hello everyone thank you so much for joining the office hours plus live stream so we combined both of those uh together this week since uh there are a lot of announcements that happened this week uh just wanted to update the community on what all came out um also wanted to give a chance to ask questions or provide any feedback you had so far uh so the the big announcement is that langlow 1.0 is out um there are tons of features so uh we're going to cover a few today um and then we'll also talk about the cloud version and how you can access that uh and also a few additional things so um yeah a lot of things that happen and uh uh we thought who's going to be better than to have one of the co-founders of Lang flow with us so Rodrigo is with us today hello Rodrigo hello hi everyone lot of people uh we're trying to do this more informal and not very pretentious uh office hours over here are alive if you will uh and try to cover a little bit of llow 1.0 the features uh the announcements that came this week and uh maybe take some of the questions that you all are going to have so thanks everyone to to join us and it's it's really exciting awesome thank you so much so yeah at a high level Rodrigo you know I think uh just to cover um there was the announcement about 1.0 and then you know just been know tons of exciting features in there uh about Cloud announcement you know maybe I'll I'll quickly walk through how one could access U langlow um it's just the the the really nice thing I like is one could just sign up and get started in just maybe 20 seconds and start experimenting with Lang FL right so it's a hosted version uh one could start experimenting right away um and yeah other than that is also the integration with lsmith so that's awesome so a lot to cover how about Rodrigo we get started uh just maybe a few quick things about what 1.0 is and uh maybe just playing around with it how that looks like absolutely yeah uh very very very exciting to us uh this week came with three three huge announcements uh langlow 1.0 it's been a while that we've been talking about getting that that together we've been uh we gave you like the preview so that everyone can see uh what that would start uh to become uh but then finally we made it uh in a very very short time uh short period of time and the entire team uh gave their their uh strengthen and put together a huge amount of effort to make this happen as quick as we could right so we hope you're going to enjoy L FL one all and with L FL one all we come new feedback we come new uh things that we want to hear from the community and we're super as always open to uh get your uh your opinion about it right about the features about uh how it's looking I hope to cover a little bit of this today uh and and then we can move to maybe some um demo as well show a little bit of it uh second point was the Lang Smith integration right so now you can everything that you're doing in Lang Lang flow uh can be with a single string being passed can be now monitored by Langs Smith and that is awesome right and then finally and most importantly we have a llow data Stacks Lang flow which is the the cloud version of Lang flow itself right so after long year of development of a of a project that is only open source and locco we finally have a hosted version of it uh by data STX and it's awesome and it works you just enter it over there try it play with it it's a preview version it's completely free okay so you can you can play with it you can play with Astra DB for for free as well right so try it out uh give it a go and let's see uh let us know what you think are you're on mute is but yeah sorry uh so just wanted to real quick mention it's a lot to cover so we'll Dive Right In in a minute just wanted to talk about some quick uh things about the platform so if you've uh used um crowdcast before great if not then there are a couple of things you can send any message in the chat window on the right side uh you can also um ask questions in the Q&A section we'll monitor both of those and also if there are any polls if we send out we can you know we can have an interactive session so yeah it's going to be a bit casual today of course it's Friday and then also it's our office hours so grab your tea coffee and let's get started Rodrigo it's Brazilian coffee by the way oh wow nice so so I'm going to share my screen Miss and try to go through some of the uh features that we've been talking about I'm gonna just share entire entire screen and if you come to docs. LF flow. org you're going to see in the first part of the documentation we wrote this quick review of what we wanted to address in langlow 1.0 right so of course I'm not going to read hope everyone is gonna is Gonna Take a Look by themselves but I'll try to go through like the main parts of this and why it matters right and this is a long time of it like uh conversations with the the internal team together with feedback from the community and trying to address the main problems the first one I think being that langlow was not what we can call touring machine right it it wasn't touring complete meaning we couldn't Loop we couldn't split we could not do if and else or other logic operations right and that that was one of the biggest additions in my opinion to Leng flow 1.0 so same components multiple outputs right now link flow accepts uh allows you to create components that can split outputs and Route those outputs to different directions right so you can imagine just with that combined with composition L length chain components and agent and tools Etc with this kind of if else logic can be very very powerful right so I think that would be one of the most important features in LF flow 1.0 you Branch outputs right you create uh orchestrators of Agents Etc right so imagine one big uh deiser component that is going to route through multiple outputs or mul multiple models or or agents Etc uh many decisions right this is a core