Meeting Title: AI Service Daily Recap - Blockers + Realign Date: 2026-05-06 Meeting participants: Pranav, Casie Aviles, Mustafa Raja


WEBVTT

1 00:01:32.040 00:01:33.100 Casie Aviles: Hey, Bradnov.

2 00:01:36.330 00:01:37.080 Pranav: Hey, Casey.

3 00:01:38.260 00:01:41.830 Casie Aviles: Let me ping Mustafa if he’s coming.

4 00:02:10.699 00:02:14.740 Pranav: Did Sam join the… Call a few minutes ago when you guys were there?

5 00:02:15.630 00:02:19.340 Casie Aviles: Oh, I think… no, no, I think he’s out. He’s out for now.

6 00:02:20.370 00:02:24.220 Pranav: Okay, did he… I didn’t see any message. I might have missed it.

7 00:02:29.190 00:02:29.970 Mustafa Raja: Hey…

8 00:02:35.190 00:02:36.410 Mustafa Raja: Hey, can you hear me?

9 00:02:36.800 00:02:37.490 Pranav: Yep, we can hear you.

10 00:02:37.490 00:02:38.040 Casie Aviles: Yes.

11 00:02:40.780 00:02:41.370 Mustafa Raja: Cool.

12 00:02:42.800 00:02:47.539 Pranav: Okay. Did, either of you guys sync up with Sam to see where he left things off?

13 00:02:50.030 00:02:52.519 Mustafa Raja: Well, I didn’t get a chance to talk to him, so…

14 00:02:53.050 00:02:53.660 Pranav: Okay.

15 00:02:54.900 00:02:59.730 Casie Aviles: Yeah, I think it’s just the last message he sent in the channel, that that’s…

16 00:02:59.940 00:03:01.530 Casie Aviles: The last one I saw.

17 00:03:04.630 00:03:07.419 Pranav: Which is just, like, the… the HTML.

18 00:03:08.100 00:03:09.000 Casie Aviles: Yeah, yeah.

19 00:03:09.170 00:03:15.309 Pranav: Okay. Yeah, that’s fine. Alright, how are… how are things going with, the tickets for today?

20 00:03:17.940 00:03:20.879 Mustafa Raja: Yeah, I have, I have the workflow,

21 00:03:21.210 00:03:27.829 Mustafa Raja: almost good. I’m going to test it, and then I’m going to share results, and then we can configure…

22 00:03:28.220 00:03:31.700 Mustafa Raja: what’s it called? A Gmail account with it.

23 00:03:31.850 00:03:35.359 Mustafa Raja: That we would want to send the email through, you know?

24 00:03:36.020 00:03:41.870 Mustafa Raja: So, do you know which email… which Gmail account is it going to be? Is it your Eden account?

25 00:03:42.630 00:03:44.550 Pranav: To send from or to send to?

26 00:03:44.740 00:03:45.970 Mustafa Raja: to send from.

27 00:03:46.860 00:03:48.570 Pranav: Yeah, you can send it for mine.

28 00:03:49.320 00:03:52.500 Mustafa Raja: And whom am I going to send that to?

29 00:03:52.930 00:03:54.510 Pranav: Should send to Danny.

30 00:03:55.360 00:03:56.010 Mustafa Raja: Oh, okay.

31 00:03:56.920 00:04:14.789 Mustafa Raja: Okay, yeah, so once… once I test it, I’ll share results within the channel, and then if the results are good, I’ll move to the next step, that is, you know, configuring it with Gmail. We were talking about this in the morning, that, at least for first two weeks, we might want to have a human in the loop.

32 00:04:16.480 00:04:17.319 Mustafa Raja: You know?

33 00:04:17.860 00:04:19.469 Mustafa Raja: Does that make sense with you?

34 00:04:19.860 00:04:31.999 Pranav: Yeah, so basically it just, like, creates the email. Yeah, I like that. How about what we do is, let’s still wire it up fully, but instead of sending it automatically from my account, let’s just put it into the drafts.

