Meeting Title: Restructured Central Doc QA Session Date: 2026-03-11 Meeting participants: Pranav Narahari, Mustafa Raja, Samuel Roberts, Uttam Kumaran, Amber Lin
WEBVTT
1 00:00:19.410 ⇒ 00:00:20.300 Mustafa Raja: Hey.
2 00:00:21.980 ⇒ 00:00:22.950 Pranav Narahari: Hey, Mustavo.
3 00:00:23.330 ⇒ 00:00:24.209 Mustafa Raja: How are you?
4 00:00:24.860 ⇒ 00:00:26.400 Pranav Narahari: Pretty good. How’s your day been?
5 00:00:26.910 ⇒ 00:00:29.240 Mustafa Raja: Yeah, pretty good.
6 00:00:30.060 ⇒ 00:00:31.560 Pranav Narahari: Pretty good. Busy?
7 00:00:32.560 ⇒ 00:00:33.879 Mustafa Raja: Yeah, pretty much.
8 00:00:35.240 ⇒ 00:00:35.970 Mustafa Raja: How’s yours?
9 00:00:36.570 ⇒ 00:00:42.220 Pranav Narahari: It’s been, yeah, it’s been busy. A little bit lighter than, the last couple days.
10 00:00:43.020 ⇒ 00:00:44.900 Mustafa Raja: Yeah, same boat.
11 00:00:45.210 ⇒ 00:00:45.840 Pranav Narahari: Yeah, go ahead.
12 00:00:45.840 ⇒ 00:00:52.189 Mustafa Raja: Yeah, this week has been lighter for me, compared to past couple weeks, especially when we were…
13 00:00:52.880 ⇒ 00:00:56.560 Mustafa Raja: You know, migrating some of the Eden stuff.
14 00:00:56.700 ⇒ 00:00:57.960 Mustafa Raja: Over to Omni?
15 00:00:58.460 ⇒ 00:01:02.380 Mustafa Raja: This weekend has been lighter than that, but still work.
16 00:01:02.970 ⇒ 00:01:04.010 Pranav Narahari: That’s good to hear.
17 00:01:04.290 ⇒ 00:01:04.670 Mustafa Raja: Yeah.
18 00:01:05.640 ⇒ 00:01:11.529 Pranav Narahari: Yeah, for me, I’ve just been getting, like, pulled in, and it’s, like, not pulled in, because I’m, like, enjoying it, but…
19 00:01:11.890 ⇒ 00:01:14.990 Pranav Narahari: like, sim sales stuff,
20 00:01:16.550 ⇒ 00:01:23.020 Pranav Narahari: Then there’s, like, also some, like, cool, like, marketing things that we’re doing, too. Like, last week we did, like, an office hours.
21 00:01:23.740 ⇒ 00:01:30.109 Pranav Narahari: But the most exciting stuff is, like, the new AI projects that I think are gonna be…
22 00:01:30.720 ⇒ 00:01:37.690 Pranav Narahari: Hopefully getting signed really soon, maybe even starting next Monday, there’s… there might be a new Eden AI project.
23 00:01:38.430 ⇒ 00:01:39.090 Mustafa Raja: Oh.
24 00:01:39.400 ⇒ 00:01:40.630 Mustafa Raja: Yeah, that’s cool.
25 00:01:40.960 ⇒ 00:01:45.400 Pranav Narahari: Yeah, yeah, that one’s pretty cool, and it’s like… .
26 00:01:46.000 ⇒ 00:01:50.250 Mustafa Raja: It’s the same Eden, right, that we already… our clients were there, right?
27 00:01:50.250 ⇒ 00:01:57.899 Pranav Narahari: Yeah, yeah, same one. I haven’t really touched the Eden Project, so I don’t really know any of the contacts, but maybe Danny you’ve worked with?
28 00:01:58.680 ⇒ 00:02:05.560 Mustafa Raja: No, I just got pulled in a couple weeks ago, so I don’t know anyone, you know…
29 00:02:05.870 ⇒ 00:02:07.080 Mustafa Raja: on their side.
30 00:02:07.390 ⇒ 00:02:09.719 Mustafa Raja: I’ve just been working for migrations.
31 00:02:10.199 ⇒ 00:02:16.689 Pranav Narahari: Gotcha, gotcha. Yeah, well, this project seems pretty cool. It’s, like, MCP servers for, like, Slack.
32 00:02:16.869 ⇒ 00:02:23.999 Pranav Narahari: And then, also, Google Workspace, but, I think…
33 00:02:24.000 ⇒ 00:02:25.659 Mustafa Raja: I mean, Google works with…
34 00:02:26.930 ⇒ 00:02:27.690 Pranav Narahari: What’s that?
35 00:02:28.110 ⇒ 00:02:31.659 Mustafa Raja: Is it going to be an MCP for Google Workspace also?
36 00:02:32.600 ⇒ 00:02:47.039 Pranav Narahari: I’m not sure yet. For the first phase, it might be integrated into Gemini, so since Gemini already has the integration with Google Workspace, we might not need to build anything there. The main thing might just be Slack.
37 00:02:47.190 ⇒ 00:02:50.289 Pranav Narahari: But I’m not entirely sure yet.
38 00:02:51.110 ⇒ 00:02:54.529 Mustafa Raja: Yeah, I think they… they also released a CLI tool.
39 00:02:55.950 ⇒ 00:02:59.459 Mustafa Raja: workspace, so… Yeah, pretty good options.
40 00:03:00.130 ⇒ 00:03:01.019 Pranav Narahari: Okay, cool.
41 00:03:02.180 ⇒ 00:03:10.290 Pranav Narahari: Yeah, well, yeah, so this, these documents are insane, they’re so big, I’m wondering…
42 00:03:10.290 ⇒ 00:03:17.319 Mustafa Raja: They are so conflicting also, if you read through them, you’ll find some conflicting information also there.
43 00:03:17.650 ⇒ 00:03:23.550 Pranav Narahari: Yeah, and so… there’s a couple ways that we can go about QA-ing this,
44 00:03:24.390 ⇒ 00:03:32.329 Pranav Narahari: you know, I was asking Cursor, of course, just, like, create me, like, a manual test plan, things of that nature. But then also just, like, scrolling through things and, like.
45 00:03:32.770 ⇒ 00:03:38.360 Pranav Narahari: comparing it to the Golden Dock, I think, is gonna be helpful. Also, just, like.
46 00:03:38.690 ⇒ 00:03:46.610 Pranav Narahari: my eyes looking at yours, but then more importantly, like, your eyes looking at mine, I think, is going to also help us out.
47 00:03:46.610 ⇒ 00:03:49.000 Mustafa Raja: What’s the golden dock here?
48 00:03:49.480 ⇒ 00:03:50.989 Pranav Narahari: is the pest.
49 00:03:51.740 ⇒ 00:03:52.690 Mustafa Raja: Oh.
50 00:03:52.690 ⇒ 00:03:55.509 Pranav Narahari: That’s what at least I saw it referred to as.
51 00:03:57.070 ⇒ 00:04:03.200 Mustafa Raja: Yeah, I think, either way, we are also going to, you know, restructure the test document also now.
52 00:04:05.690 ⇒ 00:04:08.800 Pranav Narahari: Yeah, I think it’s already been restructured, right?
53 00:04:09.210 ⇒ 00:04:10.230 Mustafa Raja: Yeah, yeah.
54 00:04:10.230 ⇒ 00:04:10.820 Pranav Narahari: Okay.
55 00:04:11.710 ⇒ 00:04:12.270 Mustafa Raja: So.
56 00:04:13.020 ⇒ 00:04:17.310 Pranav Narahari: Oh, so are you saying the restructured pest document is not the Golden Doc?
57 00:04:17.950 ⇒ 00:04:20.019 Mustafa Raja: Yeah, so,
58 00:04:20.279 ⇒ 00:04:34.109 Mustafa Raja: So the reason this document would be called Golden Doc is because Amber has worked, so much on it, that, Andy, almost never has, you know, any…
59 00:04:34.420 ⇒ 00:04:43.910 Mustafa Raja: conflicting outputs from that. But since we’re restructuring it, I don’t know if we can consider that as a golden dock anymore, you know?
60 00:04:43.910 ⇒ 00:04:47.689 Pranav Narahari: I see. Okay. Well, that’s good to know, so then…
61 00:04:48.070 ⇒ 00:04:50.660 Pranav Narahari: We may have a good amount of work to do here.
62 00:04:51.960 ⇒ 00:04:56.129 Pranav Narahari: Yeah, let’s just… let’s just get started. I’ll, let me share my screen.
63 00:04:58.230 ⇒ 00:05:03.270 Mustafa Raja: Also, do you not have, upload permissions here?
64 00:05:03.270 ⇒ 00:05:05.809 Pranav Narahari: Yeah, it wasn’t cutting…
65 00:05:06.350 ⇒ 00:05:14.120 Mustafa Raja: I think this is just… let me share it. Let me try sharing… turn off… Okay.
66 00:05:25.830 ⇒ 00:05:28.170 Mustafa Raja: Yeah, right now.
67 00:05:32.600 ⇒ 00:05:38.190 Mustafa Raja: You should be able to… Add… Stuff.
68 00:05:39.020 ⇒ 00:05:44.320 Mustafa Raja: Oh, so… I gave you permission to restructure central docks.
69 00:05:45.400 ⇒ 00:05:46.209 Pranav Narahari: Oh, okay.
70 00:05:47.010 ⇒ 00:05:52.870 Mustafa Raja: Yeah, try that, and let me give you permission to this ABC thing.
71 00:05:54.770 ⇒ 00:05:58.079 Mustafa Raja: Because you will be needing that, I believe.
72 00:06:11.940 ⇒ 00:06:16.800 Pranav Narahari: And so, one thing I didn’t do to mine yet is removing the images,
73 00:06:19.140 ⇒ 00:06:21.449 Pranav Narahari: I wasn’t sure if that’s the… Sorry, go ahead.
74 00:06:21.450 ⇒ 00:06:25.419 Mustafa Raja: Did Gurser restructure your document?
75 00:06:27.070 ⇒ 00:06:29.829 Pranav Narahari: I use Cursor, yeah. Why?
