samuel Roberts: That then will kick off.
522 00:43:54.510 ⇒ 00:44:08.140 Samuel Roberts: your code here, the agent I’m talking about, and then that’ll come back into Linear and say, hey, this is the proposal we’re changing, we think, based on what is in this ticket, that this part needs to be changed, added, or removed, or whatever.
523 00:44:08.360 ⇒ 00:44:15.140 Samuel Roberts: And then that’s when Yvette or Janiece would do the approval or, you know, rejection, or modification, or whatever.
524 00:44:16.140 ⇒ 00:44:19.460 Samuel Roberts: And then that would get batched into actually doing the change.
525 00:44:20.850 ⇒ 00:44:22.860 Casie Aviles: Oh, okay, okay.
526 00:44:23.520 ⇒ 00:44:28.789 Samuel Roberts: That’s where the human-in-the-loop step needs to be, because otherwise it would just start making changes based on everything they submit, so…
527 00:44:29.470 ⇒ 00:44:33.480 Samuel Roberts: Does that make a little more sense? I’m sorry if this was not super clear from the beginning.
528 00:44:34.500 ⇒ 00:44:46.959 Casie Aviles: Yeah, that makes more sense, yeah, like, I guess I was… at first, like, I was just… I feel like there’s a lot to look at, so I’m not, like, entirely sure what the exact flow is. I think,
529 00:44:47.950 ⇒ 00:44:51.100 Casie Aviles: Like, a flow of, like, the overall…
530 00:44:51.910 ⇒ 00:44:55.280 Samuel Roberts: I should have put something like that together in that document, you’re right. That would have been better.
531 00:44:55.670 ⇒ 00:45:01.510 Casie Aviles: Yeah, I think that would help me as well. I’m a bit more visual for…
532 00:45:01.510 ⇒ 00:45:14.880 Samuel Roberts: Definitely, no, no, no, you’re good. And honestly, what I had was base… I was basing this on the, the SOW, or the, proposal that Pranav had put together for this, so, like, I had a better picture of it based on that. So there is a Notion doc somewhere,
533 00:45:15.900 ⇒ 00:45:17.099 Samuel Roberts: See if I can find it.
534 00:45:17.490 ⇒ 00:45:18.430 Samuel Roberts: That’s not it.
535 00:45:22.680 ⇒ 00:45:30.130 Samuel Roberts: It was… This one. Okay.
536 00:45:38.960 ⇒ 00:45:43.199 Samuel Roberts: So, let me share this here… where’s the chat?
537 00:45:45.280 ⇒ 00:45:51.559 Samuel Roberts: So this is what Pranavid put together. He had basically done… Everything but section…
538 00:45:51.770 ⇒ 00:45:59.430 Samuel Roberts: 5, which is what I kind of put together, so that was where the target process is, and that then informed the tickets that got created.
539 00:45:59.560 ⇒ 00:46:00.390 Samuel Roberts: Okay.
540 00:46:01.140 ⇒ 00:46:08.509 Samuel Roberts: So for context here, this is kind of the new… Proposal kind of style that…
541 00:46:08.960 ⇒ 00:46:15.760 Samuel Roberts: we’re doing for, like, delivery team stuff. So, the idea is that the CSO would put together Proposal…
542 00:46:16.000 ⇒ 00:46:21.239 Samuel Roberts: For the initiatives, the projects under those initiatives, the milestones that they ideally want.
543 00:46:22.030 ⇒ 00:46:23.920 Samuel Roberts: And then…
544 00:46:24.230 ⇒ 00:46:29.110 Samuel Roberts: I came in and said, okay, if we want to get this done, here’s a rough technical approach.
545 00:46:29.300 ⇒ 00:46:37.040 Samuel Roberts: And so you see in here, there’s also the, the transcript stuff at the bottom there, too, under the central doc, so… it’s collapsed. No, sorry, not quite the bottom. Go up a little bit.
