slack 聊天机器人

by Bertrand Fan

通过范伯特朗

无法筹集资金的Slack机器人 (Unfundable Slack bots)

In December, Slack announced a $80 million fund to invest in software projects that complement its technology. As an early adopter of the Slack API, here are some “bets” that I’ve made on the Slack platform:

去年12月，Slack宣布了8000万美元的基金，用于投资补充其技术的软件项目。 作为Slack API的早期采用者，以下是我在Slack平台上进行的一些“下注”：

每10秒钟播放一帧自动播放《星球大战特别版》(汉斯拍摄的第一张！)的机器人。 整个过程大约需要20个小时。 (A bot that plays all of Star Wars Despecialized Edition (Han shoots first!) one frame every ten seconds. It takes about 20 hours to get through the whole thing.)

托管“抵抗”游戏的机器人(可以选择使用“隐藏议程”扩展程序中的“刺客模块”)。 秘密投票是通过bot通过DM进行的，而公共操作则在渠道中进行。 (A bot that hosts games of The Resistance (optionally with the Assassin Module from the Hidden Agenda expansion). Secret voting is conducted via DM with the bot, while public actions are done in the channel.)

通过发出命令(向左，向右，向上，向下，打开和开火)让您玩Wolfenstein 3D的机器人。 您可以选择指定向左和向右旋转多少度，但是默认情况下您可以旋转45度。 我从未成功完成第一关，但我设法杀死了几名后卫。 (A bot that lets you play Wolfenstein 3D by issuing commands (left, right, up, down, open, and fire). You can optionally specify how many degrees to turn left and right, but by default you turn 45-degrees. I’ve never successfully completed the first level, but I have managed to kill several guards.)

I’m happy to announce that I’ve completed my latest bot, Vandelay Industries. Vandelay Industries is a shameless rip-off of Frinkiac that lets you search for animated gifs from every episode of Seinfeld in Slack. It has every line of dialogue ever spoken in Seinfeld.

我很高兴地宣布，我已经完成了最新的机器人Vandelay Industries 。 Vandelay Industries是Frinkiac的无耻剥夺者 ，它使您可以从Slack的Seinfeld的每一集中搜索GIF动画。 它拥有塞恩费尔德曾经说过的所有对话方式。

If you’d like to try it out in your Slack team, just go to Vandelay Industries and click Add to Slack. It will ask you for permission to add a new slash command, /vandelay and once authorized, it should be ready to use immediately within Slack.

如果您想在Slack团队中尝试一下，只需转到Vandelay Industries ，然后点击添加到Slack即可 。 它会要求您添加新的斜杠命令/ vandelay的权限，并且一旦获得授权，就应该可以立即在Slack中使用它。

If you’re interested in the technical details of how it works, keep on reading! If not, you’ve managed to beat the estimated reading time for this article. Congratulations!

如果您对它如何工作的技术细节感兴趣，请继续阅读！ 如果没有，您已经超过了本文的预计阅读时间。 恭喜你！

处理中 (Processing)

I initially started with 91G of all nine seasons of Seinfeld in surprisingly high quality 720p. I ultimately ended up with 111G worth of animated GIFs which tells you everything you need to know about the efficiency of that file format. This section is all about how I went from one to the other.

最初，我以整个720p的高画质开始使用Seinfeld的所有九个季节的91G。 我最终得到了价值111G的动画GIF，可以告诉您有关该文件格式效率的所有信息。 这一部分是关于我如何从一个人到另一个人的。

The files were in MKV containers, so I was able to use MKVToolNix to extract the subtitles from them. You can use mkvinfo to list the various segment tracks:

这些文件位于MKV容器中，因此我能够使用MKVToolNix从它们中提取字幕。 您可以使用mkvinfo列出各种段轨道：

mkvinfo Seinfeld.S01E01.The.Seinfeld.Chronicles.mkv

It’s a little hard to read, but the track that we’re interested in is Track number 3, the first subtitles track. After noting the track ID (2), we can extract the subtitles into an SRT file like this:

有点难读，但是我们感兴趣的音轨是音轨编号3，第一个字幕音轨。 注意轨道ID(2)之后，我们可以将字幕提取到SRT文件中，如下所示：

mkvextract tracks Seinfeld.S01E01.The.Seinfeld.Chronicles.mkv 2:S01E01.srt

The SRT format is fairly straightforward. It contains a counter, start time, end time, and the subtitle text. Using a parser like subtitles-parser, we can easily iterate over the subtitles.

SRT格式相当简单。 它包含一个计数器，开始时间，结束时间和字幕文本。 使用像subtitles- parser这样的解析器 ，我们可以轻松地对字幕进行迭代。

The next step is to loop through each subtitle and extract that time range from the MKV into an animated GIF. There’s an excellent article about using ffmpeg to encode high quality GIF but if you don’t want to read that right now, the trick is to extract a specialized palette from the section of the video that you’re interested in and then use that to encode the GIF.