35 00:04:33.370 00:04:35.190 Mustafa Raja: Oh, yeah, that works. Yeah, yeah.

36 00:04:35.310 00:04:36.429 Mustafa Raja: Yeah, we could’ve done.

37 00:04:36.730 00:04:37.500 Pranav: Okay, cool.

38 00:04:37.680 00:04:42.479 Pranav: That way, it should be pretty simple to just… Flip it to AutoSend.

39 00:04:43.150 00:04:43.960 Mustafa Raja: Yeah.

40 00:04:45.230 00:04:48.399 Mustafa Raja: Yeah, once we feel it is consistent, we can just, you know…

41 00:04:51.350 00:04:52.850 Pranav: Okay. Cool.

42 00:04:53.140 00:05:00.860 Pranav: Okay, yeah, that sounds great. Casey, I know you were working on some ABC stuff. It sounds like you embedded both those documents.

43 00:05:01.310 00:05:01.810 Pranav: Yeah.

44 00:05:01.810 00:05:03.339 Casie Aviles: Yeah, they should be good.

45 00:05:03.620 00:05:04.180 Pranav: Cool.

46 00:05:05.270 00:05:06.570 Pranav: Okay. Any other…

47 00:05:06.570 00:05:08.430 Mustafa Raja: We identified a new issue with it.

48 00:05:08.570 00:05:09.189 Mustafa Raja: Kisha, do you want.

49 00:05:09.190 00:05:09.700 Casie Aviles: Oh, yeah.

50 00:05:10.350 00:05:11.320 Mustafa Raja: Internet.

51 00:05:12.150 00:05:21.769 Casie Aviles: Yeah, we did, so, when I was adding the documents, we were… we needed to add… or we needed to fix, like, the tags for the departments.

52 00:05:23.050 00:05:30.849 Casie Aviles: So I’ll just need to add that, but for now, we did, like, a quick workaround that just lets Andy

53 00:05:31.420 00:05:35.690 Casie Aviles: At least be able to, like, retrieve it from the knowledge base.

54 00:05:39.100 00:05:41.039 Pranav: Okay, what was the issue with tags?

55 00:05:42.870 00:05:44.039 Casie Aviles: It wasn’t…

56 00:05:44.890 00:05:55.120 Casie Aviles: Because for the universal documents, it was being treated as, like, one complete document, if that… if I’m… let me know if I’m wrong there, Mustafa, but yeah.

57 00:05:56.430 00:06:16.349 Mustafa Raja: Yeah, so, we have multiple universal docs, right? And when embedding pipeline, when it tries to calculate the diff between new and old, it considers all of those universal documents as one single document, because in the department section, we are naming them the same.

58 00:06:16.790 00:06:17.420 Mustafa Raja: You know?

59 00:06:17.420 00:06:21.670 Pranav: Oh, so it’s basically trying to re-embed all of them when we add a new doc? Is that what you’re saying?

60 00:06:23.170 00:06:41.750 Mustafa Raja: So the pipeline runs every day, listening for any changes, right? So, it just re… it just thinks, okay, because something changed in this other document, it considers all of those documents into one big document, right?

61 00:06:43.020 00:07:01.430 Mustafa Raja: So, what could happen is two different documents may have the same, heading, right? And that could create a confusion, and, you know, we might miss some content. So, we may need, some better, department names for,

62 00:07:01.510 00:07:04.889 Mustafa Raja: for the Central, sorry, for the Universal Dogs.

63 00:07:06.260 00:07:07.230 Pranav: Okay, okay.

64 00:07:07.230 00:07:10.800 Mustafa Raja: just named them all as Universal Doc, is what I’m trying to say.

65 00:07:12.530 00:07:16.240 Pranav: Okay, but they do currently have different names, right? Like, some of them say Pricing…

66 00:07:16.240 00:07:28.620 Mustafa Raja: The documents themselves have the different names, but in the metadata, they are being tagged as… the tagged department is universal info.