piece in my opinion of a language model real pipelines because in in in uh real problems you need control you don't want the agent to do everything sometimes you you want an agent to do something send a message over to another agent or maybe if something happens it goes that way if something else happens it's going to to go at this this other route right so multiple outputs I think that that would be the the number one uh feature for Leng flow 1.0 and that comes together with uh an entirely new uh component code okay so if I I'm going to actually come to my Leng flow demo running on my screen now just put a component any component over here in the screen a chat input for example now we have the code of the component over here as always right in Python uh but you can set and Define outputs if you can can you see my screen misba yes so now you basically can set and Define your inputs and outputs right and your outputs can be of n size you decide right so I'm going to assign that and output runs a specific function and this function or method is going to be run uh once the once the the component once the component is built okay there is no more for those who use U custom components there is no more of that now it's it's the custom component there is no more of the build function sorry so now it's multiple outputs decide the number of outputs you want the number of inputs you want and you decide the functions that are going to run and how they're going to run right so this is complete freedom to create anything python within a a visual framework right and I hope that the this is going to make things more uh simple easy and and intuitive for everyone uh okay and then um Miss but the second point that we're that we're very happy with Lang flow 1.0 is the fact that all of that also leads to a different Paradigm for how data and information is passed right you can imagine flow style pipelines and composition style pipelines flow meaning uh data is going to be bested between across these components right so one component process data sends the data to the next component composition meaning uh one component process data and sends the state to the other component or an object to the other component right maybe being the class of of the of the first component itself so that the next component can execute it later right so you can imagine data being processed so some data over here going to be processed by the second component second component sends the the information to this third component and third component might be might work with composition right so composition meaning all of these little guys over here are components and they are tools okay it runs maybe for a recurrent uh some some recurrent steps and then it sends information over again and it continues on the flow style right so you can have both worlds now in Lang flow composition and flow does that make sense yeah and and and perhaps on that U Rodrigo there is a question saying that you know the um one would like to set up an API and then from there go on to send a uh maybe instead of string message some send some file or so as an input and process the file through the notes and get something back is that something possible yeah absolutely so uh what I what I'd say here is L Leng flow basically have different types of uh inputs and outputs now and they this these types can be very easily recognized by the colors of the handles okay these little circles over here are what we call handles and uh right now when you when you pass a message right let me send anything over here and press play I can inspect the message that I passed and on this message I'm going to see there is text there is a sender sender name there's a session ID and there are files right so meaning this component didn't pass any file of course because it's not a a file sender or anything but the the message object enables files to be passed through it meaning Playground now can on the playground I can send images or even if I want uh the file component itself right can send something I can have it over here uh it's going to I'm G to grab anything on my uh my stuff don't know if this is going to read but in any case File paath is here the content of the file is there I don't know if that answers the question meaning uh the inputs and output types are more advanced and more complex right now meaning you can carry not only text but files images uh and whatever suits you and it's python uh it's accepted by any python function I think that's the that's the overall yeah perfect um so yeah thanks Mike in case if you have a follow up feel free to drop that question comment so should I move on to next feure M yeah let's keep going oh all right uh so basically going back to the new features that we have um one of the most exciting of them is also memories right so I think we covered that on our last office hours but I but I wanted to show this again over here and basically what you can have with Lang flow right now is let's let's suppose I'm going to create something very stupid here just an input and an output right so no model there is no processing anything that I input is going to be passed to the output right so high high high right nothing nothing happened at all so the model is just being echoing whatever I send on cool uh but now everything that I sent is stored in me in memory okay so we have remember this previous message I sent over there it's on my one of my memories right hi by the user hi by the machine hello by the user hello by the machine and I have a complete history of every interaction human machine that happened so what you can do is you can create chatbots now that have that can use this previous memories uh to do something useful right so suppose I wna uh take the memory and pass it over let let me play this memory just to see what comes out user hi hello hello da I can pass this over to a prompt for example and say uh history right given the