76 00:06:30.210 ⇒ 00:06:32.060 Mustafa Raja: It’s a cursor kept the images.
77 00:06:34.040 ⇒ 00:06:34.870 Pranav Narahari: Yeah.
78 00:06:35.940 ⇒ 00:06:36.650 Mustafa Raja: Okay.
79 00:06:37.680 ⇒ 00:06:43.100 Pranav Narahari: Well, so also… Are you… are you… is that surprising? Like, would… would… did cursors.
80 00:06:43.100 ⇒ 00:06:46.639 Mustafa Raja: Yeah, for me, cursor didn’t, preserve any of the images.
81 00:06:47.110 ⇒ 00:06:48.570 Pranav Narahari: Okay, yeah, so…
82 00:06:48.920 ⇒ 00:06:56.289 Pranav Narahari: it gave me… so I asked for, like, a really specific test plan, like, I went through, like, 37 steps. I can show you, like, the full…
83 00:06:56.420 ⇒ 00:07:05.789 Pranav Narahari: I can show you the full thread, but it did tell me, like, yeah, well, images you’re not even utilizing in your RAG application anyways, so…
84 00:07:05.920 ⇒ 00:07:12.869 Pranav Narahari: you can remove them. The reason why I didn’t remove them was just because
85 00:07:13.290 ⇒ 00:07:19.490 Pranav Narahari: I didn’t want to remove any, like, any extra information.
86 00:07:20.450 ⇒ 00:07:26.769 Pranav Narahari: So… and I… that… by removing certain images, I thought that could also affect just, like, the overall structure, potentially.
87 00:07:26.990 ⇒ 00:07:31.700 Pranav Narahari: Interesting. This… mine is way bigger.
88 00:07:31.840 ⇒ 00:07:32.899 Pranav Narahari: I wonder if that was…
89 00:07:33.720 ⇒ 00:07:35.429 Mustafa Raja: I think that’s just commercial.
90 00:07:36.440 ⇒ 00:07:38.420 Pranav Narahari: Yeah, with images, it makes sense, yeah.
91 00:07:41.810 ⇒ 00:07:43.660 Mustafa Raja: You’ll have to open with ya.
92 00:07:44.060 ⇒ 00:07:44.690 Pranav Narahari: Yo.
93 00:07:48.290 ⇒ 00:07:53.560 Pranav Narahari: So I think it’s okay if there’s a little bit of variation in formatting,
94 00:07:53.930 ⇒ 00:07:59.080 Pranav Narahari: you have a better understanding of the design of the RAG system?
95 00:07:59.740 ⇒ 00:08:06.430 Pranav Narahari: And actually… Given that, I should probably see if Sam can join this too, since I think…
96 00:08:08.120 ⇒ 00:08:14.070 Pranav Narahari: His understanding of the… The, like, the rag design will also be helpful here.
97 00:08:14.960 ⇒ 00:08:16.029 Pranav Narahari: Let me add them.
98 00:08:38.740 ⇒ 00:08:49.140 Pranav Narahari: Okay, so… Off the bat, do you see any…
99 00:08:50.000 ⇒ 00:08:53.330 Pranav Narahari: Anything that’s standing out with the table of contents?
100 00:08:54.780 ⇒ 00:08:55.800 Mustafa Raja: This is good.
101 00:08:56.190 ⇒ 00:08:57.679 Pranav Narahari: This looks good. Okay.
102 00:08:58.080 ⇒ 00:09:06.689 Mustafa Raja: Yeah, but I don’t think that this one is managed by… what’s it called? Google Docs itself, no?
103 00:09:07.850 ⇒ 00:09:23.850 Mustafa Raja: So, meaning, if we go, if we add, or if we change any headings here, or add new headings, what happens is we would want, Google Docs to, update the…
104 00:09:24.300 ⇒ 00:09:34.490 Mustafa Raja: table of contents itself, no? So, we could remove this one, and then add a new one, going into insert tab, and then, you know…
105 00:09:35.650 ⇒ 00:09:42.150 Pranav Narahari: Okay, let me… you’ll probably do this quicker than me. Let me just let you share screen, and then, you can show me what you mean.
106 00:09:57.210 ⇒ 00:10:00.470 Pranav Narahari: Oh, so it generates the table of contents, okay, interesting.
107 00:10:02.490 ⇒ 00:10:09.170 Mustafa Raja: So I’ll just remove this, and then what I will do is… Nope.
108 00:10:12.620 ⇒ 00:10:13.969 Mustafa Raja: Let me go there.
109 00:10:16.620 ⇒ 00:10:19.739 Mustafa Raja: And this will be maintained by Google Docs.
110 00:10:20.740 ⇒ 00:10:21.750 Pranav Narahari: Oh, okay.
111 00:10:21.750 ⇒ 00:10:27.510 Mustafa Raja: And what it does is it takes a look at all of the headings, and then, you know, regenerates the table of contents.
112 00:10:27.510 ⇒ 00:10:28.870 Pranav Narahari: Nice. Okay.
113 00:10:29.300 ⇒ 00:10:30.280 Pranav Narahari: That’s perfect.
114 00:10:30.890 ⇒ 00:10:44.499 Pranav Narahari: Okay, so let’s just do, like… there’s a high level, just, like, skim, like, you can skim kind of pretty quickly through this, and if something pops up, we can talk about it. So these open links, right, if you scroll up a little bit…
115 00:10:46.680 ⇒ 00:10:46.859 Mustafa Raja: Oh.
116 00:10:47.040 ⇒ 00:10:47.700 Pranav Narahari: Excuse me?
117 00:10:50.520 ⇒ 00:10:55.140 Pranav Narahari: We probably shouldn’t have, like… should we just have the… the link itself there?
118 00:10:57.230 ⇒ 00:10:59.610 Mustafa Raja: I think the link, the,
119 00:10:59.840 ⇒ 00:11:07.020 Mustafa Raja: It doesn’t matter, because when we extract from the JSON, it does get included, you know?
120 00:11:07.420 ⇒ 00:11:11.249 Pranav Narahari: Yeah, true. Okay, cool. Because, yeah, we’re going to be passing the mark down, right?
121 00:11:11.500 ⇒ 00:11:21.129 Mustafa Raja: Yeah, not the Markdown. We cannot download it in Markdown programmatically, we can only download it in JSON format.
122 00:11:22.170 ⇒ 00:11:23.000 Pranav Narahari: Oh, okay.
123 00:11:23.280 ⇒ 00:11:25.430 Mustafa Raja: Yeah, is that insane.
124 00:11:25.430 ⇒ 00:11:26.660 Samuel Roberts: Fire message just now?
125 00:11:27.060 ⇒ 00:11:30.779 Pranav Narahari: Oh yeah, I appreciate that. Oh, I didn’t even send you the link yet. I don’t even know how you found it.
126 00:11:30.780 ⇒ 00:11:31.720 Samuel Roberts: I found it.
127 00:11:31.900 ⇒ 00:11:33.970 Samuel Roberts: Magic. Now the calendar, I figured…
128 00:11:34.100 ⇒ 00:11:37.359 Samuel Roberts: I figured one of you two would have it on your calendars.
129 00:11:37.540 ⇒ 00:11:38.829 Pranav Narahari: Perfect, yeah, appreciate that.
130 00:11:40.300 ⇒ 00:11:41.249 Samuel Roberts: Okay, what am I, what am I…
131 00:11:41.250 ⇒ 00:11:51.609 Pranav Narahari: But, yeah, first off, okay, that makes sense. Sam, right now, what we’re doing is just kind of doing, like, a high-level skim of, Commercial Central Doc right now, which is the one I created.
132 00:11:51.610 ⇒ 00:11:52.030 Samuel Roberts: Okay.
133 00:11:52.560 ⇒ 00:11:57.650 Pranav Narahari: one thing that we were just talking about that I was like, okay, let’s bring in Sam here, is,
134 00:11:58.310 ⇒ 00:12:04.259 Pranav Narahari: with… how we’re building out the RAG pipeline,
135 00:12:04.870 ⇒ 00:12:14.490 Pranav Narahari: what I didn’t do for this central doc yet is removing the images and the associated, like, text with those images. I thought that…
136 00:12:15.460 ⇒ 00:12:25.279 Pranav Narahari: I don’t know if that’s reducing accuracy by having the images there. It’s of course not adding additional context. My worry is, is just that, like, we’ll cut out some of the…
137 00:12:25.670 ⇒ 00:12:27.850 Samuel Roberts: Important information with that.
138 00:12:27.850 ⇒ 00:12:36.370 Pranav Narahari: But if you feel like having these images in there are going to reduce the… The accuracy, then…
139 00:12:36.840 ⇒ 00:12:40.889 Pranav Narahari: I could just remove them. I’m just kind of wondering what you think here.
140 00:12:41.360 ⇒ 00:12:49.219 Samuel Roberts: Yeah, I mean, we talked about, like, processing the images so we get the content out, right?
141 00:12:49.220 ⇒ 00:12:56.429 Pranav Narahari: Yeah, that’s another thing, too. I was like, I don’t want to delete them and then have to, like, figure out how to add them. No, I wouldn’t… I wouldn’t get rid of them. I imagine…
142 00:12:57.150 ⇒ 00:12:59.830 Samuel Roberts: I don’t know exactly how they’re gonna come down in the…
143 00:13:00.640 ⇒ 00:13:06.199 Samuel Roberts: Mustafa, do you know what, like, happens when we look at the JSON of a doc that has a bunch of images?
144 00:13:06.490 ⇒ 00:13:12.670 Mustafa Raja: Ignore these images, and then, whatever the text is, we convert it into a markdown format.
145 00:13:12.940 ⇒ 00:13:13.440 Samuel Roberts: Right.
146 00:13:13.690 ⇒ 00:13:18.999 Mustafa Raja: And then we convert it in modern format, and that’s all just… that’s just all text.
