MLflow

🚀 MLflow Community, Your Insights Drive Innovation! 🧠 Our H2 2024 Community Survey is now open. In just 10-15 minutes, you can: - Directly influence MLflow's development - Share challenges in ML workflow management - Help shape the future of machine learning infrastructure Whether you're working with traditional ML, generative AI, or cutting-edge research, your voice matters to us. Take the survey and be a catalyst for change: Visit the Survey here: https://lnkd.in/gxExx_qb #MachineLearning #MLflow #GenAI #MLOps

Join the MLflow team for our first office hours of 2025! Whether you're new to MLflow or a seasoned contributor, this is your chance to: 🔍 Get help with your first (or next) contribution 💡 Learn about the latest GenAI features 🚀 Discuss the project roadmap with core committers 💬 Connect with the broader MLflow community The team will be available to answer questions, review PRs, and help you get the most out of MLflow's capabilities. When: January 8, 2024 Time: 5:00 PM PST / 8:00 PM EST Where: Virtual — https://lu.ma/x3fuq2o9 Register now to join us! #MLflow #OpenSource #MLOps #ML #LLMOps #AI https://lu.ma/x3fuq2o9

Check out this step-by-step guide to combining LlamaIndex, MLflow, Qdrant, and Ollama: 🔍 Integrate LlamaIndex with Qdrant for efficient vector storage and search 🚀 Use MLflow for model tracking, packaging, and evaluation 🔄 Implement Change Data Capture for real-time document updates ⚡ Optimize RAG workflows Read the full article: https://lnkd.in/gCnm9JgE

This week: MLflow Office Hours! Join us this Wednesday, January 8 at 5PM PST / 8PM EST. Bring your questions about contributing, get hands-on help with your PRs, and learn about our newest GenAI features. Whether you're a first-time contributor or longtime user, the MLflow team is here to answer your questions. Register here: https://lu.ma/x3fuq2o9 #MLflow #OpenSource #MLOps #AI #LLMOps

Join the MLflow team for our first office hours of 2025! Whether you're new to MLflow or a seasoned contributor, this is your chance to: 🔍 Get help with your first (or next) contribution 💡 Learn about the latest GenAI features 🚀 Discuss the project roadmap with core committers 💬 Connect with the broader MLflow community The team will be available to answer questions, review PRs, and help you get the most out of MLflow's capabilities. When: January 8, 2024 Time: 5:00 PM PST / 8:00 PM EST Where: Virtual — https://lu.ma/x3fuq2o9 Register now to join us! #MLflow #OpenSource #MLOps #ML #LLMOps #AI https://lu.ma/x3fuq2o9

MLflow's ChatModel class lets you build LLM applications with tool calling in just a few steps, with built-in support for tracing and deployment. ChatModel handles the complexity of LLM inputs and outputs for you, while giving you full flexibility to implement custom tools. In this tutorial, we show how to: 🔧 Build a weather tool with complete type safety using FunctionToolDefinition 📊 Get automatic tracing of all OpenAI calls and tool executions 🚀 Deploy your model with one CLI command using MLflow Models ChatModel is a subclass of PythonModel, so you get all the benefits of MLflow's production-ready model packaging and serving capabilities. Check out the full tutorial to build your first tool-calling model ⬇️ https://lnkd.in/gdTw8N2S #MLflow #LLMOps #AI #MachineLearning

Hello there #MLflow users! This new survey that's up is a great chance for you to really let your thoughts be known about many of the features that we've been building and releasing over the last year (and some older ones too!). If you could spare the time, it would greatly help the maintainer team to know where we can best focus our efforts over the next year. We truly want to address the pain points you're encountering, build the features that you really want, and refine the experiences that you already enjoy using.

MLflow 2.19 introduces built-in tracing support for CrewAI, giving you complete visibility into complex multi-agent workflows with just one line of code. CrewAI lets you build systems where multiple AI agents collaborate, each with specialized roles and tools. Multiple agents engaging in complex interaction patterns, retrieving outside information, and invoking tools can result in highly complex systems that are difficult to understand and debug. MLflow's CrewAI autologging makes debugging and iteration simple: 🔍 Track every agent interaction and tool call in real-time 📊 Visualize the complete hierarchy of tasks and handoffs ⚡ Enable comprehensive tracing with mlflow.crewai.autolog() 🔄 See exactly how your agents collaborate and delegate tasks The trace view provides full context for each operation, including task completion status, agent outputs, tool calls, and execution flow. This visibility is especially valuable when developing and refining multi-agent systems where understanding the interaction patterns is key to improving performance. #MLflow #AI #Agents #LLMOps

MLflow's ChatModel and PythonModel classes serve different needs when deploying GenAI applications. Here's when to use each: ChatModel simplifies GenAI deployment with standardized OpenAI-compatible interfaces. This means: 🔗 Immediate compatibility with existing OpenAI-based tools and workflows 🚀 Pre-defined model signatures that work out of the box 📊 Streamlined integration with MLflow's tracking and evaluation features PythonModel is your choice when you need complete control over: 🛠️ Custom input/output schemas for specialized use cases 🔄 Complex data transformations beyond standard chat patterns ⚙️ Fine-grained model behavior and deployment configurations For most conversational AI applications, ChatModel's standardized approach helps you avoid common deployment pitfalls while maintaining consistent interfaces across your GenAI services. Consider PythonModel when your use case requires specialized data handling or custom interaction patterns. See the comment below for links to in-depth tutorials on ChatModel 👇 #MLflow #LLMOps #MachineLearning #GenAI

Awesome place to jumpstart your GenAI project, packed with best practices and powerful tools! Excited to see the growing adoption of the MLflow + LLMOps combo!😄