history answer my question and then I can have a variable called question and then answer for the model to answer it and there you go so I can pass my chat chat memory to one of the prompt variables and then take the input being that to the prompt as well and now I can maybe throw over here a language model and then and then you keep your your usual pipeline right but basically when I run this prompt this prompt message that will go to the openai will carry the history that I wanted it to to carry right so memory is something that is becoming much more powerful and friendly inside Lang flow you have your session I can even come over here and delete some of these right I say I don't I don't want uh modotto to store these messages anymore right or or I can even say I didn't say hi I said I hate you right oops oh my god oh okay error for some reason this feature is not oh I know why it's because I'm using the that version not the main one use the main one but there you go and then and then that's it m should we cover more features we wanna uh any questions do yeah actually U there is a a question as well also on the previous one maybe we can come back to it um if you could showcase if you know if there's a file uh or something that accepts file that knowde how the API looks like right so how it allows want to send a file over the uh API column maybe we can cover that in a little bit uh the other question was is it possible to iterate data rows and run flow for each entry like list of questions so from a CSV tables um and have answers in form of data that's a very good question yeah so this is this is hard it's not easy but it's already possible Right with link flow so if you want to it at uh with data over here there are two two main ways I I'd say uh give it a chance right number one would be uh you're going to create a component with a for Loop inside of it that would be number one so let's say I have a a file loader right this gu is for for instance going to load um CSV file let's see if I have anything maybe I do maybe not all right data. CSV okay this guy is going to load it and uh the data that comes out of this should be passed over to a component that will Loop uh through the rows right and now we don't have this component pre-butt but we can easily create one so custom component let's say I want to use I don't know bundas for this right I can literally come here and import bundas as PD uh my function is basically going to take the data and it will create a for Loop so uh from uh pd. read CSV da d da and you're gonna use your code over here uh to Define to Define uh whatever whatever is it you want and now you're going to create your inputs over here that are going to be the file input and so on and then you can Loop inside a component uh but that's not very fun right okay you can do it but you can do it in a in a code way uh there is a there is another way right it's experimental right so if you come down to experimental there's something called prototypes and these are the super super uh new and uh cool but also very weird maybe uh unstable features right so you try it out and the idea here would be having using the notify and listen right uh this is something that uh Gabriel came together with and we were discussing what the what it means for L flow notifying listen are are components that allow for Loop to happen within the platform because what the notify is doing is basically taking any data and it's sending to another component Wi-Fi right Wireless mode okay so so let's say I have some data over here I'm gonna pass a key a key name let's say my key and this key is going to be received by this component so whatever this guy spits out is going to come inside of this listen component I can play it and receive the output on the listen which means if this one is in the end of a loop in the end of a a a sequence of components you can imagine how you can create all sorts of creative loops with this so notify the last data the last piece of data in your flow and listen in the beginning for another flow or with some conditionals over here using these two components you can basically iterate through lists through uh tables or whatever right but answering the question not natively you can try you can do it if you want to explore a little bit further okay very good thank you rrio um yeah let's let's carry with the the exciting features okay cool and then U more features like overall giving overall control in terms of how you can what you can do with your components and and and the the pipeline in general one of them is the is the freeze uh feature which is much much prettier in the dark mode so freeze is basically doing let's stop the component from running again right let's say I loaded something from a vector star and I want to keep that component there but I don't want it to load again or I loaded some uh uh data from a website I parse the website I scraped something I don't want to do that again and again and again so first time I run it then I freeze it and just stick with the last uh result right uh that that's very useful for a lot of cases for avoid spending extra tokens and multiple uh different situations that you're going to find in your way where you you're going to try to use the uh the freeze feature and then we have the the output preview I just showed this from now on every component that you run basically have little eye on the side so that you can see what comes up right super useful for inspection uh my suggestion don't do anything without understanding what comes out of your components understanding types in Lang flow especially now that we have these very clearly defined types like the message type uh the data type okay and uh text and models that me show another example here we have open AI model spitting out text but it also spits out a language model meaning you can you can use the output of this model or you can use the model that was built itself maybe