下一步是遍历每个字幕，并将该时间范围从MKV提取到动画GIF中。 有一篇关于使用ffmpeg编码高质量GIF的出色文章，但是如果您现在不想阅读，诀窍是从您感兴趣的视频部分中提取专门的调色板，然后将其用于编码GIF。

Here’s a script that I adapted for the purposes of this processing step:

这是我为了该处理步骤而修改的脚本：

320 refers to how many pixels wide the resulting GIF will be. You’ll notice that instead of specifying a start time and an end time, I instead specify a duration. Although ffmpeg claims to support end times, no matter which version I tried I could not get it to properly extract the right range, so I ended up calculating the duration by subtracting the start time from the end time and abusing the unix epoch like this:

320表示生成的GIF将有多少像素宽。 您会注意到，我没有指定开始时间和结束时间，而是指定了持续时间。 尽管ffmpeg声称支持结束时间，但是无论我尝试使用哪个版本，我都无法使其正确提取正确的范围，因此最终我通过从结束时间中减去开始时间并滥用unix历元来计算持续时间，如下所示：

After applying the gifenc.sh script, we’re left with a nice animated GIF of the correct extracted range like this:

应用gifenc.sh脚本后，剩下的动画GIF正确提取范围如下：

But I wanted to display the subtitle text at the bottom of the GIF and after digging through the ImageMagick documentation, I came up with this:

但是我想将字幕文本显示在GIF的底部，并且在浏览了ImageMagick文档后，我想到了：

It’s not the most elegant solution, but it gets the job done. Our final gif looks like this:

这不是最优雅的解决方案，但可以完成工作。 我们的最终gif如下所示：

Now just do that about 104,782 more times and you’re done. Running an entire episode through ffmpeg and ImageMagick took about 30 minutes on my top-of-the-line-in-2011 Macbook Pro. This is the part of the story where I’d like to tell you that I managed to leverage Amazon Elastic Transcoder or made a hadoop job to distribute the load to all the computers in my house, but really what I did was put all of this on my reconditioned budget OVH server and let it run for 5 days while I continued to live my life.

现在只需再执行104,782次，您就完成了。 在ffmpeg和ImageMagick上运行一整集，大约花了30分钟，这是我在2011年最先进的Macbook Pro上完成的。 这是故事的一部分，我想告诉您，我设法利用Amazon Elastic Transcoder或做了一个工作，将负载分配到家里的所有计算机上，但实际上我所做的一切在经过重新调整的预算OVH服务器上运行5天，同时继续生活。

Once it was done encoding all the GIFs, I just let Apache serve them statically with a long Expires header and put Cloudflare in front of the whole domain. Only time will tell if that will actually hold up with traffic demands.

完成对所有GIF的编码后，我就让Apache通过长的Expires标头静态地为它们提供服务，并将Cloudflare置于整个域的前面。 只有时间才能证明这是否真的能满足交通需求。

正在搜寻 (Searching)

I installed Elasticsearch and indexed the contents of the SRT files. Here’s where I encountered some non-technical snags: In Season 6, episodes 14 and 15 are an hour-long clip show called Highlights of a Hundred where Jerry Seinfeld shows you a bunch of old clips from previous episodes. And in the final episode of Seinfeld, Season 9, Episode 23, they do a ton of flashbacks to previous episodes. Both of these would routinely get returned in the search results, so I just excluded them from the search query. There’s probably a better way to just lower the quality of those episode scores, but the documentation for doing so in Elasticsearch is about as easy to read as the above mentioned ImageMagick documentation. And at the end of the day, no one wants to see clips from either of those two episodes. Sorry, Larry David, the last episode was terrible.

我安装了Elasticsearch并为SRT文件的内容建立了索引。 在这里，我遇到了一些非技术难题：在第6季中，第14和15集是一个长达一个小时的剪辑节目，称为“一百集精彩集锦”，杰里·塞恩菲尔德(Jerry Seinfeld)向您展示了前几集中的一堆旧剪辑。 在《塞恩菲尔德》的最后一集，第9季，第23集中，他们对前几集进行了大量的回溯。 这两种方法通常都会在搜索结果中返回，因此我只是从搜索查询中排除了它们。 降低这些情节得分的质量可能是更好的方法，但是Elasticsearch中的文档与上述ImageMagick文档一样容易阅读。 最终，没有人希望看到这两集中的任何一个的剪辑。 抱歉，拉里·戴维(Larry David)，最后一集很糟糕。

斜杠命令 (Slash Command)

The final step was just to wrap it all together with a Slack slash command, which is just a simple Express app that acts as a client to the Elasticsearch instance. There’s some OAuth to package the command as a Slack App and handle the Add to Slack button, but I don’t really need to check for authentication when the requests come through so I’m not saving the authorization tokens. The code for the server is available here: vandelayindustries-slack-server.

最后一步只是将所有内容与Slack斜杠命令包装在一起，这是一个简单的Express应用程序，充当Elasticsearch实例的客户端。 有一些OAuth可以将命令打包为Slack应用程序并处理“添加到Slack”按钮，但是当请求通过时，我真的不需要检查身份验证，因此我不保存授权令牌。 该服务器的代码在此处可用： vandelayindustries-slack-server 。

That’s it! I hope this technical writeup helps the next person that wants to extract GIFs from an entire television show for little to no actual reason.

而已！ 我希望这项技术文章能帮助下一个想要从整个电视节目中提取GIF的人，几乎不需要任何理由。

翻译自: https://www.freecodecamp.org/news/unfundable-slack-bots-9369a75fdd/

slack 聊天机器人