546 00:46:37.600 ⇒ 00:46:40.809 Samuel Roberts: Yeah, so this is the department-based insights, that’s the transcript stuff.
547 00:46:43.310 ⇒ 00:46:48.349 Samuel Roberts: So this was just the rough, technical side that I then took and started
548 00:46:49.060 ⇒ 00:46:56.290 Samuel Roberts: chatting with Cursor to put together fleshed-out tickets. But I probably should have given you a higher level view to start, I apologize with that.
549 00:46:57.540 ⇒ 00:47:04.780 Casie Aviles: Yeah, no problem. Yeah, I think that’s… that helps now that we’ve clarified.
550 00:47:05.240 ⇒ 00:47:12.970 Samuel Roberts: Okay, cool, yeah, so I think the basic structure is… and this was unclear, kind of, from the beginning, that’s why the option A or B was kind of there.
551 00:47:13.160 ⇒ 00:47:13.770 Casie Aviles: So…
552 00:47:15.930 ⇒ 00:47:20.019 Samuel Roberts: The plan is, yeah, they… ticket comes in from a thumbs down.
553 00:47:20.340 ⇒ 00:47:23.239 Samuel Roberts: Someone goes in, puts together the comment.
554 00:47:24.310 ⇒ 00:47:27.580 Samuel Roberts: That comment kicks off a job to figure out where it would go.
555 00:47:28.890 ⇒ 00:47:32.350 Samuel Roberts: That comes back, updates the ticket with a proposal, basically.
556 00:47:33.460 ⇒ 00:47:41.149 Samuel Roberts: That proposal gets approved or not, and then if it’s approved, it gets batched, and then on a certain… once a day, they’ll… all those updates run.
557 00:47:41.560 ⇒ 00:47:42.820 Samuel Roberts: And update the docs.
558 00:47:43.870 ⇒ 00:47:44.640 Casie Aviles: Okay.
559 00:47:44.950 ⇒ 00:47:46.760 Casie Aviles: Yeah, that makes more sense.
560 00:47:46.760 ⇒ 00:47:58.510 Samuel Roberts: Okay, yeah, sorry, sorry for the lack of clarity there. I think I was just deep in it and didn’t think about putting together, like, a higher-level document view, but there was also… we weren’t sure which way we were going with it at first until Pranav kind of.
561 00:47:58.820 ⇒ 00:47:59.270 Casie Aviles: Fair enough.
562 00:47:59.270 ⇒ 00:47:59.940 Samuel Roberts: through.
563 00:48:00.300 ⇒ 00:48:00.950 Casie Aviles: Yeah, I’ll agree.
564 00:48:00.950 ⇒ 00:48:08.270 Samuel Roberts: a few things, but, yeah, that’s the rough plan. So I think you’re kind of handling that side of the linear.
565 00:48:08.470 ⇒ 00:48:11.830 Samuel Roberts: Reading the linear, and then ideally writing back to linear.
566 00:48:12.230 ⇒ 00:48:14.420 Casie Aviles: Okay. Yes. So…
567 00:48:14.420 ⇒ 00:48:20.670 Samuel Roberts: I will work on the agent side of it and figure out if the CLI or the embedding or whatever identifies that best.
568 00:48:21.110 ⇒ 00:48:25.190 Samuel Roberts: And then we’ll kind of do, like, a…
569 00:48:25.370 ⇒ 00:48:31.970 Samuel Roberts: You know, that’s where, like, the linear automation will hand off to the agent, and the agent will hand back to the linear automation.
570 00:48:33.280 ⇒ 00:48:34.929 Casie Aviles: Mmm, okay, okay.
571 00:48:34.930 ⇒ 00:48:35.580 Samuel Roberts: Okay.
572 00:48:37.640 ⇒ 00:48:43.059 Casie Aviles: Yes, so that handles, like… yeah, okay, the right back here then, okay.
