Hey guys! I just finished the project I was showing on the last call. For those who were not there, it's a benchmarking environment to compare different semantic chunking methods from Langchain in both scientific and medical domains. It has a custom scoring system which consists of both chunk size distribution and quality of retrieval. Feel free to read the paper or actually run some tests yourself: https://github.com/monami44/Langchain-Semantic-Chunking-Arena For a video walkthrough go here: https://www.youtube.com/watch?v=7_YNeJvcAtM

Hey everyone! 👋 I’ve got to cancel this week’s coaching call. CrewAI, my main gig, is crazy busy right now, and we’re deep into Agents Week. If you’re curious about what we’re up to, check it out here: 👉 https://week.crewai.com Also, at the exact same time as our usual call, CrewAI will be hosting a live event on YouTube. If you want to catch some cool stuff happening in real-time, you can tune in here: 🎥 youtube.com/@crewAIInc Oh, and be sure to keep an eye out tomorrow—CrewAI has a huge announcement coming your way! 🚀 Quick Flows Update 🌀 • Bad news: I need to re-record the Flows tutorial because we made some tweaks to the tool. • Good news: The tool is now simpler than ever! I’ll be recording the new video late tonight, and my editor will have it ready by late Wednesday or early Thursday. Sorry again for the short notice! I hope you all have an amazing Tuesday, and I’ll see you on next week’s call. Thanks so much for your understanding—you guys are the best! 🙏

Langchain recently uploaded a tutorial video to build a ReAct-based voice agent using the OpenAI Realtime API!! They also showed how to set the prompt instructions and tools to fully customize your agent 👌 Link to the video: https://www.youtube.com/watch?v=TdZtr1nrhJg

Hey everyone! I am working on building a RAG application using crewai and unable to get past PDFSearchTool tool from rag-deep-dive video. If I enter custom embedding model configurations, I get an error saying "TypeError: cannot pickle "classmethod" object And if I don't provide embedding model configurations and just give the api key and base url, there is an SSLZeroReturnError where inserting batches in chromadb fails. Would really appreciate help!