 Good afternoon and thanks for coming to my session and my talk today is how did we Improvement of the open source commuting management with the chat GBT or GBT for So before we start I want to introduce myself first I'm the I I work for second stage is a startup company and the way don't need to our open source project Call the world's magic to see if we call the native computing Foundation, I saw some books projects right now so I'm responsible them since they have was my community and I also organized Some more developer events in Asia. I'm one of the organized organizers of Russ China com It's a it's a big conference for a lot of developers in China This year we have 400 people 400 rest developers to attend this event and I also organize some small meetups In Taipei, Singapore and Beijing if you want to go to if you want to reach out some Russ developers in China You can also reach out to me and I'm also the other over since the Apple or wasn't landscape It's similar to them since I've learned the scale, but it's a fox on the web simply run a web similar ecosystem So So this image is not generated by a LGBT by the way, it's generated by my friends So before I started I want to do a little survey I will ask three questions. So first them I want to know how you ever use them LGBT or GPT4 in your work. If so, can you raise your hands? Let me know Well, almost 100% So the next question is how often did you use to LGBT? Okay, for me, I use LGBT Daily every day. I can't live with LGBT if it should be crushed. I will feel very bad So the last question is that some how you ever tried other open source like language models like number two Can you please also raise your hand? Okay, almost 50% thank you. Thank you so LGBT is Is really is really hot But how or how we apply to be to be into our open source community or open source community management This is a agenda for my talk. So I will give you a brief introduction of what's my so, you know What is what's my community and what what I what what I'm working for and I also will share my understanding of community management and Then I will share how did we use to the GPT or GP4 to enhance the open source community management and Then last if we have time I will do Tutorial to show you how to build a community assessment for them for your own source community By the way, all the devil tools I mentioned below I built with and was much and the rest and as they are also used in our Was my community you can also? Check out this page. It has all the devil tools. I will mention By the way, you don't need to master was much or us It's a it's a it's a sus it's come. It's a it's don't it's don't require you to master programming you can just set up some more Some parameters like what a GitHub organization and then get a repose to set as the automation So a brief introduction of what's much what's my is since the other son projects Also mentioned before why was my is since the other son project From this page you can see that it's them It's the since the annual survey of 2022. It's the service Can find it a key findings? Containers as a new normal and the web simply is a future. So web simply is a new runtime or new container in the cognitive area and In one shot. What's magic is a high performance alternative To these containers it's called is cross platform. It's less. It's less waste and it's a secure I don't I will not introduce what's magic magic here if we want to know to learn What's magic? deeply you can check out our website so The edu funnel I think most of us are familiar with so on the top of the funnel is Awareness is where you introduce people to your to your open source products And what kind of problem it can solve so in this stage Developers will learn your will know and learn your product. We want to get more and more More developers to know our product and in them in the middle of the funnel It's a interest and the desire this is where people identify a problem and as a This research solutions. So at this stage developers May will try your product to see if it can solve your problems in for the open source community is We can we can use GitHub issues as an indicator to see if there are any developers are interested in our open source product and The bottom of the funnel is the action in the trial in the traditional marketing action means that The user by by the products but in In our in the open source community in our culture is so that developers will become a contributor and Adopter so based on the open source community funnel I At least some more simple metrics that we will monitor that we take care of so for the first for the top of funnel awareness We will monitor our get have stars Docs page views and the report visitors and For the for the middle of the funnel interest and desire We will monitor How many people how many members in our discord server and how often they ask a question And then we also might monitor get have folks because we all know that get have folks is the first step to be a contributor and we also will monitor that's not the get have issues and the number of get have issues and there's a number of different to get have a issue creators and If they use our products and they might a problem They they will go to our get have a report and the report what they might out They will just the require request a new a new feature. So we will also monitor that is the software downloads So docker image post number on there's a there's indeed her desire them the developers are interested in using our products to solve their problem and then in some in the In the bottom of the funnel action, we will monitor the number of Ops and How many different people how many the number of different PR creators and we also will monitor How many people that use was my into their production your ment? But I'll I list some metrics, but I want to highlight some open source management is just a number is just About the numbers They are just one way to matter our walk But we can't so we can't use the numbers to account as the community's healthy, so As Linus has said in the morning People relationship is a heart. So the community experience is also matters We can use GP for GP for is not magic is a console other problems, but either can be can be a way to To enhance the community experience So how do we do so I will list three use cases the first one is get have issues management Get have issues is very important in the open source project we will use get have issues to track our progress and We also will use get have issues to