573 00:48:43.620 ⇒ 00:48:50.610 Samuel Roberts: Yeah, so yeah, it will… it will come up with the proposal, and then, you know, it’ll pass that to, kind of, the linear…
574 00:48:51.320 ⇒ 00:48:56.209 Samuel Roberts: Automation you’ve built out should have, like, a hook or an endpoint or whatever to do that.
575 00:48:56.320 ⇒ 00:48:59.140 Samuel Roberts: Like, respond to the comment or something, probably.
576 00:48:59.810 ⇒ 00:49:00.630 Samuel Roberts: Right.
577 00:49:00.870 ⇒ 00:49:01.790 Casie Aviles: Yes.
578 00:49:01.790 ⇒ 00:49:05.890 Samuel Roberts: I think, yeah, the idea… and I mean, we’ve got that… you have got a few things there already for that, so, like…
579 00:49:06.580 ⇒ 00:49:08.420 Samuel Roberts: I think we’re good to start.
580 00:49:08.680 ⇒ 00:49:11.759 Samuel Roberts: Thinking that through a little bit, and then,
581 00:49:13.010 ⇒ 00:49:17.430 Samuel Roberts: Yeah, so effectively, I think, yeah, someone will make this comment, The agent will kick off.
582 00:49:17.880 ⇒ 00:49:21.070 Samuel Roberts: And leave a comment here with a proposal, basically.
583 00:49:23.100 ⇒ 00:49:27.370 Casie Aviles: Okay. And then, yeah, that would have gone through, like, the…
584 00:49:27.620 ⇒ 00:49:30.140 Casie Aviles: Sent the docs already, and then…
585 00:49:30.140 ⇒ 00:49:39.139 Samuel Roberts: It will have gone through the docs and proposed what changes need to be made, and then once that’s approved, we’ll take that information and make those changes.
586 00:49:39.900 ⇒ 00:49:40.600 Casie Aviles: Okay.
587 00:49:42.340 ⇒ 00:49:48.080 Samuel Roberts: So then that’s the other side of this, is actually getting the changes, but that’s kind of related to the agent I’m putting together now, too.
588 00:49:50.110 ⇒ 00:49:53.049 Casie Aviles: Okay, okay. Yeah, that makes more sense, no?
589 00:49:53.050 ⇒ 00:49:57.199 Samuel Roberts: Okay, yeah, sorry about that. I know I kind of threw a bunch of these tickets at you yesterday, but
590 00:49:58.590 ⇒ 00:50:02.380 Samuel Roberts: I… yeah, I should have had a little more clarity there for you, my higher level.
591 00:50:02.990 ⇒ 00:50:09.520 Casie Aviles: Yeah, all good. Okay, so… I’ll just work on these tickets today, then, and…
592 00:50:10.380 ⇒ 00:50:13.939 Casie Aviles: I think I should be able… I should be close to 3.1.
593 00:50:14.060 ⇒ 00:50:15.020 Samuel Roberts: Okay.
594 00:50:17.580 ⇒ 00:50:18.490 Casie Aviles: Alright, and then…
595 00:50:18.490 ⇒ 00:50:23.770 Samuel Roberts: And I think, yeah, 3.2 is just, like, the triggering, right, basically.
596 00:50:26.880 ⇒ 00:50:34.149 Samuel Roberts: Yeah, the triggering, the proper formatting, and the retries in case something goes wrong.
597 00:50:34.760 ⇒ 00:50:36.319 Samuel Roberts: But… don’t…
598 00:50:37.020 ⇒ 00:50:43.300 Samuel Roberts: stress too much about the retries yet, I guess. Like, we’ll focus on these ones, I guess. Definitely…
599 00:50:43.530 ⇒ 00:50:47.890 Samuel Roberts: think about the retries, but I might have to handle retries at the agent level anyway.
600 00:50:48.130 ⇒ 00:50:51.200 Samuel Roberts: Depending on which way we go with it.