147 00:13:19.450 ⇒ 00:13:29.480 Samuel Roberts: Yeah, so the images, do they just get ignored at that point? Yes, they totally get ignored. Okay, in that case, I’d say leave them. We’ll figure out how to extract that info, probably as part of the…
148 00:13:30.640 ⇒ 00:13:35.669 Samuel Roberts: process… I don’t know where exactly in the process yet, but yeah, I think that’ll…
149 00:13:35.920 ⇒ 00:13:37.490 Samuel Roberts: It’s better to leave him for now.
150 00:13:38.130 ⇒ 00:13:39.610 Pranav Narahari: Perfect. Okay, cool.
151 00:13:42.730 ⇒ 00:13:48.429 Pranav Narahari: Mustafa, anything else that you noticed, just… and I know we haven’t gone that far in, but before we continue?
152 00:13:49.570 ⇒ 00:13:54.609 Mustafa Raja: No, this looks pretty good, to be… to be honest. At the first glance, this is pretty good.
153 00:13:54.850 ⇒ 00:13:55.500 Pranav Narahari: Okay.
154 00:13:56.060 ⇒ 00:13:56.840 Pranav Narahari: Cool.
155 00:14:00.510 ⇒ 00:14:02.760 Pranav Narahari: And I think what we’ll do, too, is, like…
156 00:14:03.140 ⇒ 00:14:09.270 Pranav Narahari: It’s impossible to do a deep dive on 5 different documents with 200 pages, right?
157 00:14:09.460 ⇒ 00:14:19.869 Pranav Narahari: What we’ll do is just, like, okay, let’s send this out to them, they’ll give us their feedback, but also, when we run through things, if there’s a big difference in accuracy, like, then…
158 00:14:20.040 ⇒ 00:14:37.170 Pranav Narahari: we’ll see the associated message that was incorrect, but previously was correct, and then we’ll make the updates into the central docs. I think this is one of those things where it’s like, it’s actually better to move a little bit quicker, instead of just, like, trying to perfect it on the first time.
159 00:14:37.740 ⇒ 00:14:38.270 Mustafa Raja: Okay.
160 00:14:38.270 ⇒ 00:14:39.069 Samuel Roberts: No, I agree.
161 00:14:39.260 ⇒ 00:14:46.909 Samuel Roberts: I think getting in front of them, getting their feedback, because there’s the two… there’s the… it has to be good for the chunking and the AI, but someone…
162 00:14:47.070 ⇒ 00:14:51.610 Samuel Roberts: In some form, we’ll be maintaining this, whether or not we put a system on top, in forms or whatever, but, like…
163 00:14:51.770 ⇒ 00:14:53.580 Samuel Roberts: It still needs to be somewhat human.
164 00:14:54.320 ⇒ 00:14:59.060 Samuel Roberts: readable and have everything. So I think, like, getting their feedback is gonna be good, even if it’s not a finalized thing.
165 00:15:00.540 ⇒ 00:15:01.320 Mustafa Raja: Okay.
166 00:15:01.970 ⇒ 00:15:02.510 Mustafa Raja: Yes.
167 00:15:02.510 ⇒ 00:15:11.139 Pranav Narahari: I had a couple other things. So, the owner… I noticed that there’s an owner for each one of these, or at least for the commercial central dock.
168 00:15:11.510 ⇒ 00:15:15.399 Pranav Narahari: Is that something that we need to verify?
169 00:15:16.220 ⇒ 00:15:18.449 Pranav Narahari: I don’t think that’s gonna be changing, right?
170 00:15:18.830 ⇒ 00:15:23.530 Samuel Roberts: The owner, meaning… There’s, like, listed in the…
171 00:15:24.360 ⇒ 00:15:25.940 Pranav Narahari: Yeah, in the central dock.
172 00:15:29.460 ⇒ 00:15:31.960 Samuel Roberts: Like, Mustafa, if you clicked on commercial again?
173 00:15:35.070 ⇒ 00:15:40.500 Pranav Narahari: And you scroll down right into that first section… Yeah, right there.
174 00:15:41.100 ⇒ 00:15:42.190 Samuel Roberts: Oh, I see, yeah.
175 00:15:44.490 ⇒ 00:15:50.389 Pranav Narahari: I’m just wondering if this is important. If it’s not important, like, it’s never coming up, like, we don’t even need to talk about it.
176 00:15:52.660 ⇒ 00:15:54.460 Samuel Roberts: I don’t know if that’s ever come up.
177 00:15:54.930 ⇒ 00:15:55.740 Pranav Narahari: Okay.
178 00:15:56.660 ⇒ 00:15:59.939 Samuel Roberts: wouldn’t… I wouldn’t want to remove too much yet, you know what I mean? Like…
179 00:15:59.940 ⇒ 00:16:04.280 Pranav Narahari: Exactly. I think we should keep it, I was just wondering if we need to update it, but I think we’re good.
180 00:16:04.280 ⇒ 00:16:08.340 Samuel Roberts: Oh, yeah, I mean, if it has to be updated, they may be able to tell us that once we share.
181 00:16:08.620 ⇒ 00:16:11.310 Pranav Narahari: Perfect. Yeah, that’s a good point.
182 00:16:11.630 ⇒ 00:16:14.600 Pranav Narahari: Can you go to 2.3, Mustafa?
183 00:16:15.100 ⇒ 00:16:19.689 Pranav Narahari: And I can also share screen if that’s easier, but I think you got it, either way.
184 00:16:20.400 ⇒ 00:16:22.950 Pranav Narahari: Hmm…
185 00:16:22.950 ⇒ 00:16:23.700 Mustafa Raja: District.
186 00:16:23.700 ⇒ 00:16:24.649 Samuel Roberts: Oh, man.
187 00:16:27.140 ⇒ 00:16:30.500 Pranav Narahari: It’s… it’s, emergency contacts.
188 00:16:41.570 ⇒ 00:16:42.240 Samuel Roberts: Cool.
189 00:16:42.240 ⇒ 00:16:43.320 Pranav Narahari: Yeah.
190 00:16:45.630 ⇒ 00:16:48.010 Pranav Narahari: So, do we need to add emails here?
191 00:16:51.600 ⇒ 00:16:53.050 Samuel Roberts: Oh,
192 00:16:56.790 ⇒ 00:17:02.930 Samuel Roberts: I don’t… Know if that’s gonna be important, because they’ll probably… they’re probably able to…
193 00:17:04.280 ⇒ 00:17:06.939 Samuel Roberts: find that. There was some…
194 00:17:08.400 ⇒ 00:17:12.800 Samuel Roberts: we were talking at some point about a little more information like that, but I think…
195 00:17:12.980 ⇒ 00:17:27.849 Samuel Roberts: what we ended up doing, and it’s not quite related to this in terms of emergency contacts, but for people, they have a system, I forget what their, like, HR system is, and one of the things that they wanted us to add was when things are asked, it says specifically go to…
196 00:17:28.760 ⇒ 00:17:29.960 Samuel Roberts: that system.
197 00:17:30.840 ⇒ 00:17:37.610 Samuel Roberts: I don’t… Remember… do you remember what it’s called, Mustafa?
198 00:17:38.640 ⇒ 00:17:40.030 Mustafa Raja: No…
199 00:17:42.410 ⇒ 00:17:43.170 Samuel Roberts: But yeah, I think…
200 00:17:43.170 ⇒ 00:17:43.770 Pranav Narahari: Here’s another…
201 00:17:43.770 ⇒ 00:17:48.349 Samuel Roberts: There’s, like, an HR system we’re not really integrating with that I think we probably just… their names are probably fine.
202 00:17:49.680 ⇒ 00:17:53.399 Pranav Narahari: Oh, I think… yeah, I think I know what you’re talking about.
203 00:17:53.400 ⇒ 00:17:54.820 Samuel Roberts: I can’t think of the name.
204 00:18:04.920 ⇒ 00:18:10.099 Samuel Roberts: Is it Paylocity? Is that what I’m thinking of?
205 00:18:10.100 ⇒ 00:18:11.560 Mustafa Raja: Oh, yeah, I think it is.
206 00:18:11.710 ⇒ 00:18:24.999 Samuel Roberts: Okay, yeah, I found that in an email. Okay, cool. Yeah, so I think for other, like, for other things, we had them just respond with a message that, like, go to that, rather than try to integrate anything. So I’m probably inclined to say we don’t need them here, unless…
207 00:18:25.300 ⇒ 00:18:27.090 Samuel Roberts: We get feedback from them that…
208 00:18:28.470 ⇒ 00:18:30.289 Samuel Roberts: They definitely wanted here, but then that’s a whole.
209 00:18:30.290 ⇒ 00:18:30.650 Pranav Narahari: Yeah.
210 00:18:30.650 ⇒ 00:18:32.110 Samuel Roberts: maintaining that much, or…
211 00:18:32.520 ⇒ 00:18:33.529 Pranav Narahari: Okay, sounds good.
212 00:18:34.110 ⇒ 00:18:39.639 Pranav Narahari: Mustafa, can you just remove that, like, mess, like, that, line right above? Verify names and add emails?
213 00:18:55.000 ⇒ 00:18:59.420 Pranav Narahari: Alright… Let’s see what else…
214 00:19:12.340 ⇒ 00:19:15.950 Pranav Narahari: I also noticed that there’s, like, a commercial pest section.
215 00:19:16.420 ⇒ 00:19:19.310 Pranav Narahari: Is that supposed to be in pests? Is that supposed to be in both?
216 00:19:21.290 ⇒ 00:19:22.260 Mustafa Raja: No, I didn’t matter.
217 00:19:22.260 ⇒ 00:19:23.550 Samuel Roberts: Yeah.
218 00:19:24.690 ⇒ 00:19:26.230 Mustafa Raja: I don’t know much about that.
219 00:19:27.410 ⇒ 00:19:28.120 Pranav Narahari: Okay.
220 00:19:29.120 ⇒ 00:19:33.800 Pranav Narahari: I think we leave it in, right? It’s not a big deal if it’s in. It’s a big deal if it’s not in.
221 00:19:34.140 ⇒ 00:19:37.980 Samuel Roberts: Yeah, I would say leave it. I mean, you could definitely ask that and, like, highlight it, maybe.
222 00:19:38.730 ⇒ 00:19:39.160 Pranav Narahari: Yeah.