main transparency of on the technical front Everyone can check as a get have issues. There is no secret in the open source community and them so the developers or the users will create an issue if they Might some problems as they were running the software report a bug or request a new feature and they will interact with the So maintainers so we are get have issues so get have a issue is Many get have issues get have issues management is also important to them to the to the Open source community We want to respond to every get have issues quickly and And now I want to help solve their their problem help solve the problems sooner Get have issues in worse the in worse the middle of the funnel It's it shows some the interest or the there is that developer want to use the our product to solve their problem. So the first one is I make a I make a slack boss because our company Communicates on the slack. So now this is like boss that can this this So one one a new issue is created We will the boss will call the culture GPT to analyze the sentiment of the issue if we charge It defines the issue is urgent. So the bottle will send the issue to the Maintenance channel directly. So maintenance will pay attention to the issue this on this This tool can help maintain our diverse to to Respond to the urgent to get have a issue effectively The next one is not involved in is not involved with the GPT for but it's an some kind of automation so The image on the Right lives and lives is it's also a slack but what is a slack slack about do it's It's an answer the issue reminder Because sometimes we want to respond to the get have a issue later But then we will forgot to forgot to so so we will have some issues that some don't have any response in several days It's by the four them the issue creator so So this boss can help can remind the maintenance that There is a issue that you didn't respond to please take care so The lives bottom of the the left bottom image This one so He created as the issue two days ago, but so we didn't respond to this issue In two days, so so I said about the same as them as soon as this issue to them for the channel And then the maintainer will know Oh, I have a how I have an issue that I didn't respond to so the the maintainer will respond to the get happy issue This is another automation Ecosystem is also important to some open source open source software. So I did a tool I did another boss automation tool to record all the get happy issues that Mention our open source project. It will save with the results to an air table Table table air table is a table tool so you can use it to to do some It's just like a cell so It will help me to understand that What can the ecosystem or was much which which open source projects are using was much and if them I will also look at The issue of the get happy issues if there are something wrong with was much I will also Cross-poss this issue to to our GitHub repo. So These two can can can help me monitor them with my ecosystem So this is the get happy issue management I also have now we also have one tool that can summarize them the get happy issue comments But I forego to listen here Some another use case is that is the new contributor about it to be honest When I see when I see a new PR is in our GitHub repo. I'm really very happy so A new a new pull request Indicates that a new contributor is joining the team. It's them. It's a it's a bottom of the funnel We want to we want them our community has more contributors and then be our maintainers so and the process for a new contributor to contribute to our our GitHub repo our open-source projects is the following so now first first The potential contributor will find the GitHub issue that is suitable for him and our heart and then he or she will write the code and then he created the PR after they he created upper he or see create create the PR our Maintenance will reveal the PR it may take all it may take several days and After everything is okay. The PR will be in March. So then we have a new contributor onboarding so From the first step we can filter the issues that some community can contribute to This is a discord bot We use the way our discord is our community member base better Our company uses like it is kind of 3d. I know so this is a discord bottle for in our In our in the world's mighty disclose over so when some all it's it's a GitHub issue Tracker. So if a issue is labeled with good first issue have one day GSOD our effects mentorship and OSPP The issue will be sent to them soon to the GitHub channel directly and as a channel name is Sinking something to contribute to so we fill it as that some fill it as the issues as that some Should both for them community or the newcomer to come to contribute to It's a it's a it's it's kind of help with them potential contributor to find as a find as the issue that they can contribute to then we are we are Yes, we can use to GB to record but But we we don't even move this process So we we will go through to the fast forward to the PR reveal phase This is an image from Lannaby from this image. We can see that some now PM PR reveal is It will take almost five days to reveal but We will we just to spend 19 hours to coding So PR PR reveal is takes the most time during the whole process. How can we? Accelerate this process So we made a PR reveal bot it's also based on to the GPT our GPT form the The GitHub boss will reveal the new PR Automatically it's kind of it will summarize the PR content It will summarize them as a key changes over this PR And we also point out to the now potential users the potential see can help with the maintainers guys the point over them Over this PR. So it's it's kind and So as a period boss will respond to the PM Soonly so and so it's kind it's kind respond to them contributor silly So this PR reveal boss can see what the maintainers time and respond to the PR creator in no time and it will enhance them the creators experience and For me is this PR real boss can also help me understand what does what does the PR do in plain English? I don't need to ask our maintainers So it's it's also can help me to learn what is what is the computers are doing and Then after the PI is merged and we want to we want to build the connection with the contributor We want to say we want to say they contributed to our repo One more time. So I will so I said Sets a set automation that's a send a second email to the contributor So in this this just a sample. So in the email, I will I will say thank you to the contributor and I will