601 00:50:53.440 ⇒ 00:50:54.790 Casie Aviles: Okay, okay.
602 00:50:55.340 ⇒ 00:51:00.440 Casie Aviles: I think if I have more questions, then I’ll just… I’ll just ping you.
603 00:51:00.640 ⇒ 00:51:02.969 Casie Aviles: If there’s still something unclear.
604 00:51:03.590 ⇒ 00:51:05.109 Samuel Roberts: Yeah, and I can, I mean…
605 00:51:05.460 ⇒ 00:51:10.100 Samuel Roberts: I’m trying to think, is it helpful for me to put, like, a flow together, or is this clear enough now?
606 00:51:13.190 ⇒ 00:51:22.610 Casie Aviles: If it’s not too… if it’s not… it’s not gonna take too much time, I think it can be helpful, so I can, like, look back at it.
607 00:51:23.480 ⇒ 00:51:29.760 Samuel Roberts: Yeah, okay. I can try to either draw something out or mermaid something out real quick, and then I’ll get into the agent stuff.
608 00:51:30.560 ⇒ 00:51:33.310 Casie Aviles: Okay, that’s lovely cool. Thank you.
609 00:51:33.310 ⇒ 00:51:44.779 Samuel Roberts: Awesome. All right. Yeah, sorry again, I thought… I was so deep in it, I wasn’t thinking about how taking a step back and giving you the whole picture, and just, you know, it’s hard to sometimes just pull tickets and not know where they’re going, so…
610 00:51:45.430 ⇒ 00:51:50.189 Casie Aviles: I think it’s a new process for us as well, anyway, so…
611 00:51:50.190 ⇒ 00:51:51.590 Samuel Roberts: Definitely. Definitely, yeah.
612 00:51:51.590 ⇒ 00:51:52.030 Casie Aviles: Fine.
613 00:51:52.710 ⇒ 00:51:53.600 Samuel Roberts: Okay.
614 00:51:54.290 ⇒ 00:51:59.330 Casie Aviles: And… yeah, yeah, I think that’s all. What else?
615 00:52:02.120 ⇒ 00:52:04.850 Casie Aviles: Yeah, I don’t think I have any more right now, but…
616 00:52:05.810 ⇒ 00:52:12.380 Samuel Roberts: Okay. Yeah, if anything’s… if any questions do arise, do… yeah, just ping me on Slack, or the channel, or tag me there.
617 00:52:12.540 ⇒ 00:52:17.689 Samuel Roberts: And I can… I do have a dentist appointment today in about an hour.
618 00:52:17.940 ⇒ 00:52:19.220 Casie Aviles: Oh, okay.
619 00:52:19.620 ⇒ 00:52:30.109 Samuel Roberts: But I’ll be back after that, so that… from 1 to 2, at least, I’ll probably be not able to respond Eastern, but, which I think is, like, 12 hours from you, so yeah, 1 to 2 your time, so…
620 00:52:32.540 ⇒ 00:52:34.679 Samuel Roberts: But besides that, I’ll be on Slack.
621 00:52:35.930 ⇒ 00:52:36.830 Casie Aviles: Okay, okay.
622 00:52:38.340 ⇒ 00:52:38.960 Samuel Roberts: Alright.
623 00:52:40.110 ⇒ 00:52:41.689 Casie Aviles: Alright, yeah, thank you, Sam.
624 00:52:42.180 ⇒ 00:52:46.180 Samuel Roberts: Yeah, sorry again, hopefully that’s clear enough, and I’ll do a diagram right now.
625 00:52:46.300 ⇒ 00:52:47.720 Samuel Roberts: Where I head out.
626 00:52:48.720 ⇒ 00:52:49.470 Casie Aviles: Thank you.
627 00:52:49.910 ⇒ 00:52:50.600 Samuel Roberts: Alrighty.
628 00:52:51.320 ⇒ 00:52:52.179 Samuel Roberts: See you later.