223 00:19:39.160 ⇒ 00:19:44.870 Samuel Roberts: This is, again, where some of the stuff, like, they have… It’s not like a clean… kind of…
224 00:19:44.870 ⇒ 00:19:52.130 Pranav Narahari: data in general, like, there’s lots of weird exceptions and things that I don’t fully understand. I think Amber has a little bit better…
225 00:19:52.280 ⇒ 00:19:55.460 Samuel Roberts: Context of that from working through some of the stuff previously, but…
226 00:19:57.160 ⇒ 00:19:59.940 Samuel Roberts: Yeah, the… I would say leave it, definitely.
227 00:20:00.810 ⇒ 00:20:01.480 Pranav Narahari: Okay.
228 00:20:05.170 ⇒ 00:20:11.330 Pranav Narahari: Yeah, so there’s a few notes in here that I’ll just, like, clean up after this call, before you send it out.
229 00:20:12.130 ⇒ 00:20:14.800 Pranav Narahari: Let me see if there’s anything big I wanted to talk about…
230 00:20:23.220 ⇒ 00:20:26.080 Pranav Narahari: Yeah, I feel pretty good.
231 00:20:27.060 ⇒ 00:20:28.580 Pranav Narahari: My main thing was just, like.
232 00:20:28.750 ⇒ 00:20:35.899 Pranav Narahari: Is Mustafa gonna look at this and be like, oh, this is completely off from, like, what I did, and, like, we gotta figure that out, but… okay.
233 00:20:36.060 ⇒ 00:20:38.160 Pranav Narahari: I think, I think we’re looking pretty good here, then.
234 00:20:38.530 ⇒ 00:20:43.169 Samuel Roberts: Okay, yeah, the only thing I was curious about, just having… just scrolling through, was the…
235 00:20:43.370 ⇒ 00:20:48.500 Samuel Roberts: headers and the numbering system. I know Mustafa and I talked about that as a way to, like…
236 00:20:49.410 ⇒ 00:20:51.899 Samuel Roberts: Keep the headers and the hierarchy.
237 00:20:52.170 ⇒ 00:20:56.750 Samuel Roberts: Probably not critical until we actually start doing the embedding, but just to…
238 00:20:56.750 ⇒ 00:21:01.650 Mustafa Raja: Yeah, I think… yeah, I think what we want to do is we just want to…
239 00:21:02.270 ⇒ 00:21:06.549 Mustafa Raja: Let me see… yeah, so you see… yeah.
240 00:21:10.470 ⇒ 00:21:14.909 Mustafa Raja: So you see here how the headers are numbered. I think…
241 00:21:15.650 ⇒ 00:21:18.959 Mustafa Raja: I don’t know why AI left these out, but…
242 00:21:18.960 ⇒ 00:21:25.110 Samuel Roberts: Yeah, that’s why, yeah. But I’m just, I mean, I… and it might be something we need to test once we actually do the embeddings by area and group.
243 00:21:25.350 ⇒ 00:21:29.070 Samuel Roberts: Rather than just… arbitrary limits.
244 00:21:29.210 ⇒ 00:21:29.930 Samuel Roberts: But…
245 00:21:29.980 ⇒ 00:21:30.820 Mustafa Raja: Yeah.
246 00:21:34.470 ⇒ 00:21:36.470 Samuel Roberts: Yeah, okay, yeah, that one.
247 00:21:37.930 ⇒ 00:21:42.080 Samuel Roberts: Yeah, see, there’s something there. I just, I feel like consistency is probably the most important part there, but…
248 00:21:42.530 ⇒ 00:21:43.200 Mustafa Raja: Yeah.
249 00:21:43.200 ⇒ 00:21:49.750 Samuel Roberts: Again, it’s not critical yet, either. It’s gonna be more critical, I think, later. If… it might be more critical later, I’ll put it that way, I don’t wanna…
250 00:21:55.040 ⇒ 00:21:56.250 Pranav Narahari: Cool, yeah, let me…
251 00:21:56.380 ⇒ 00:21:58.490 Mustafa Raja: My only concern here is…
252 00:22:02.920 ⇒ 00:22:08.629 Mustafa Raja: My document is 68 pages long, and this one is 160.
253 00:22:10.020 ⇒ 00:22:10.839 Samuel Roberts: Oh, boy.
254 00:22:11.120 ⇒ 00:22:12.529 Samuel Roberts: And this, this is…
255 00:22:12.530 ⇒ 00:22:14.460 Mustafa Raja: by AI.
256 00:22:16.760 ⇒ 00:22:19.449 Mustafa Raja: By, I mean, I mean cursor.
257 00:22:23.110 ⇒ 00:22:31.930 Mustafa Raja: But I think the headers are all there. I don’t know if the 100 pages are going to be these.
258 00:22:32.640 ⇒ 00:22:36.550 Samuel Roberts: Could be part of that. It could be just duplicate information that we condensed a little bit more.
259 00:22:36.910 ⇒ 00:22:38.840 Mustafa Raja: Yeah, and then these images also.
260 00:22:38.840 ⇒ 00:22:40.809 Samuel Roberts: The images add a lot, you’re right.
261 00:22:42.760 ⇒ 00:22:46.649 Mustafa Raja: But if I were to search something, let’s take a phrase…
262 00:22:51.610 ⇒ 00:22:54.179 Mustafa Raja: So I’m able to find that.
263 00:22:56.840 ⇒ 00:22:59.279 Samuel Roberts: Wait, fertilization? Composting?
264 00:22:59.470 ⇒ 00:23:00.960 Mustafa Raja: Wait.
265 00:23:01.900 ⇒ 00:23:02.430 Samuel Roberts: Nope.
266 00:23:02.870 ⇒ 00:23:04.229 Samuel Roberts: There you go, okay, yeah.
267 00:23:04.550 ⇒ 00:23:06.059 Mustafa Raja: Where is this? Oh, it is.
268 00:23:06.430 ⇒ 00:23:07.260 Mustafa Raja: Yeah.
269 00:23:07.600 ⇒ 00:23:13.459 Mustafa Raja: So I’m able to find this, let’s try… Let’s try this.
270 00:23:15.890 ⇒ 00:23:16.910 Mustafa Raja: Yeah.
271 00:23:17.450 ⇒ 00:23:20.870 Samuel Roberts: Okay. I mean, we probably want to go the other way, too, and make sure that things are…
272 00:23:21.980 ⇒ 00:23:24.280 Samuel Roberts: Not lost, yeah. Hope.
273 00:23:24.490 ⇒ 00:23:25.160 Mustafa Raja: Oh.
274 00:23:31.040 ⇒ 00:23:34.400 Mustafa Raja: Home procedures for… yeah, you see, this whole thing…
275 00:23:34.400 ⇒ 00:23:37.920 Samuel Roberts: Wow, oh my goodness, yeah, so all that was getting ignored anyway, right?
276 00:23:38.260 ⇒ 00:23:39.030 Mustafa Raja: Yeah.
277 00:23:39.620 ⇒ 00:23:41.279 Samuel Roberts: So we might need to do…
278 00:23:42.370 ⇒ 00:23:48.720 Mustafa Raja: Yeah, we might want to do something with this… Let me do this.
279 00:23:51.610 ⇒ 00:23:53.370 Mustafa Raja: Yeah, it’s able to find.
280 00:23:55.220 ⇒ 00:23:56.350 Samuel Roberts: Yeah, the images, though.
281 00:23:56.350 ⇒ 00:23:58.370 Mustafa Raja: Seeing a lot of images.
282 00:23:58.670 ⇒ 00:24:00.810 Samuel Roberts: A lot more than I realized, I will say.
283 00:24:01.200 ⇒ 00:24:09.150 Pranav Narahari: How long did it take to run this, Mustafa? Like, if you wanted to redo this and, like, include the images, would that be a big task?
284 00:24:09.150 ⇒ 00:24:14.290 Mustafa Raja: So, I think, include the images. Do we…
285 00:24:15.220 ⇒ 00:24:23.930 Mustafa Raja: I mean, I could just ask it to add the images, because I see in my cursor chart that it’s telling me images are not worth it.
286 00:24:25.100 ⇒ 00:24:32.950 Mustafa Raja: Right. What I can do is, I can just go ahead and ask it, you know, go ahead and include all the images, I don’t want to lose any information.
287 00:24:33.270 ⇒ 00:24:38.200 Pranav Narahari: I guess my just main concern is, and honestly, like.
288 00:24:38.290 ⇒ 00:24:54.009 Pranav Narahari: the images look like they’re taking up a lot of space, right? So it could explain the main difference in page length. But that would just kind of confirm it for us, right? If you add the images back in, the page length is around the same. It’s just, like, another…
289 00:24:54.060 ⇒ 00:25:01.940 Pranav Narahari: way to QA. And we talked about how in the future we might want to extract information from these images, so… it’s gonna need to be done at some point.
290 00:25:03.540 ⇒ 00:25:04.130 Mustafa Raja: Okay.
291 00:25:04.130 ⇒ 00:25:09.020 Samuel Roberts: I don’t know what… like, some of the images are probably more valuable than others, too. I imagine some are…
292 00:25:09.020 ⇒ 00:25:10.320 Pranav Narahari: Yeah, that’s true, like.
293 00:25:10.320 ⇒ 00:25:10.820 Samuel Roberts: Bye.
294 00:25:11.320 ⇒ 00:25:18.810 Pranav Narahari: What we’re gonna probably do is, like, run it through a script, pull in all the images, categorize them into, like, whatever we define as
295 00:25:19.780 ⇒ 00:25:21.370 Pranav Narahari: You know, important, not important.
296 00:25:21.370 ⇒ 00:25:25.329 Samuel Roberts: Right, like a description, text, kind of, yeah, we can probably pull all that out.
297 00:25:25.620 ⇒ 00:25:26.260 Pranav Narahari: Yup.
298 00:25:30.880 ⇒ 00:25:37.699 Pranav Narahari: And so… sorry, one more time for, like, the numbers and the headings for the… in the table of contents.
299 00:25:37.950 ⇒ 00:25:43.460 Pranav Narahari: So that’s gonna be necessary for creating, embeddings, is what we’re saying, right?
