For those who couldn't attend my 'Keynote' at @aiDotEngineer
> Pydantic is all you need
Making LLMs, software 3.0, backwards compatible with software 1.0
youtube.com/live/veShHxQYPzo…
other links in the thread!
Combining @pydantic, @OpenAI function calling, and @langchain for data analysis is insane!
Suppose you want structured data on healthcare startups, but you're met with nothing but blogs and articles.
How can you transform this mess into the key info you need?
It's easy! 🧵
Now we can move on to the fun part! Using @langchain, we initialize an @OpenAI chat model. We then convert our Pydantic model into a function and pass it along with the original text to the chat model.
And that's it! We now have our structured data extracted from the text. In just a few lines of code, we can easily parse the output and have it ready to be saved, displayed, or whatever else you intend to do with it.
We're releasing 60 lessons & practical projects on LangChain & Vector Databases in Production as our inaugural course within the Gen AI 360 Certification.
Enroll at learn.activeloop.ai
Why should you take it?
1. Generate Picture Books with AI for free (code open-source👇) with @OpenAI Function Calling, @langchain, #DeepLake, & @StabilityAI.
- Prompt -> a PDF storybook with illustrations.
- Stores image & text pairs in the multimodal #DeepLake VectorDB for model training/finetuning!
(1/14) Introducing: SalesCopilot, an open-source alternative to @Gong_io and @chorus_ai, built with #DeepLake & @langchain.
- Transcribes calls in real-time
- Detects customer objections
- Suggests responses based on a custom knowledge base stored in Deep Lake with GPT-4
Demo
I just finished building SalesCopilot, an open-source AI-powered sales call assistant - real-time transcription, automated objection detection and handling, GPT-3.5/4 powered chat, and more! reddit.com/r/MachineLearning…
I just added the option to use hypothetical document embeddings (HyDE) with BriefGPT, which improves retrieval performance. I also added a page to compare HyDE vs normal retrieval, check out the difference for yourself! Thanks @langchain@streamlitgithub.com/e-johnstonn/Brief…
Sales call assistant that I'm working on: reads the transcript of the call in real-time, detects objections within 5 seconds, classifies the objection, then tells you how to respond. You can choose the guidelines that the assistant uses, I used blog.hubspot.com/sales/handl…
Sales call assistant that I'm working on: reads the transcript of the call in real-time, detects objections within 5 seconds, classifies the objection, then tells you how to respond. You can choose the guidelines that the assistant uses, I used blog.hubspot.com/sales/handl…
I just published my latest project, a tool that transcribes audio in real-time and integrates with ChatGPT to give you a bot who knows the conversation better than you. Lots of cool applications!
Thanks to @langchain@trychroma@OpenAI@hwchase17!
github.com/e-johnstonn/wingm…
Oh, and you can input a name and save conversation transcripts with them. You can then load them and continue saving/adding to them, building a database of conversations. This database is then queried for contextually relevant information to give even more context to the bot!
Built my own locally hosted tool for document summarization and querying, and just added compatibility with locally-run LLM's! Thank you @langchain@hwchase17 for the awesome work you do that enabled this!
github.com/e-johnstonn/Brief…#langchain