67 00:07:28.870 00:07:30.879 Mustafa Raja: Right? Oh, that’s what we probably do.

68 00:07:30.880 00:07:42.350 Pranav: Because having a department there doesn’t make sense. We need to have, like, a tagged topic, I guess. And then we can just have that be the title of the document.

69 00:07:42.910 00:07:57.830 Mustafa Raja: I mean, the pipeline is set to, you know, look into the department. I think the department works just as tag, and we can just update it and have each document have its own

70 00:07:58.520 00:08:00.119 Mustafa Raja: Tag in the, you know?

71 00:08:01.660 00:08:06.240 Pranav: But then, that means we had to add a tag every time a new document is in here.

72 00:08:07.610 00:08:11.969 Mustafa Raja: Yeah, yeah, that is correct. Does that not make sense?

73 00:08:12.590 00:08:19.800 Pranav: That seems like, kind of a burden for us, right? Because every time we add a tag, we have to redeploy the app, right?

74 00:08:23.790 00:08:24.250 Casie Aviles: Hmm.

75 00:08:24.250 00:08:25.959 Mustafa Raja: Every time we add a tag,

76 00:08:26.700 00:08:29.939 Mustafa Raja: Yes, every time we add a document, yeah.

77 00:08:30.670 00:08:39.699 Pranav: Yeah, so we don’t want that process. We just should have a process where we can just add the document in here, and then it auto… auto-embeds… Embeds.

78 00:08:39.900 00:08:43.980 Pranav: Honestly. So…

79 00:08:43.980 00:08:46.340 Mustafa Raja: I mean, I don’t see…

80 00:08:46.340 00:08:51.700 Pranav: why the tagging is super important here. Like, what is the reason for the tags? Like, how does that help with retrieval?

81 00:08:51.700 00:09:15.579 Mustafa Raja: Yeah, so, the tags here help, help Andy, look into specific departments only. So, if a CSR goes on to ask a question specific… specifically about lawn department, with this metadata, Andy is able to look into specifically lawn department, right?

82 00:09:15.740 00:09:19.810 Mustafa Raja: If we remove this, and he loses that ability.

83 00:09:25.020 00:09:26.979 Pranav: Okay, yeah.

84 00:09:29.630 00:09:30.160 Casie Aviles: Yeah, this…

85 00:09:30.160 00:09:34.549 Mustafa Raja: Also, what we should also do is,

86 00:09:35.100 00:09:41.520 Mustafa Raja: we have to manually add IDs of all of the documents that we want to embed. Maybe what should happen is…

87 00:09:41.760 00:09:45.020 Mustafa Raja: They should look into just the, folder.

88 00:09:45.220 00:09:48.870 Mustafa Raja: And automatically get all of the… all of the documents, you know.

89 00:09:50.810 00:09:53.419 Mustafa Raja: Yeah. Rather than us having to give IDs, yeah.

90 00:09:53.840 00:09:57.469 Pranav: Yeah, we definitely shouldn’t support it like that.

91 00:09:58.000 00:10:02.199 Mustafa Raja: And maybe this department tag could live within the document also.

92 00:10:05.770 00:10:07.219 Pranav: So what is the one drop that you guys do?

93 00:10:07.220 00:10:13.230 Mustafa Raja: The benefit of that would be, we wouldn’t have to update code, all the time, you know.

94 00:10:14.220 00:10:17.310 Mustafa Raja: Sorry, I couldn’t get your question.

95 00:10:17.310 00:10:19.180 Pranav: What is the workout we did?

96 00:10:20.490 00:10:24.439 Mustafa Raja: So let’s… can we go back to the thing?

97 00:10:25.480 00:10:27.700 Mustafa Raja: Yeah, so if you… yeah, so…

98 00:10:27.900 00:10:35.569 Mustafa Raja: This was tagged as Universal Info, and we tagged it as ops, and it did work, you know?

99 00:10:38.260 00:10:40.920 Pranav: So why did that work and not universal info? I’m still not understanding that.