300 00:25:43.460 ⇒ 00:26:03.240 Samuel Roberts: Yeah, I mean, I… I don’t know if it’s necessary, because part of it was we had issues before where it was an arbitrary token limit that was being chunked, and so we started adding metadata to those to know, oh, this is nested well within whatever. So my thought was to just keep those numbers there, but if we’re gonna start chunking by sections instead.
301 00:26:03.240 ⇒ 00:26:08.259 Samuel Roberts: We probably still want something to… the metadata to tie it to, like, you know, if you’re deep in…
302 00:26:08.480 ⇒ 00:26:11.210 Samuel Roberts: The lawn mowing… Down here.
303 00:26:11.520 ⇒ 00:26:12.820 Samuel Roberts: And it, like…
304 00:26:13.510 ⇒ 00:26:24.049 Pranav Narahari: How I’m thinking we would do that is, like, essentially, like, we would just take that, and the… I agree, though, having the number there is the most, consistent way to do it, because.
305 00:26:24.050 ⇒ 00:26:24.430 Samuel Roberts: Yeah.
306 00:26:24.430 ⇒ 00:26:40.920 Pranav Narahari: if we search up lawn mowing, and we just say, okay, pull the text under lawn mowing, and then figure out how many, like, tokens that chunk should be, or how many chunks we should put that into, it might not be that consistent. So, okay, it definitely makes sense to have numbers there. Now that I say it out loud, it makes a lot more sense.
307 00:26:41.100 ⇒ 00:26:50.650 Samuel Roberts: Yeah, and I think some things will be mitigated by the fact that if we chunk, like, this whole section, it’ll be different, but then if we also need to know anything from above, we’ll still have that metadata.
308 00:26:50.810 ⇒ 00:26:53.769 Samuel Roberts: So I think it’s kind of the best of both worlds there.
309 00:26:54.400 ⇒ 00:26:56.519 Pranav Narahari: Perfect, okay. Let me create a ticket for that.
310 00:26:58.530 ⇒ 00:27:04.100 Mustafa Raja: Yeah, I think… I think for the text content, it’s all in there.
311 00:27:04.710 ⇒ 00:27:10.869 Mustafa Raja: It’s just the… Images might be what these documents are missing.
312 00:27:11.550 ⇒ 00:27:12.230 Pranav Narahari: Yeah.
313 00:27:13.800 ⇒ 00:27:15.190 Pranav Narahari: Okay, sounds good.
314 00:27:19.760 ⇒ 00:27:20.390 Samuel Roberts: Okay.
315 00:27:22.250 ⇒ 00:27:27.229 Pranav Narahari: Yeah, like I said, what we probably should just… Sorry, go ahead.
316 00:27:27.460 ⇒ 00:27:30.050 Mustafa Raja: This is a lot of text, but this is just an image.
317 00:27:30.740 ⇒ 00:27:31.950 Samuel Roberts: An image of text.
318 00:27:32.400 ⇒ 00:27:33.120 Mustafa Raja: Yeah.
319 00:27:36.580 ⇒ 00:27:37.330 Pranav Narahari: Yeah.
320 00:27:37.870 ⇒ 00:27:39.199 Pranav Narahari: And so…
321 00:27:39.380 ⇒ 00:27:40.830 Mustafa Raja: Underneath also.
322 00:27:41.030 ⇒ 00:27:58.329 Pranav Narahari: What we’re gonna do is we’re gonna create, like, a bunch of reports for them, essentially, like, one report is going to be extracting all these images, categorizing them into, like, okay, we know these are useless, we know these are relevant, there’s a group that may or may not be,
323 00:27:58.340 ⇒ 00:28:04.900 Pranav Narahari: And also, within that, we’ll categorize based on just, like, what are these images showing?
324 00:28:04.960 ⇒ 00:28:17.480 Pranav Narahari: like, are they click-by-click instructions where to, like, find more information? Are they, like what Mustafa just showed, just, like, a screenshot of text, and for whatever reason, they just didn’t paste that in?
325 00:28:18.160 ⇒ 00:28:22.670 Pranav Narahari: So… Yeah, we’ll do that. I think…
326 00:28:22.890 ⇒ 00:28:34.560 Pranav Narahari: this just is, like, I’m happy we’re doing this, like, ASAP now, we’re sending this over to them today, just so we can now get started on, like, okay, we have, like, a few of these, like… well, first thing to do is, like.
327 00:28:34.790 ⇒ 00:28:38.029 Pranav Narahari: Let’s get… let’s create embeddings for this.
328 00:28:38.320 ⇒ 00:28:46.650 Pranav Narahari: and then test it against all of the different messages that Andy has gone through for the current version, test it into this new version.
329 00:28:46.830 ⇒ 00:28:50.409 Pranav Narahari: Let’s… let’s see where the accuracy is at.
330 00:28:50.580 ⇒ 00:28:52.049 Pranav Narahari: That’s gonna tell us a lot.
331 00:28:52.550 ⇒ 00:28:53.100 Samuel Roberts: Yeah.
332 00:28:53.830 ⇒ 00:28:59.940 Mustafa Raja: And that we would want to do with an embedding pipeline in Master, right?
333 00:29:02.610 ⇒ 00:29:09.210 Samuel Roberts: Yeah, I mean, we could do it with the current embedding pipeline, but I don’t… I mean, there’s already problems with that, so I don’t know if that’s even worth…
334 00:29:10.980 ⇒ 00:29:11.750 Mustafa Raja: Yeah.
335 00:29:15.430 ⇒ 00:29:20.750 Samuel Roberts: Yeah, I think we want the new one. I mean, two changes, but I think it’s still… Worth it.
336 00:29:21.510 ⇒ 00:29:22.300 Mustafa Raja: Yeah.
337 00:29:23.360 ⇒ 00:29:29.849 Mustafa Raja: Yeah, I think the biggest advantage is going to be us being able to embed the whole section, rather than…
338 00:29:30.090 ⇒ 00:29:33.720 Mustafa Raja: Dividing the section in a number of characters. Yeah.
339 00:29:33.720 ⇒ 00:29:40.059 Samuel Roberts: That’s gonna be a big one for the embedding, and then the fact that we hopefully have, you know, deduplicated things a little bit, and condensed things that are…
340 00:29:40.310 ⇒ 00:29:44.089 Samuel Roberts: Phone numbers and hours that we’re getting confused and stuff.
341 00:29:44.420 ⇒ 00:29:45.350 Samuel Roberts: Yep.
342 00:29:45.450 ⇒ 00:29:47.249 Samuel Roberts: I think both of those together.
343 00:29:47.500 ⇒ 00:29:56.660 Samuel Roberts: Should improve a lot of those kind of weird edge cases where it’s, like, lawn this goes somewhere, but lawn this goes somewhere else, and…
344 00:29:56.660 ⇒ 00:30:03.270 Mustafa Raja: That was the first thing that I checked, if this is including lawn care stuff and lawn mowing or not.
345 00:30:03.610 ⇒ 00:30:07.020 Samuel Roberts: Yeah, exactly, and this is the sound of stuff that… yeah, you’re right.
346 00:30:11.030 ⇒ 00:30:11.640 Samuel Roberts: Alright.
347 00:30:11.640 ⇒ 00:30:12.250 Pranav Narahari: Cool.
348 00:30:15.300 ⇒ 00:30:17.180 Pranav Narahari: You want to run through another one, Mustafa?
349 00:30:18.850 ⇒ 00:30:20.549 Mustafa Raja: Yes…
350 00:30:20.550 ⇒ 00:30:26.180 Pranav Narahari: I know this one was, like, primarily concerning because of, like, the big page difference, but… I think it.
351 00:30:26.180 ⇒ 00:30:37.409 Mustafa Raja: Just like that, because the page difference is going to be there, but I think that the text is all there. This is the difference coming from the images themselves, yeah.
352 00:30:38.280 ⇒ 00:30:38.950 Pranav Narahari: Yep.
353 00:30:41.140 ⇒ 00:30:43.689 Mustafa Raja: So, this is the other one, home improvement.
354 00:30:45.820 ⇒ 00:30:52.670 Pranav Narahari: Also, one other thing, I… what we should be doing is uploading these MD files into GitHub.
355 00:30:52.790 ⇒ 00:31:06.670 Pranav Narahari: that way we make changes, it’s a little bit better, unless you guys already have a system for, like, how changes are being made, and, like, just tracking that. I think it would be good for us to just have that all happening within GitHub, and then…
356 00:31:07.060 ⇒ 00:31:15.419 Pranav Narahari: The problem is, of course, like, they’re making changes in Google Docs, we’re making it in GitHub.
357 00:31:16.350 ⇒ 00:31:35.660 Pranav Narahari: I think, though, like, we’re gonna change the system so they’re not going to be updating the doc directly, right? And so… what we’ll probably do is just, like, have them update a form. It’s actually not gonna update the doc itself, actually, it’s just gonna update the MD file. Maybe that creates, like, a PR automatically?
358 00:31:35.660 ⇒ 00:31:36.140 Samuel Roberts: Hmm.
359 00:31:36.140 ⇒ 00:31:38.610 Pranav Narahari: And then Sam, you and I review that.
360 00:31:38.740 ⇒ 00:31:42.240 Pranav Narahari: And… Our workflow would be pretty similar as well.
361 00:31:42.650 ⇒ 00:31:46.799 Pranav Narahari: If we want to update something, we create a PR for it.
362 00:31:49.550 ⇒ 00:31:50.040 Samuel Roberts: Yeah.
363 00:31:50.040 ⇒ 00:31:51.300 Pranav Narahari: What do you think about that?
364 00:31:51.720 ⇒ 00:31:54.159 Samuel Roberts: I think that could work. I’m trying to think.
365 00:31:54.890 ⇒ 00:31:56.030 Samuel Roberts: Does any…
366 00:31:56.820 ⇒ 00:32:02.420 Samuel Roberts: I just don’t want things to fall out of sync, is sort of my… but if we’re doing an update system of some sort.
367 00:32:02.550 ⇒ 00:32:05.770 Samuel Roberts: I don’t know what that’s exactly gonna look like yet, but yeah, I think…
368 00:32:10.870 ⇒ 00:32:16.280 Samuel Roberts: I mean, there might also be a way that we could just do a diff of, like, the generated markdown from the dock every time.
