LangSmith Highlights: Getting Started

Key Takeaways

in this video we're going to show you how you can easily send your linkchain API traces to link Smith or SAS platform so you can use those traces to debug to test and evaluate them to also monitor any production application and in this video we'll specifically tailor it to linkchain code but it's also possible to send your traces if you're building another llm powered application outside of the linkchain framework and we'll show that in a separate video but if you are using linkchain it's really simple to get up and running you can do it in a matter of seconds all you have to do is set a few environment variables the first is link chain tracing to true you want to give your application a project name in this case I'm calling it planetary walkthrough demo if you don't give it a name it will just set to default and all of the runs for this application will get collected and aggregated together in this planetary walkthrough demo project which I'll show in a second you want to give a destination in this case we're sending all of our traces to our Cloud hosted platform at api. smith. link chain.com and then finally you want to give it your linkchain API key this is really the API key that you create in your link Smith account and it's tied to your tenant uh and I'll be sure to delete this API key after the demo to make sure that there's not any Miss use of of my token but that's all you have to do to get up and running and I'll walk through this code pretty quickly but just a reminder the way it works is Lang Smith is not hosting my application in this scenario the code is actually running on my computer but it's collecting all of my traces because when we install the link Smith SDK there's a callback function to that API destination that I've Set uh up top along with my authentication which is my API key and then all of my traces will get sent to L Smith which is hosted in the cloud we have options if you want to self-host link Smith as well but the code is actually going to run on my computer and the traces are getting stored in link Smith think about it as an aggregator of traces just like an observability tool so if I take a quick look at this code we're not going to go too deep into it but uh I'm running an agent uh using bling chain expression language it has access to a couple of tools a Search tool and an llm math tool calculator tool it's going to use open AI functions as the underlying LM uh we've given it a prompt that we've pulled in from our Hub in link Smith uh this this prompt is is a pretty simple one it just says that you are a smart AI uh assistant which I can show in a separate video um how to define prompts and how to pull it into your code so just take a mental note of that uh and then it's going going to use a scratch pad to keep track of its thoughts as well as a chat history uh and then we're going to run this agent against these questions these put that I've seated it and we're going to see what that looks like in lsmith so we'll run it now and then if I go to lsmith this is uh in my projects this is the planetary walkthrough demo and we can now see that there are a number of runs uh happening right now they're in action which is this uh purple running symbol and a few have already completed these are running in parallel so you don't have to serly have the agent answer these questions and we're going to show exactly what are the outputs of these traces in a in the next video