in an agent right or maybe in another component that uses a language model so you just have to build it once so this is super powerful and uh useful uh especially when you start to get more familiar and comfortable with using these uh different outputs okay and then we have uh the custom endpoint name if you come to your flow over here and you w to uh you want to set up an endpoint name now for your flow you just give it a name right you don't need anymore to use the the ID of this this is useful as well because you get a lot of control over how one or more apis you're creating are going to uh receive send information right so set up your your endpoint name over here and and it's going to be accessed through the API once you once you run it right so the flow ID is going to be the name over there instead um actually Rodrigo while you're on the API page um could you show the uh the tweaks available especially for the file um that that was one of the questions earlier right how does that look yeah so basically let let me add a file compon oh I have one okay let's just have the file component in my workpace and uh let me add some file here let's say this data CSV all right uh API is going to have I'm going to have some tweaks and I have a path right so so basically I can when I'm going to export either my python code let's say I'm going to use the python code right we have the tweaks uh the tweaks variable over here and I can set each one of these variables inside a dictionary right so in in the example here if I let's say let's say I'm going to open up the tweaks look at this so in the Run curl this is another a better example tweaks is here and the file component can be adjusted inside of this dictionary by opening up the tweaks the tweaks inside of that I can set up the path that comes inside of that field of the file component all right and that's how you set up uh every every component's uh fields and variables it's through the tweaks right and to see every one of them a bit more clearly we have this front uh this uh little table over here where you can inspect each one and change them in place here right perfect yeah I think that answers hopefully that answers the question Mike um about the API usage awesome yeah let's uh keep going through the list Rodrigo cool all right so getting back and please forgive me if I'm just going through this very fast uh keep sending questions and I'll just try to answer them as best as possible so next one is going to be logs and monitoring right uh from now on langlow has uh two new features one is going to be logs themselves right so every component sends information to the other one what is being sent what was received by home and um how did this information go is there an error status uh source and Target all right so you can monitor everything that's going on from now on that's one and second you have uh complete integration with uh LS me lmth sorry so come over here check uh your your linkchain API key set it up set up your uh L chain project and once you work with L flow just run some components you're going to be able to see your output uh in Lang Smith uh natively right so any any component that's that you're running inside L flow has already an inbu uh Trace being run behind the scenes right so just just explore it uh try it out and tell us what you feel about it yeah that's actually quite nice you know visualizing not just in langlow but also one could do that in in L langmi with the integration yeah exactly uh I think we have a couple more and number one yeah and then the other one and this is super cool is going to be multimodel right so if I open a new chat here and try it out with GPT 40 or another multimodel model so let me run it over here oh my god oh okay my format my bad let's try the dog again and there you go right so uh you can send dogs now and cats be happy uh all right uh this what this means is it's not just image right what it what it means is much more it's the entire Lang flow structure uh behind the scenes worked with text only and having images now means that in a very short period period of time you're going to see other models coming together very easily right it means that L flow will soon accept audio and video and any other type of data really because the components now natively accept this kind of transition or our our um data structures accept this right and it was a big refactoring a huge change that that is now completely passed through so imag is just the first one I'd say uh then we have and then we have folders and the playground right if if you don't if you haven't explored yet uh the playground is exactly what you see over here when you have uh when you just have the chat input or output you're going to see exactly this screen here but if you have like text inputs and text outputs for example use see these inputs and outputs on the sidebar as well and these can be used to monitor too right imagine you want to uh depending on the situation you want to send some fixed text in to this box over here right and you don't want it to come to the chat for some reason it's very very useful for inspection okay so text input text output memories uh you can create multiple memories to your model so uh your chat your chat input and output they they take uh as as one of the fields the session ID right and you can by default the session ID is going to be the same as the flow ID so if you pass nothing it's just going to consider uh it's a default session ID but if you want to create different session ID and just come over here and put anything and you run it you should see like another memory right and you and you can have unlimited amount of number number of memories over here so create multiple memories and imagine working with agents that send information to these different memories right imagine having agents that store different information in different uh places and then also retrieve different informations