369 00:32:17.580 ⇒ 00:32:18.320 Pranav Narahari: Yeah.
370 00:32:18.480 ⇒ 00:32:21.680 Samuel Roberts: If we don’t… you know what I mean? Like, if that becomes too…
371 00:32:23.050 ⇒ 00:32:26.959 Samuel Roberts: cumbersome for them to have to, you know, depending on what they’re trying to change, I don’t know how we’re gonna…
372 00:32:27.840 ⇒ 00:32:37.849 Samuel Roberts: take that in yet. So if it’s still, like, if they’re updating this doc, we could still do, like, a diff of a markdown file that’s generated and still have the same kind of system. I think that could work. Okay.
373 00:32:38.740 ⇒ 00:32:47.859 Pranav Narahari: Yeah, and at the end of the day, like, them making updates to this doc aren’t making real-time embeddings and updating Andy in real time, right?
374 00:32:50.180 ⇒ 00:32:54.009 Samuel Roberts: At this point, I believe when they make a change…
375 00:32:55.300 ⇒ 00:32:56.730 Mustafa Raja: Yeah, this is real time.
376 00:32:56.730 ⇒ 00:32:58.339 Samuel Roberts: Things are getting re-embedded.
377 00:32:58.500 ⇒ 00:32:59.899 Pranav Narahari: Oh, in real time, okay.
378 00:33:00.290 ⇒ 00:33:01.390 Samuel Roberts: Yeah, there were some.
379 00:33:01.390 ⇒ 00:33:04.009 Mustafa Raja: But it’s only the section that’s being changed, you know.
380 00:33:04.350 ⇒ 00:33:05.280 Samuel Roberts: Yeah.
381 00:33:06.930 ⇒ 00:33:19.059 Mustafa Raja: We have a diff logic, similar to what GitHub would have, where the section that’s been changed. We remove the section and re-embed only that section.
382 00:33:19.990 ⇒ 00:33:21.260 Pranav Narahari: Gotcha.
383 00:33:21.260 ⇒ 00:33:24.230 Samuel Roberts: And that’s happening relatively real time, yeah.
384 00:33:25.890 ⇒ 00:33:27.250 Pranav Narahari: Okay. Well…
385 00:33:27.250 ⇒ 00:33:29.360 Samuel Roberts: some issues, I think, with, like, the… the…
386 00:33:30.790 ⇒ 00:33:33.679 Samuel Roberts: The table of content wasn’t updated or something.
387 00:33:33.680 ⇒ 00:33:35.310 Mustafa Raja: Oh, yeah, so…
388 00:33:35.310 ⇒ 00:33:36.479 Samuel Roberts: I think that’s sorted, right?
389 00:33:38.110 ⇒ 00:33:53.189 Mustafa Raja: And, so the issue is that if they add a new heading, and the table of contents is not updated, because we take our table of contents as our source of truth for sections.
390 00:33:53.330 ⇒ 00:33:58.229 Mustafa Raja: So, that particular heading or section would not be considered a section, but…
391 00:33:58.530 ⇒ 00:34:03.719 Mustafa Raja: part of some other section, until they, refresh the TUC.
392 00:34:08.600 ⇒ 00:34:09.929 Mustafa Raja: Does that make sense?
393 00:34:10.300 ⇒ 00:34:16.100 Samuel Roberts: Yeah, I see, I see… I’m wondering if that’s… I won’t do it going forward or not, with the new embedding pipeline, maybe we…
394 00:34:18.620 ⇒ 00:34:23.960 Samuel Roberts: Have to think about that for the next… Okay.
395 00:34:24.659 ⇒ 00:34:28.159 Samuel Roberts: But yeah, I mean, either way, it was still, yeah, there was still some… it was getting updated.
396 00:34:28.360 ⇒ 00:34:30.660 Samuel Roberts: So, yeah, another approval process.
397 00:34:31.780 ⇒ 00:34:37.059 Samuel Roberts: the… not ideal, but it’s also, like, I think we need something to keep it.
398 00:34:38.020 ⇒ 00:34:39.270 Samuel Roberts: organized.
399 00:34:39.409 ⇒ 00:34:43.959 Samuel Roberts: So, just having them go in and do whatever, we still need something, I think.
400 00:34:44.350 ⇒ 00:34:48.350 Pranav Narahari: Sam, do we already have a design doc on this embedding pipeline?
401 00:34:51.170 ⇒ 00:34:55.660 Samuel Roberts: No, I don’t… I don’t think we have anything for that yet. I imagine…
402 00:34:55.770 ⇒ 00:35:02.459 Samuel Roberts: it’s not gonna be too dissimilar from the, like, overall structure of the N8N, except for the way we do the chunking is gonna be more…
403 00:35:02.570 ⇒ 00:35:06.730 Samuel Roberts: section-based, and it, I don’t think was able to do that at the time, even.
404 00:35:07.260 ⇒ 00:35:08.050 Samuel Roberts: So that’s why.
405 00:35:08.050 ⇒ 00:35:08.590 Pranav Narahari: Gotcha.
406 00:35:08.590 ⇒ 00:35:12.130 Samuel Roberts: Or a token, or character limit, whatever it was.
407 00:35:12.130 ⇒ 00:35:26.579 Pranav Narahari: So for even, like, re-embedding based on updates they make to the dock, that’s going to be a little bit of a different process, too. I don’t know if we want to re-embed every sec- like, the entire section based on changes,
408 00:35:27.250 ⇒ 00:35:32.060 Pranav Narahari: That seems to make sense, it seems like to be the most… seems like the most safe way to do things.
409 00:35:32.300 ⇒ 00:35:32.650 Samuel Roberts: Yeah.
410 00:35:35.580 ⇒ 00:35:46.060 Samuel Roberts: Yeah, whenever they’re making changes, you know, if they’re making changes that reference other things, they’re probably changing that reference as well, or it was already there, so it shouldn’t be…
411 00:35:47.280 ⇒ 00:35:50.609 Samuel Roberts: a problem. Yeah, I don’t think everything needs to get re-embedded.
412 00:35:51.150 ⇒ 00:36:02.480 Pranav Narahari: Yeah, definitely. Okay, maybe we can have a… so this is actually good, because, like, we’ll have a little bit of time after we send this over to them, like, while they’re making updates, we’ll have time to think about how are we gonna…
413 00:36:02.750 ⇒ 00:36:09.030 Pranav Narahari: create this embedding pipeline. I believe that work hasn’t started yet, so, that’ll be next.
414 00:36:12.340 ⇒ 00:36:17.729 Pranav Narahari: Yeah, we can… we can probably talk about this tomorrow, which is… which is good. We’ll have a little bit of time for that.
415 00:36:18.090 ⇒ 00:36:25.279 Pranav Narahari: Yeah, I feel pretty good. I know Amber… I just like.
416 00:36:25.280 ⇒ 00:36:26.520 Amber Lin: Hello!
417 00:36:26.520 ⇒ 00:36:27.040 Samuel Roberts: Hey.
418 00:36:27.040 ⇒ 00:36:40.499 Amber Lin: I’m here to check on, like, if certain details are included, or how we word things. I could serve, like, a first layer of client checks, because I know usually what they’re looking for.
419 00:36:40.630 ⇒ 00:36:57.320 Amber Lin: So Mustafa, if you can send me the link to the drive, I can take a look, because I do want to spot check, like, certain emails, are they included? When they escalate, or when they refer to different departments, are those…
420 00:36:57.470 ⇒ 00:37:02.570 Amber Lin: like, record it correctly, so… I can help do a spot check.
421 00:37:03.060 ⇒ 00:37:09.639 Mustafa Raja: Okay, yeah, it’s in the ABC, drive, in the restructured central docks.
422 00:37:09.780 ⇒ 00:37:11.629 Amber Lin: Okay, let me check.
423 00:37:11.630 ⇒ 00:37:16.990 Mustafa Raja: I’m going to work some more on home improvement. I’m seeing some…
424 00:37:17.410 ⇒ 00:37:20.670 Mustafa Raja: Misinforma- missing information here.
425 00:37:20.790 ⇒ 00:37:23.940 Mustafa Raja: But the lawn one and the rest, I think, should be good.
426 00:37:23.940 ⇒ 00:37:31.519 Amber Lin: I see. Is there, like, a structured way we’re checking for missing content, or are we just doing manual spot checks?
427 00:37:31.720 ⇒ 00:37:33.619 Mustafa Raja: I am doing manual support shots.
428 00:37:33.620 ⇒ 00:37:40.089 Amber Lin: I see. Do you guys think it’ll be helpful to run, like, a workflow to check that? Is that possible?
429 00:37:44.620 ⇒ 00:37:45.530 Samuel Roberts: Yeah, I mean, we could…
430 00:37:45.530 ⇒ 00:37:52.190 Pranav Narahari: You’re saying manual workflow for… Specifically, like, the headings, or for, like, every piece of content?
431 00:37:52.250 ⇒ 00:38:06.469 Amber Lin: I think for every piece of content to make sure we don’t lose their details, right now, I think we’re only able to manually spot check. We have no way to systematically say for sure, hey, we included everything.
432 00:38:08.640 ⇒ 00:38:16.009 Pranav Narahari: Yeah. So, Amber, there’s two ways that I’m thinking about this, and how we can just, like, end goal being accuracy, is,
433 00:38:16.410 ⇒ 00:38:29.499 Pranav Narahari: let’s… we kind of iterate quickly, so we send this over, we’ll get some comments, but at the same time, too, after we embed this, we’ll run it against all of the previous messages that the current Andy is getting.
434 00:38:29.860 ⇒ 00:38:36.519 Pranav Narahari: And then we’ll assess accuracy, and that’s gonna give us more context onto, like, okay, what things might be, like, malformatted.
435 00:38:36.800 ⇒ 00:38:38.060 Amber Lin: Okay, I like that.
436 00:38:39.020 ⇒ 00:38:43.050 Pranav Narahari: These docs are just way too big, like, the commercial one is, like, 200 pages.
