Pablo Elgueta

ChatGPT for lawyers! Using @langchain and @pinecone I've created a legal research assistant that can search through legal documents in seconds. Currently it supports basic legal codes and can search government websites for new regulation. @hwchase17 shorturl.at/euvAV

@moondrencht Wilson

Today, we're proud to launch Wilson to the world – your very own legal superagent. Think Cursor for legal contracts. To celebrate our launch, we're giving away a month of our Pro plan for free -- comment "Wilson" and we will send you a promo code. No waitlist -- try Wilson today!

@matiii Would love to join!

Trying to do the same for London! If you’re around and want to join the invite list for the next event with other builders and founders, drop me a note or comment if this could be interesting to you!

@womchile hola, tengo una esim con un numero chileno y se me hecho a perder el celular, y vivo fuera de Chile, que puedo hacer para recuperar la esim?

@fpingham We didn't see any improvements on evals with 4.1, and latency was worse than gpt-4o, so we didn't make the switch.

are people finding that 4.1 for agents makes some surprisingly stupid mistakes? are people using other models for convserational agents?

@BraceSproul Exampled project detailed! It helps a lot for visualizing more complex flows in practical apps.

I want to improve the applied AI content & work we put out at LangChain, and I'm curious to hear what people find the most valuable: Which (or multi) of the following do you find the most valuable, and want more of? - YouTube video - short, high level/conceptual walkthrough. - YouTube video - longer in depth code walkthrough. - Example project - smaller, weekend projects which make it easy to clone and get started locally - Example project - detailed, production ready apps (e.g Open Canvas or Social Media Agent) which require more setup locally, but are more powerful and actually useful. - Documentation - simple, single file type examples of real world "problems", but simplified down to make the concepts behind the implementation easier to grok Or other! If you have suggestions, please drop them below! My goal with this is to make it easier to onboard or improve your LangGraph skills, and build powerful and useful applications using it!

@ludwigABAP You are looking for LangGraph. It does exactly this.

LLM chat interfaces need tree-like branching conversations and the ability to jump back in at any splits don't tell me about context windows i do not care

I'm one week gone and Tencent drops probably one of the most excited updates to AI video generation since ControlNet! ToonCrafter is nuts 🤯 If you haven't heard, it lets you control the interpolation process by providing sparse keyframes. 10 wild examples:

Live with Restream twitter.com/i/broadcasts/1…

Check my new blog post on the inner workings of a modern GenAI app: affiliate.ai/post/a-technic… @hwchase17

Deal of the Day May 7: New liveProject series! Save 45% on my liveProject "Build a Custom Chatbot Using LangChain and ChatGPT" and other selected titles @ManningBooks: mng.bz/WrEx @hwchase17

🚀 Excited to launch my new course on Building a Chatbot with LangChain, in collaboration with Manning! 🔹 Master Retrieval Augmented Generation 📚 Experience real-time feedback & support on Manning's platform. Enroll now🔗shortener.manning.com/VxYX #AI #Chatbots #LangChain

Introducing my newest project: FriendGPT, a curated AI companion on Telegram that is just for you! Using the new @llama, @langchain, @supabase, @Redisinc, @render and @telegram. See a quick demo below and find it at friendgpt.app.

there's (exactly) seven ways to optimize latency in an llm application – just published a guide in the @openai docs that covers them! go check it out! (and let me know what you think) - link 👇-

@HomamMalk Good to know others are talking about this, it's the next big step to make llms more human-like. Check LangMem from the Langchain team, it's in alpha.

@MikePFrank Why don't you make it a 2 step process, where the inner monologue feeds another prompt that outputs the actual response.

I’ve noticed since giving my bots inner monologue capability that over time they seem to spend more and more of their time just inner monologuing, and increasingly forget to respond externally at all

@cyrusofeden @hwchase17 @jerryjliu0 Have you tried having a table per user?

The biggest frontier for LlamaIndex / LangChain right now is that they’re hard to use for multitenant apps — as in, most apps ever. I spent an hour trying to find an escape hatch, please show me the way @hwchase17 & @jerryjliu0 🙏

@granawkins That's assuming the winner will be decided by the end consumer. Normal people will probably never know what an LLM even is, while using apps powered by OpenAI, Google and OSS models in the background. Enterprise and SME is the battleground.

Apple will probably win LLMs. They'll release a model that runs locally/native on your iPhone, is good enough, has great privacy/security and has access to all your personal data. The advantages of larger foundation models will be negligible for 99% of tasks.

✒️Adding long term memory to OpenGPTs Wrote a blog on how we think about long term memory It's a bit abstract to think about, so included a practical example of how we added an implementation to OpenGPTs Blog: blog.langchain.dev/adding-long-te… Mini-🧵

@xaviermalina @minimaxir Because it outputs the json with the arguments so then you can pass them to a function.

@minimaxir When is function calling advantageous over a raw implementation of it in python? I couldn't figure out why it was useful

If OpenAI doesn't announce anything new for ChatGPT's function calling / structured data at their conference next week then I'm writing a comprehensive blog post on it. It's literally the most impactful feature of ChatGPT since ChatGPT but no one talks about it.