different pieces of information from different um sources right this is very powerful and hopefully it can be used uh well with some creativity great um actually on that Rodrigo um the the pre previous one right about files so what type of file uh could be sent right now um wondering if there's some documentation right on on what all could be sent yes okay uh so let me open the file component and we have inside of this we have a uh just a second text file types if I'm going to I'm going to look inside the code itself to answer this question right but this is all this is all going to the documentation very shortly so text if I come to file and I look at this so these are the the file types that are currently accepted by that component but and I'm I'm gonna ping this on the screen on the on our chat here one sec perfect there you go so these are the the types that are currently accepted by the bi component and that that doesn't mean this is what link flow accepts this means that this component right as we are using it right now accept these files now you want to open you want to create your own component that will accept uh something else it's a couple lines of code it's not a big deal right so you say well let me plug my favorite Library here or let me make a connection with unstructured IO or grow bid for PDF or whatever you want right so ju just write your component if you have little bit of python background or even if you don't I have a great GPT uh instance creating components for us I'll try to make it public so it's super easy to create your own component and make like uh get out of out of the the basics with Leng flow yeah actually that's that was another question about how to build these custom components right so uh um chat PT custom GPT probably was one of the source um is there any sort of tool or uh a rack component which helps kind of write this uh custom component code yeah we have a coule we have a couple of them right uh but since we refactored uh the way that components are are written we're trying to make one that is that is going to be new and up to date right because the the entire the the entire structure of how inputs and outputs uh changed right now with llow 1.0 so let's I can we can share whatever we have over here my suggestion is to is to take a couple of examples if you take the ones that are everything that is in L flow on the sidebar is open to see the code right so if you take all of these any any one of these components they were written written using uh the new code structure that means you can just use them as examples feed them to your favorite chat bot and ask it to to create a new one like that but that does something else okay I think you're going to get like 80% of the way there and feel free to use the Discord Channel and we'll try to answer and try to create the components that you want to you want to see in the sidebar right the goal now is to be increasing the number of components that are going to be natively useful especially like helpers for example combine text future data uh merge data parse data split text Etc and then increase the number of integration bundles like the notion integration bundle that we put together on llow 1.0 preval or an unstructured IO bundle or uh take a Discord bundle Etc bundle would be a pack of components right and that's our goal right now to put together all of these component packs out so that people can either download from the store or take from our sidebar over over here and start using them great and maybe one last thing about you know file types or so so you you showed us how you know this um chat can now acccept multimodal uh inputs right so you should image in there uh what all is possible right now and um yeah if there are new things you know audio or video so how how does that look like if one were to to send that info uh I I don't I don't know if I get the question right MBA so you want to know you want to know how yeah how and and what is accepted in this in this chat window so you showed one image right and it was some specific format right so what what all formats are accepted in here oh yeah so the formats for for images are these ones let me paste them as well right nice but so far there is nothing apart from image right so multimodel is multimodel right now is image only okay if you if you want to use on the chat some different type of uh of structure that is not image or text that's to come right the rest can be passed through uh the file component can be passed and converted into text Etc but natively on the chat it's it's just image for now yeah and the good thing is you know one could always take custom component Python block right and then kind of write whatever the logic is for video or any other format and then send that data over so there's always that flexibility using custom components great yeah Rodrigo back to you on on the list of features okay I think we're we're done actually I don't know if we have many many more to say and I I I think that people will be tired of hearing me uh speaking about everything so I think I what I'll do is I'll just try to try to conclude with what we want from this from this 1.0 right and I think the idea is overall to add control to Lang flow right we want language models to be able to create real applications and we want these applications to be completely monitored uh easy and inter inter active interactively built okay and these models they sort of need an interface for the human to validate and experiment and prototype really fast uh the way that that we do that on code works but it's still very slow right and we're trying to make length flow and interface to solve that and become hopefully the number one piece of soft Ware for people who who are enthusiasts about Ai and when I start building something very powerful very quickly right and uh and uh we've been Super H happy to see how the community is growing in in the last couple of of