437 00:38:43.520 ⇒ 00:38:44.010 Pranav Narahari: like…
438 00:38:44.170 ⇒ 00:38:46.590 Amber Lin: I see. Yeah. Okay.
439 00:38:46.590 ⇒ 00:38:53.719 Samuel Roberts: good, kind of, Automate the manual spot checking, if we wanted, and do, like, something a little broader, but…
440 00:38:53.840 ⇒ 00:38:56.740 Samuel Roberts: I mean, the questions are gonna be the final…
441 00:38:57.120 ⇒ 00:38:57.450 Amber Lin: I see.
442 00:38:57.450 ⇒ 00:38:58.660 Samuel Roberts: Okay, okay.
443 00:38:58.660 ⇒ 00:39:08.200 Amber Lin: Sounds good. So, like, after we embed everything and do the checks, I think our next steps after that would be sizing these down, because they are…
444 00:39:08.340 ⇒ 00:39:09.860 Amber Lin: It’s pretty massive.
445 00:39:12.320 ⇒ 00:39:19.570 Pranav Narahari: So, is there a file size that we’re going for?
446 00:39:19.960 ⇒ 00:39:36.879 Amber Lin: Not really. My… my reasoning for that is because even as we restructured it, we only moved information around. We haven’t consolidated, say, very similar SOPs, or overlapping info, so…
447 00:39:37.110 ⇒ 00:39:40.180 Pranav Narahari: Yeah. We’ll do a check. Yeah.
448 00:39:40.180 ⇒ 00:39:49.439 Samuel Roberts: Yeah, there’s definitely more to do there. I think we did some of the kind of low-hanging fruit, like hours and things like that, that were consolidated, but there’s definitely probably more.
449 00:39:50.890 ⇒ 00:39:55.350 Samuel Roberts: Less structured data that we could… Do the same thing, too.
450 00:39:56.130 ⇒ 00:39:56.630 Amber Lin: Nope.
451 00:39:56.860 ⇒ 00:40:03.689 Amber Lin: Awesome, okay overall looks good to me. I’ll do some spot checks, and then…
452 00:40:03.870 ⇒ 00:40:05.819 Amber Lin: We’ll send this over today.
453 00:40:06.720 ⇒ 00:40:10.089 Pranav Narahari: Yeah, the plan… I’m gonna send this over to them on,
454 00:40:10.410 ⇒ 00:40:12.130 Pranav Narahari: Kind of like the Google Messaging.
455 00:40:12.240 ⇒ 00:40:13.460 Amber Lin: Yay, awesome.
456 00:40:13.460 ⇒ 00:40:14.000 Pranav Narahari: Yeah.
457 00:40:14.000 ⇒ 00:40:25.820 Amber Lin: Maybe also, like, do an email, because sometimes they might… because this is a more formal, like, hey, we did this. It’ll be great if we can have, like, a record.
458 00:40:26.280 ⇒ 00:40:30.049 Pranav Narahari: Yeah, good point. I’ll, I’ll do that as well.
459 00:40:30.290 ⇒ 00:40:40.749 Amber Lin: I think you can find all their emails in the Google Chat, or my, like, weekly summary email has the… has the folks in it.
460 00:40:41.910 ⇒ 00:40:43.209 Pranav Narahari: Okay, so let me…
461 00:40:43.210 ⇒ 00:40:44.890 Amber Lin: Hold them for you.
462 00:40:45.270 ⇒ 00:40:48.270 Pranav Narahari: I was thinking just Janiece and Yvette,
463 00:40:48.600 ⇒ 00:40:50.629 Pranav Narahari: And then I’ll let them probably, like.
464 00:40:50.630 ⇒ 00:40:51.650 Amber Lin: forward it. Okay.
465 00:40:51.650 ⇒ 00:40:52.570 Pranav Narahari: Yeah, toward it.
466 00:40:52.570 ⇒ 00:40:53.620 Amber Lin: Yeah, that works.
467 00:40:53.860 ⇒ 00:40:55.120 Pranav Narahari: Okay, cool.
468 00:40:55.120 ⇒ 00:40:55.770 Amber Lin: Awesome.
469 00:40:56.390 ⇒ 00:41:02.310 Mustafa Raja: Yeah, so… Pranav, you would want me to add the images too, right?
470 00:41:02.810 ⇒ 00:41:04.299 Mustafa Raja: In these documents?
471 00:41:08.400 ⇒ 00:41:19.320 Pranav Narahari: Yeah, I would like that. But actually, you know what? You don’t need to add it to these documents, you know, like, just create a copy, and then re-add them.
472 00:41:19.430 ⇒ 00:41:31.099 Pranav Narahari: I’ll just send over this email to them, like, right after this call. So then, maybe, like, yeah, later today, and then tomorrow you can just, like, re-add those images. Just because I don’t want that to be, like…
473 00:41:31.100 ⇒ 00:41:36.730 Mustafa Raja: Yeah, I’m going to update this one, this home improvement one. I saw some missing information, so I’ll just…
474 00:41:37.760 ⇒ 00:41:41.859 Mustafa Raja: Maybe, wait a few, minutes after this call.
475 00:41:41.980 ⇒ 00:41:42.790 Mustafa Raja: Okay.
476 00:41:42.790 ⇒ 00:41:45.740 Pranav Narahari: How long do you think it would take to get all the images back in?
477 00:41:46.840 ⇒ 00:41:52.849 Mustafa Raja: I just don’t know that, you know, I’m going to ask her, so I don’t know how long it’s going to take.
478 00:41:53.180 ⇒ 00:41:59.559 Pranav Narahari: Okay, yeah, so I’ll give you, like, I think by tomorrow, we can probably get that done for sure, like, conservatively, right?
479 00:42:00.760 ⇒ 00:42:08.999 Pranav Narahari: Yeah, so I’ll just send over the email without the images then. But yeah, let me know once you’re done with the home improvement, or if there are any other things that you want to QA.
480 00:42:09.480 ⇒ 00:42:12.720 Pranav Narahari: And then once you kind of give me that final, like.
481 00:42:12.840 ⇒ 00:42:14.860 Pranav Narahari: Just, like, thumbs up, I can send it over.
482 00:42:15.730 ⇒ 00:42:16.400 Mustafa Raja: Okay.
483 00:42:21.250 ⇒ 00:42:26.919 Pranav Narahari: Cool, guys. Oh, Sam, one other thing that I wanted to ask about, I just thought of it, but then we kind of…
484 00:42:27.090 ⇒ 00:42:28.270 Pranav Narahari: change topics.
485 00:42:28.680 ⇒ 00:42:29.480 Pranav Narahari: with…
486 00:42:30.360 ⇒ 00:42:36.240 Pranav Narahari: Is a good way to also assess accuracy is just to, like, see if there’s, like… is there a way that we can just, like…
487 00:42:37.560 ⇒ 00:42:41.070 Pranav Narahari: pull… When we’re doing, like, the…
488 00:42:41.220 ⇒ 00:42:51.960 Pranav Narahari: the similarity check. And I’m sure there is a way to do this, just kind of see, like, what pieces of information are coming in, and then we do, like, a… like, LLM as a judge to see if they’re…
489 00:42:52.290 ⇒ 00:42:59.979 Pranav Narahari: the same piece of information. That way, we can know if, like, okay, are we just duplicating information, or is there conflicting information?
490 00:42:59.980 ⇒ 00:43:02.130 Samuel Roberts: You’re saying when we do the vector lookup?
491 00:43:02.340 ⇒ 00:43:03.180 Pranav Narahari: Yeah.
492 00:43:05.800 ⇒ 00:43:11.580 Samuel Roberts: Mmm, okay, so you’re saying, okay, okay, so you’re saying, like, as part of the Andy… Pipeline.
493 00:43:12.900 ⇒ 00:43:18.909 Samuel Roberts: vast, it… or just, like, whatever, however the query is generated for a vector…
494 00:43:18.910 ⇒ 00:43:26.460 Pranav Narahari: I don’t think we need to do it on, like, in actual production. I think it can be just something for us to do, like, one time, against all of the previous messages.
495 00:43:26.710 ⇒ 00:43:30.139 Samuel Roberts: I think that’s probably doable, I mean, and I’m not sure…
496 00:43:30.930 ⇒ 00:43:34.669 Samuel Roberts: I mean, in Maestra, I think we can definitely do that now. I don’t know what the N8N…
497 00:43:35.410 ⇒ 00:43:38.889 Samuel Roberts: How that returns it, but we could probably tag that on there, too, if we even needed to…
498 00:43:39.200 ⇒ 00:43:42.049 Samuel Roberts: Verify how it was doing it versus how it is doing it now.
499 00:43:42.850 ⇒ 00:43:45.539 Pranav Narahari: I’ll create a task for that. Me and you can talk about that tomorrow.
500 00:43:45.540 ⇒ 00:43:46.300 Samuel Roberts: Yeah.
501 00:43:48.300 ⇒ 00:43:49.329 Samuel Roberts: That’s a good call.
502 00:43:52.800 ⇒ 00:43:55.200 Pranav Narahari: Because I’m less concerned with our…
503 00:43:55.720 ⇒ 00:44:05.460 Pranav Narahari: What we’re most interested in, which is accuracy and execution time for duplicate information, But conflicting information is where…
504 00:44:05.570 ⇒ 00:44:10.720 Pranav Narahari: You know, accuracy is just gonna be… It’s just gonna go down.
505 00:44:10.720 ⇒ 00:44:15.470 Samuel Roberts: Yeah, yeah, I mean, the duplicate versus conflicting, I feel like, is hard to…
506 00:44:16.120 ⇒ 00:44:22.219 Samuel Roberts: know in the current form of the doc sometimes. Yeah. If we can get it out, we can find out, yeah.
507 00:44:22.220 ⇒ 00:44:27.609 Pranav Narahari: Yeah, and we’ll get two birds with one stone with this, because we’ll find both the duplicate and conflicting.
508 00:44:29.370 ⇒ 00:44:30.150 Pranav Narahari: Yup.
509 00:44:30.840 ⇒ 00:44:32.490 Samuel Roberts: Okay, yeah, good, good call.