100 00:10:40.920 00:10:47.240 Mustafa Raja: Because, you know, you see that a lot of these documents have department universal info.

101 00:10:47.680 00:10:48.390 Pranav: Yeah.

102 00:10:49.140 00:10:56.119 Mustafa Raja: So, when our embedding pipeline calculates the diff between the new content and the old content.

103 00:10:56.220 00:11:00.279 Mustafa Raja: It considers all of these documents as one single document.

104 00:11:00.670 00:11:09.989 Mustafa Raja: And these documents might have… these documents might have common headings, or common TOCs, or common sections that might confuse the diff section, you know?

105 00:11:12.090 00:11:12.880 Pranav: Okay.

106 00:11:13.630 00:11:28.710 Mustafa Raja: So that is… that is why, it was just not capturing the content for this, one of the new documents, and we just changed the department so it doesn’t, consider, consider it as universal info.

107 00:11:28.920 00:11:30.390 Mustafa Raja: So…

108 00:11:30.390 00:11:37.480 Pranav: What we should probably do here, then, is just maintain one document, which just includes all of the… all of the information in there.

109 00:11:37.480 00:11:40.120 Mustafa Raja: Yeah, that also makes sense, yeah, for the universal info, right?

110 00:11:40.120 00:11:44.990 Pranav: We should just have it similar to what we have for all of the other,

111 00:11:45.390 00:11:48.219 Pranav: Yeah, basically for all of the other departments.

112 00:11:48.330 00:12:01.380 Pranav: So, yeah, let’s, I like this workaround for now. If that’s working, then we can… we can leave it as is. For all new docs that come in, I’ll just let them know, hey, we need to start… we need to allocate some time to…

113 00:12:01.750 00:12:06.459 Pranav: Migrate all of these, like, individual docs into a certain dock, into a single dock.

114 00:12:07.260 00:12:10.840 Mustafa Raja: Yeah, we might want to consolidate in… Linda.

115 00:12:11.220 00:12:11.750 Pranav: Yep.

116 00:12:12.630 00:12:17.690 Casie Aviles: So that would also prevent us from… so we’ll… we can avoid, like, adding this…

117 00:12:17.800 00:12:20.750 Casie Aviles: Every time, you know, we can just point it to one document.

118 00:12:21.050 00:12:25.310 Casie Aviles: And then we can just run the embedding again. Yeah, that makes sense.

119 00:12:25.530 00:12:27.970 Pranav: Okay, perfect. Yeah, let me create a ticket for that.

120 00:12:58.870 00:13:02.089 Pranav: Alright, cool. Anything else, guys?

121 00:13:04.280 00:13:06.830 Mustafa Raja: This is all from, Tide.

122 00:13:07.580 00:13:13.440 Casie Aviles: Yeah, that’s all I had for now, but, so I am seeing these new tickets, so this will be for…

123 00:13:13.580 00:13:18.029 Casie Aviles: this, this cycle to red, I’ll work on this as well.

124 00:13:18.440 00:13:21.719 Pranav: Yeah, I mean, we’re not really defining cycles here.

125 00:13:22.980 00:13:23.630 Casie Aviles: Okay.

126 00:13:23.890 00:13:28.580 Pranav: I’m just adding it to current cycle because it’s a priority, it’s not just something we should throw on the backlog.

127 00:13:30.120 00:13:30.940 Casie Aviles: Okay.

128 00:13:33.290 00:13:34.130 Pranav: Okay.

129 00:13:34.530 00:13:36.010 Pranav: Cool, guys. Anything else?

130 00:13:39.040 00:13:40.460 Casie Aviles: No, that’s all.

131 00:13:41.180 00:13:41.820 Pranav: Alright.

132 00:13:42.110 00:13:43.429 Pranav: Okay, talk soon.

133 00:13:43.540 00:13:44.440 Pranav: See ya.

134 00:13:45.390 00:13:45.960 Casie Aviles: Thank you.

135 00:17:14.790 00:17:15.470 Mustafa Raja: Hmm.