months and langlow uh rippo is basically getting now almost 20,000 stars and this is this is unbelievable right we are super happy with that uh we see a lot of feedback on this card we are taking that feedback sometimes we we're we're hiding uh from from the community over there or or we're we're missing from the community over there just because uh we we were super busy on this one at all but we're back there we are trying to get your Fe feedback and and literally just ping us uh if if we don't answer what what you need uh feel free to to poke us again and and we'll be there trying to trying to add and prioritize the features that the community is asking our goal here is to solve for the main problems that you all see every day on the real uh real world case applications not only um the playground or or the interface but make it in the end become an entire almost an idea right how can I create how can I learn create prototype and then deploy everything that I've been playing with right so I think that's that's the the the final uh go here missb awesome well actually that was uh quite a lot so folks uh you know there's always the video available it's recorded we'll have that posted on YouTube um and there are a couple of things that we'll cover actually Rodrigo before that maybe a few quick questions one question was with the the latest release how does it look like uh to build uh multi-agent setup um or flow does it require multiple flows or one could do it in within one flow um how how do you go about that yeah so you can already create multiagents uh in a single flow right uh the we for for safety and because after one we want everything to be more stable right so what we did is we threw agents still on the experimental tabs take a look experiment especially with the to two calling agent right check that one out and then there is something called the flow as a tool and this is where I think the meat is going to come right flow as a tool is basically create your flow create a pipeline and then you're going to say okay I want to see this pipeline as a tool to another agent and the power of this is basically Unstoppable by it's it's whatever you want literally now and we already have that over over there uh experimental right check it out we don't want to put anything above experimental if we are not certain anymore uh that it's enough uh qade or stable so that uh so that everyone so that everyone uh has the the required um maintainability right so for for for their applications and and real world cases to work uh without without issue isues I think another thing that we're we're probably going to do is we we very likely have a pre-release uh version as default as well so that experimenters can also plug that that on and play with the most recent features always right so we we'll likely have these two uh different ways for you to to play with um so yeah awesome um couple more questions actually Rodrigo that came in I think we can just answer those before we go on to data Stacks uh L Flow side so um one question is is it more simple to create uh well actually I think we we kind of went through that the the multi-agent side um are there any sample projects demonstrating the new features in in general um maybe I I could probably chime in over there um so of course this the the video series that we have right now and office hours can help you kind of go through the the new features but also we're working on getting some short tutorial videos out so you'll probably see a lot of those where you know different features and how to use uh those are going to be available to the community uh not just that uh there's also the the store section right so you'll see a few things over there how these new features are are available um yeah other than that any other resource Rodrigo that you would recommend uh yeah so I think we have a couple of we have a couple of previous webinars we have a couple of YouTube videos uh get onto our medium blog there are a couple articles with use cases although in overall langlow 1.0 just came out right so the amount of content for 1.0 is still low in relation to the previous amount of cont content so give it a give it a couple weeks right or or next week the other week you're going to see so many flows coming out because we start of nailed down into the types of inputs and and and outputs that we wanted uh to fix on right so from this point it's going to be releasing content and creating useful stuff around what's around this platform right what I like to try to convey is that L flow is not the store of components it's the it's the factory right it's the manufacturing plant so we want Lang flow to be the component Builder right and I think we reached that with one point 1.0 with a good amount of success and from now on it's it's it's the the work of putting components out right very fast so putting components out and flows out and more examples and let's make that uh list of examples to get started uh become instead of six examples 100 right so that's that's where we're aiming right now perfect and also it's it's just amazing to see how Community comes up with so many different just creative flows and and ways to actually use langlow right we usually tend to give some generic you know base level how to get started but it's always nice to see how communities involv and and yeah come up with so many solutions so U yeah other question Rodrigo is um interested then how one can use um the API token so kind of programmatically log in a user with the API token um it seems like uh let's say this is the use case where there's a front end and uh want to log in Auto log in someone using that API token how can that be done oh that's a good question that's a good question for Gabriel I'm gonna skip that one I never used it let's yeah and I came across documentation um I actually I never use myself um but that's actually a very good question so how about we hold this one