list I will I will ask him to join our discord server To to check out the other good first issues and I will also ask him to claim their developer swag so this is a whole process to to automate some the new contributor onboarding process with GPT or just automation another use case is GPT for the community assistant this is Well, that's open AI just announced a new product Last month is called assistant with the open I assess the API on Me people like me that's a man and not technical don't know how to program Can build their own assistant? I will show how to do a next so We can build a community assistant to interact with the community member and answer their questions This is I ask a same question to change the GPT and So what the boss is that I made for was much the question is very normal it's How to how to install was much on Ubuntu? It's a common question And also to be the answer is also it's also common But we don't have a was my APT report. So the answer is is wrong this boss is I built this boss with was my dogs. So I I use was my dogs to create some embeddings when One user as a question Open AI view and will search also similar responding embeddings and then send them send them and then answer the question with Embedding as a context. So this ball works very well. It's the command line is a it's a It's 100% correct with this with the command lines You can you can install was my g on your Ubuntu machine successfully. So I'm built building Gpt's GPT 4 or to the GPT is great, but some it's not an expert to form our open source project well applied to a Unique software to the GPT Open Provide wrong answers. So now so we build a rug based Gpt can help solve some the problem. So We still have time so I will show you how to build a rug based Gpt assistant with open AI and And the and the first network So the first step is that some go to the open AI platform and the sorry and then go to Assistant here and then you'll click on create To create a new assistant The instruction here is a promise for the assistant so you can custom your own and Then we choose to the model and You need to choose a GPT for 1106 preview our GPT 3.5 Turbo 1106 I will use GPT for and Then I will show the enable the retrieval feature here and Then we can add some some some some files those files will be a create a Open AI assistant API will use these files to create some embeddings for your own for your own knowledge base Sorry, I changed it on the computers Just give me a minute Sorry, this is not my computer because my computer can't connect The money so I upload I uploaded a Magnum file here, and then I click save it's done We will have a Assistant idea here and this is very important because we will use the idea to connect discord or telegram later, so Before we before we Build Connect to them this assessment to telegram of discord. We can also test the assistant into the Into the open AI playground. We can ask a question here The file I just uploaded is about how to fix that With was much on Docker, so I will ask this question It will take several minutes or seconds to respond to this question We can we can go to the next phase after we have a center ID ID we can build a build a Telegram of discord both base based on this This assistant because high discord both is the discord setup is a little complex So I will use the telegram here We can just Click click on deploy And then we will need to to fill in some parameters The first one is the assist idea which we can get from the open AI and then we need Open IK and then we will need a telegram token. So The system just answer my question That's very very correct and the Docker desktop version should be above 4.15 because this is Collaboration with was much and the Docker that happened on KubeCon and a Detroit last year. So so you can see that some The open AI assistant can answer unique questions related to your your open source software and then Then we can build the telegram bot So we need this telegram ID. Oh, sorry assistant ID just Coffee this one Fill in here, and we also need an open IK. I will create a new one Then we need a telegram token. I will use my phone to to set up the token Yeah To create a telegram token you can find us about father. It's very simple You just chat with both father and then we'll get a telegram token. It's much simpler than just called Okay, then we click on create and build It's almost done So the force network will build and deploy them the function Which so and we don't need to anything just waiting for some get have the function is is built but if This computer doesn't have a telegram button you can find them the boss where this link is still building so I Will continue because we have one we have we also have several minutes. So if we won't build any Any automation on whether they are pulling the Pulled by to the BT or not. So you can you can scan this card code to check them the automation tools I mentioned the in in my slides and then This tool is pulled by always magic and if you have any questions in building your own automations You can contact me and I put my email here Okay, so one problem that I think it's a beautiful is that it's too expensive so We can there are a lot of open source like language models and We can also use this is So we also use open source alarm to to run or build our own agent just like I did before so if we are interested in High performance and the portable AI inference with the rust and the wasm you can go to my colleague now Michael yes talk tomorrow. So That's all my all my presentation Yeah, the the telegram boss based on the OpenL citizens and as the How to get started with how to get so wasm was magic with the Docker is Is build successfully and if you want to try you can You can find the support with where this link because this computer doesn't have Doesn't have Doesn't install telegram icon to show you in live. So you can you can try it by yourself Okay, that's all. Thank you. I think we can Answer some questions if you you have Yeah, I'm quite curious the I saw you upload a MD file Creating the Docker Yeah, we mind showing us the content Yeah, just now you upload for the file interpreter, right? Yes Yes, you can just upload your yeah, maybe could you show us the content of the file that you uploaded? The Docker of the MD file. Yes. Yeah, sure. Thank you Sorry, it's not my computer. I Will show you in the was my dogs This is a file. I just upload to OpenL citizens. I can send you the link after the meeting if we don't have any other questions That's all. Thank you