510 00:44:33.090 ⇒ 00:44:35.650 Pranav Narahari: Cool. Yeah, I’ll create a ticket for that so we don’t forget.
511 00:44:36.640 ⇒ 00:44:38.589 Pranav Narahari: Yeah, but other than that,
512 00:44:39.100 ⇒ 00:44:45.590 Pranav Narahari: Mustafa, you’ll just give me a final update on home improvement when that’s done, and then we’re good to send off that email.
513 00:44:46.290 ⇒ 00:44:46.880 Mustafa Raja: Nope.
514 00:44:47.490 ⇒ 00:44:59.049 Uttam Kumaran: One, one random thought as I was just listening in, we’re facing, actually, similar things in the platform, and one thing that we’re doing on the cleanup side is I want to run a GitHub action.
515 00:44:59.170 ⇒ 00:45:00.330 Uttam Kumaran: every week.
516 00:45:00.450 ⇒ 00:45:05.050 Uttam Kumaran: that looks for… Broken links.
517 00:45:05.340 ⇒ 00:45:09.170 Uttam Kumaran: Duplicate information and proposes consolidation.
518 00:45:09.900 ⇒ 00:45:12.160 Uttam Kumaran: It’s probably something that you guys can just run.
519 00:45:12.720 ⇒ 00:45:13.740 Uttam Kumaran: Yeah.
520 00:45:14.000 ⇒ 00:45:18.149 Uttam Kumaran: The second piece is, like, we’re also thinking about this markdown problem.
521 00:45:18.240 ⇒ 00:45:19.650 Samuel Roberts: And Google Doc?
522 00:45:21.640 ⇒ 00:45:28.609 Uttam Kumaran: now that you have Google Workspace, like, CLI, it’s actually, like, really easy to write to Google Docs and Google Appsets.
523 00:45:29.110 ⇒ 00:45:34.360 Uttam Kumaran: But I’ve also heard people thinking about using, like, Obsidian, headless Obsidian.
524 00:45:34.860 ⇒ 00:45:41.829 Uttam Kumaran: So, like, the lore on the Google Doc is really just, like, this company uses Google a lot.
525 00:45:42.450 ⇒ 00:45:46.990 Uttam Kumaran: I think, yes, Sam’s right on, like, you have to maintain some type of two-way sync.
526 00:45:47.890 ⇒ 00:45:53.370 Uttam Kumaran: So, there’s a couple of ways of, like, tackling that problem.
527 00:45:53.600 ⇒ 00:46:01.889 Uttam Kumaran: You know, like, one way could be they get a version, we get a version, changes can only be made during a certain window where there’s a sync.
528 00:46:02.220 ⇒ 00:46:05.449 Uttam Kumaran: You could implement a new tool, like Obsidian.
529 00:46:05.740 ⇒ 00:46:09.509 Uttam Kumaran: Where there is… there is… they… they basically do,
530 00:46:10.150 ⇒ 00:46:13.810 Uttam Kumaran: There’s sort of, like, a round-robin structure for…
531 00:46:14.250 ⇒ 00:46:17.729 Uttam Kumaran: Merge… merging, so they can maintain two-way sync.
532 00:46:18.780 ⇒ 00:46:20.779 Uttam Kumaran: And then, I think the cleanup step
533 00:46:21.070 ⇒ 00:46:23.560 Uttam Kumaran: You should totally just run on, like.
534 00:46:23.900 ⇒ 00:46:29.059 Uttam Kumaran: Every Sunday, you should write a GitHub action that just goes and checks and sees that there’s garbage in there.
535 00:46:31.300 ⇒ 00:46:31.960 Pranav Narahari: Yeah.
536 00:46:32.820 ⇒ 00:46:37.940 Pranav Narahari: I think that’s… And this is the time to do it, too, when, like, we have a fresh dock.
537 00:46:38.350 ⇒ 00:46:43.179 Pranav Narahari: As they make more changes, like, we don’t want it to start getting messy again, so…
538 00:46:43.310 ⇒ 00:46:49.370 Pranav Narahari: Yeah, I’ll, we should definitely talk about that tomorrow as well, Sam.
539 00:46:50.310 ⇒ 00:46:53.310 Samuel Roberts: No, that’s… that’s… that’s a good call. I think, yeah.
540 00:46:53.490 ⇒ 00:46:54.740 Samuel Roberts: once we…
541 00:46:56.510 ⇒ 00:47:05.760 Samuel Roberts: Yeah, I just… we’ll have to figure out, because like I was saying before, there’s some stuff, like, where it’s conflicting versus duplicate versus very similar but distinct information.
542 00:47:06.470 ⇒ 00:47:14.910 Samuel Roberts: So that’s… that’s just the only pitfall there, but yeah, I think even just running an action that will show us what’s going on is good, especially if there’s changes being made more frequently.
543 00:47:17.750 ⇒ 00:47:21.830 Samuel Roberts: Or we don’t have a system where they’re letting us know when they make changes, even.
544 00:47:22.270 ⇒ 00:47:24.009 Samuel Roberts: Either way, like, it’s worth having.
545 00:47:28.710 ⇒ 00:47:32.429 Samuel Roberts: Yeah, I would say throw a ticket for that, we can discuss that.
546 00:47:34.400 ⇒ 00:47:34.910 Samuel Roberts: Make that.
547 00:47:34.910 ⇒ 00:47:35.530 Pranav Narahari: Yup.
548 00:47:41.070 ⇒ 00:47:41.850 Samuel Roberts: Alrighty.
549 00:47:44.480 ⇒ 00:47:45.459 Pranav Narahari: Cool, guys, I feel good.
550 00:47:45.460 ⇒ 00:47:46.689 Samuel Roberts: Excuse me. Okay.
551 00:47:49.070 ⇒ 00:47:49.510 Uttam Kumaran: Okay.
552 00:47:49.510 ⇒ 00:47:50.899 Samuel Roberts: I’m gonna hop in, unless you need it.
553 00:47:50.900 ⇒ 00:47:51.720 Uttam Kumaran: Thank you.
554 00:47:52.290 ⇒ 00:47:53.460 Pranav Narahari: Perfect. Yeah, thanks, guys.
555 00:47:53.460 ⇒ 00:47:54.150 Samuel Roberts: Very cool.
556 00:47:54.830 ⇒ 00:47:57.439 Uttam Kumaran: Sam, one question. Sam, I just had a quick question.
557 00:47:57.440 ⇒ 00:47:58.260 Samuel Roberts: Yeah, yeah.
558 00:47:58.260 ⇒ 00:48:03.579 Uttam Kumaran: On the GitHub token thing, yeah. Yeah, I thought I messed something up.
559 00:48:03.860 ⇒ 00:48:05.230 Uttam Kumaran: And then…
560 00:48:05.230 ⇒ 00:48:11.670 Samuel Roberts: I think… Maybe not. I think what happened was I… I was trying to get rid of the next one, I might not have renamed it in Railway.
561 00:48:14.310 ⇒ 00:48:22.270 Samuel Roberts: Or on, 1Password for running the env. So I made that change, so it shouldn’t need the fallback anymore, I think.
562 00:48:24.820 ⇒ 00:48:25.450 Uttam Kumaran: Okay.
563 00:48:25.640 ⇒ 00:48:26.419 Samuel Roberts: I can double check it.
564 00:48:27.050 ⇒ 00:48:30.860 Samuel Roberts: Yeah, do you want to just double check? Because I’ll revert my PR, I just didn’t see…
565 00:48:31.100 ⇒ 00:48:35.299 Uttam Kumaran: the… Little app thing running, so…
566 00:48:35.850 ⇒ 00:48:37.390 Samuel Roberts: Yeah, I think,
567 00:48:39.570 ⇒ 00:48:44.479 Samuel Roberts: I mean, there’s other stuff in that PR, I don’t know if you want to completely revert it, or just make that file change, maybe.
568 00:48:48.300 ⇒ 00:48:52.239 Uttam Kumaran: I was gonna… I was gonna listen to your comment and revert everything else.
569 00:48:52.720 ⇒ 00:48:57.969 Samuel Roberts: Okay, that’s fine. Yeah, I think it was just, like, that one file, really, or one or two files, because there was a bunch of other stuff.
570 00:48:57.970 ⇒ 00:49:00.470 Uttam Kumaran: Like, should I… but… but I guess, like,
571 00:49:00.940 ⇒ 00:49:06.919 Uttam Kumaran: should I… like, does that… that shouldn’t affect the live… Right now, right?
572 00:49:08.540 ⇒ 00:49:12.779 Samuel Roberts: No, what you did shouldn’t affect it. No, I went in and made the change, so if it wasn’t getting…
573 00:49:13.020 ⇒ 00:49:16.249 Samuel Roberts: If something wasn’t running anymore, it now should be running again.
574 00:49:16.930 ⇒ 00:49:18.339 Samuel Roberts: I’ll double-check it, though.
575 00:49:19.130 ⇒ 00:49:20.090 Uttam Kumaran: Okay, okay.
576 00:49:21.030 ⇒ 00:49:28.200 Samuel Roberts: Yeah, I made a proxy, so basically the GitHub token’s just stored in the backend, and now the calls go through that for the marketing stuff, but I might have just missed where that was.
577 00:49:28.340 ⇒ 00:49:30.519 Samuel Roberts: Important for the transcript sync, too.
578 00:49:31.360 ⇒ 00:49:36.090 Samuel Roberts: Well, no, the transcripts in… Doesn’t need the GitHub token, it’s using the other stuff, right?
579 00:49:37.160 ⇒ 00:49:41.929 Samuel Roberts: I’ll take a look. I’ll dig in. I wasn’t digging in too much when I said that, I just was like, wait a minute, and I…
580 00:49:42.310 ⇒ 00:49:43.810 Samuel Roberts: I’ll double check it.
581 00:49:44.490 ⇒ 00:49:45.070 Uttam Kumaran: Okay.
582 00:49:46.250 ⇒ 00:49:47.380 Samuel Roberts: Cool, I’ll let you know.
583 00:49:47.960 ⇒ 00:49:49.349 Uttam Kumaran: Alright, thank you guys.
584 00:49:49.460 ⇒ 00:49:50.150 Samuel Roberts: Alright.