to perhaps next office hours we'll we'll try to answer that uh with Gabriel um another question about the um fir craw integration um was yeah any any comments over there on fire crawl integration it's it's released so just go on your sidebar and check it out fir craw API and fire craw scrape API on utilities you should see so Rodrigo if you don't mind could you real quick maybe show on the screen and we'll um yeah and I think I'm gonna use the I'm gonna use my sharing again to show people something too uh look at this so this is the firecrow API right Rodrigo I think we're seeing the uh not the the L screen my bad let me try again sorry about that re sharing good screen share your screen I was sharing just a window sorry about that and yeah okay can you see it now yes all right again everything that's new on experimental right once we keway it enough we're GNA send it up that's how it's probably going to work uh and then I have uh something this but that maybe we want to cover especially to say to illustrate like some of as asked some of the new features right this is uh this is a completely like different component than than you've ever seen in Lang flow I don't know if you can notice here but this is a group component okay uh I called it an AI router and it has on the on the description right now for those who don't know this it accepts markdown right so you can create a little bit of a documentation over there to explain to people what your what Your flow or component uh is doing right and this this guy over here is basically uh try to try to follow up with me here it's it's an a router right you're taking it Taking A a condition and given the role of this condition over here plus a couple of examples for the true case and the false case it's going to send the message input either to the true route or to the false route does that make sense MH it's basically branching right here I'm saying check if the user sounds happy right two examples big news L10 is out woohoo that movie was awesome love you babe these are some true examples for happy right false examples oh my cat died no one loves me not cool right so basically what it's what it's doing is if I send a like it's a a model classifying the input into one of these cases and it's taking the input and sending in different directions all right so if I take on the playground here let me remove my previous messages and I'm just going to say let's see if the open API key is set up maybe not probably not yeah let me set up my API key inside of this component there is the advanced settings and you can basically set it up this is an old a very old component so well let me try that again and I'm going to do uh my cat died you're sad right and notice there are two memory sessions being created just because I used the different one last time but then uh yeah we are doing an awesome live with msba now and I'm happy yeah but uh why it's cool it's not because it's it's deciding if the user is sad or happy the cool thing here is that it's deciding where it's going to send to right meaning depending of a user input it can be it it can route the output to a completely different flow right and then you can say hey depending on the output you're going to use this database for rag but depending on the other output it's going to be a different database for rag right and then you can start manipulating in conditions together with the power of AI and llm um contextual building plus agents plus rag plus everything but if that makes sense awesome uh that's actually really nice especially the markdown uh I love it um I mean of course also the router it's going to see probably a lot of views from the community um there was a question from Francisco about sharing the GPT for custom compon compents so perhaps maybe you know once we update all of that with the the latest release and the documentation we can certainly share with the community um other than that maybe real quick we can talk about the uh uh data stack SL flow and uh then take a few questions so I'm G to start sharing my screen okay so this is really exciting um within 20 seconds or so you can actually go from the you know just no access to completely in langlow and experimenting right so if you want to follow along just search data Stacks Astro or data stacks langlow and you'll see something as you know result from from data Stacks um click on that link it will take you to a page which looks like this where you can sign up and it's free so you can sign up using GitHub Hub or Google or make an account uh once you do that probably 20 seconds or so once you give access you land into a page like this so real quick way of actually getting access to Lang flow and this is again within your account so anything yet that you create over here new project flows uh it's going to be within your account so this is amazing now in here you can of course start a new project um take from the templates you can basically do perhaps most of the things that you can do in your self-hosted environments um the there are a couple of things that you might notice that are a little different here um especially with the version as well as with the API access that might be a bit different but uh this is probably the the fastest easiest way to get started um I you know I started with the template and then you know this is very basic we have covered this example before um yeah again want to really emphasize if you are helping someone learn or you want to forward uh this info about langlow how someone can build cool llm apps uh highly recommend it to to get started with this um you can create you know tons of flows tons of whatever you want to build in here and uh data Stacks provided this for for the community um to for free so yeah do check it out I'll probably Maybe cover more details in one of the future ones but just wanted to kind of you know quickly go give an overview so yeah anything that you want to comment over here R yes so M we just received a question uh saying can we share flows with uh specific users or groups that's that's random jails jails I don't know how to speak that uh so yes you can you can share both flows and components in L flow uh there are multiple ways to do that one is if you click on on the the component options you are able to see uh there's a download button you can just literally download that every flow can be represented as a Json file so you send it over to to other people another way to share them is is to share them through the Lang flow store okay L flow store is a place where we keep a lot of components and flows for people to play with um it's it's Super Beta it's it's just for people to try not nothing really serious over there just yet but it's already a place where you can throw your components and and flows over there and every everything that you're building on uh just and I'm gonna just share again if that's okay just to show this because I think this is probably one of the most powerful features in Lang flow maybe powerful hidden features features in Lang flow uh is that if you if you're if you're working with uh flows and like really really complex stuff things start to get very messy and this AI router for example I just showed is actually a combination of a lot of components together so let me open it up and show you the real the real thing behind the scenes and I'm going to ungroup it okay so I ungrouped it and this is what is actually going on right there are a couple lists uh sending information to parse data parsing the data sending to prompt prompt sending to an open AI model uh there's a conditional router over here so a lot of stuff going on right and what I'm doing is I took all of these I grouped them and once you group them uh you can decide what's what stays in the in the main uh screen over here in the main component or what goes to Advanced and you can set up a name a description and so on once you do that you can save your component so let's say we're going to call this component mbus uh component right and and now and well let's add a Emoji to that okay so it's going to be the fire one cool now let me start this oh and by the way just use markdown over here right bullets and there you go and now let me store it save it right in my saved component so I'm going to save that guy now anytime I want to reuse that entire thing I maybe spent two hours working on I just drop it from my sidebar and use that and the the best thing of all is that you can group groups as well right so if this guy connects to that guy uh supposedly you can create a group from both and then you can go forever oh but this is going to become very very noisy you just come to Advanced and start hiding some of those fields right that you don't don't need the user to see so forget everything let's make it very simple and maybe uh it's going to become useful for next time use right uh all right lot of outputs over here let me hide some of them don't need these don't need these maybe these two are the only ones I need give it a name now I'm I have a group of groups and I'm going to start this again right save so I have it on my side bar if I want to share it with my friends uh I can do this in multiple ways one of them is going to be download here I can also share my entire flow from over here right in this case I can't do this because working on the da version but uh on the on the 1.0 version you're going to be using you're you're going to have this button available uh and and maybe there are some restrictions to your flow for you to share it but then but then uh that's it so create group assemble again group again share and and make uh the community reuse whatever you created so that we can we can build on top of these modular building blocks if that makes sense perfect yeah and um there was a question about grouping so that's answered um are there things that um don't group so I think we covered that so just the chat input and output so those are the ones that cannot group but everything else in between uh groups together right um I think I recovered most of it one last thing is if there are multiple lines from one connector that connects uh to the next block what's the usual execution order for something like that I don't I don't know if I if I understood the question M my bad so so let's say if there are multiple connectors going out from one block to the other maybe two blocks right uh is there some sort of execution order that happens like this one happens first or the data sent to the they're going to run in parallel right but always from the left to the right so so you're see you're going to see the first component sending information to the other to the SEC second two components and then these two may be branched into more components and they are going to always be from the left to the right right and there are multiple modes of execution one component can send information to the front or if you call the the one in the front it's going to pull information from the back so it depends from where the information was asked to come perfect awesome I think yeah we covered a lot of things there are also more questions coming in we always uh recommend and suggest for the community to post if you have questions in the Discord uh server uh we will do our best to answer as many as we can in the office hours over here uh it's uh it's amazing to see you know folks joining in and the uh interactive sessions over here this is awesome uh we'll do a multiple form format so maybe Audio Only and then something like this in the upcoming weeks so yeah it's it's great to have everyone thanks a lot Rodrigo for going in details with the with the features and uh yeah we'll we'll probably be covering some of those more so we'd love to have you more on on these thank you MB and thank you everyone who joined uh we're really happy to see you here byebye all right y all take care bye