Awesome Langchain Weekly Digest - w23/2023
Gorilla, Voyager, Andrej Karpathy at Microsoft Build, Tips, and more...
Welcome to another bustling week in the LangChain ecosystem! It's truly inspiring to witness the continuous advancements and the rise of Open Source LLM. This week, we have many new features, improvements, and resources to share, including an insightful overview of GPT from Andrej Karpathy and resources in the growing `No OpenAI` trend. Let's dive right in!
Langchain Updates
Document loaders
NotionDB ported to JS/TS
Github issues and PRs
Integrations
Momento Cache, to reduce costs
Twilio to send SMS, make sure that your agent does not start chatting with its peers! And an article demoing Build a Basketball SMS Chatbot with LangChain Prompt Templates in Python.
C Transformer Interface opens the door to connect LLMs ported by Georgi Gerganov running on lower-cost machines (like llama.cpp).
Qdrant improvements: Richer filtering support, Self-querying retrieval, Batched upserts
Amazon Bedrock Models: now has LLM and embedding model wrappers for Amazon's new Bedrock service, which serves both Amazon's own Titan models and 3rd party models like Anthropic's Claude.
Arguilla open-source data curation platform for LLMs, empowering data teams from data labeling to model monitoring.
Output Parsers
Agents
Multi-dataframe agents shows how the agent can interact with multiple csv files passed in as a list.
Generative Agents This script implements a generative agent based on the paper Generative Agents: Interactive Simulacra of Human Behavior by Park, et. al.
Changelog
This is a curated list of updates. For a comprehensive look at all updates on both Python and JS, review here:
Awesome Projects
Checkout those awesome projects showcasing the versatility of LangChain:
Gorilla: An API store for LLMs. Enables LLMs to use tools by invoking APIs. Given a natural language query, Gorilla comes up with the semantically- and syntactically- correct API to invoke.
Voyager: An LLM-powered embodied lifelong learning agent in Minecraft that continuously explores the world, acquires diverse skills and makes novel discoveries without human intervention.
WingmainAI: A tool for interacting with real-time transcription of both system and microphone audio.
Local GPT: Inspired by Private GPT with the GPT4ALL model replaced with the Vicuna-7B model and using the InstructorEmbeddings instead of LlamaEmbeddings
Articles & Videos
This week's picks include articles on implementing GPT4All with LangChain, training your own LLM using privateGPT, and LangChain Autonomous Agent - BabyAGI. Don't miss the collection of videos, including a comprehensive course on LangChain and an overview of the state of GPT by Andrej Karpathy.
Training Your Own LLM using privateGPT: Learn how to train your own language model without exposing your private data to the provider
LangChain Autonomous Agent — BabyAGI , integrated as Web App
Auto-Evaluation of Metadata Filtering by Lance Martin looks at how to improve retrieval using metadata filtering.
Interesting thread from Harisson on condensing chat history to improve performances and reduce cost using 2 models: https://twitter.com/hwchase17/status/1663576081918742530
And a visual overview of LangChain retrievers:
source: https://twitter.com/pwang_szn/status/166312305009794662
New LangChain course
When Harrisson collaborates with Andrew Ng, we get another awesome piece of content with this 1-hour course on LangChain. Check it out:
🎥 LangChain for LLM Application Development
State of GPT
One of the more comprehensive videos of this year about LLMs and prompt engineering. Summarize so many notions and concepts developed recently and a good overview and understanding of building LLMs. Andrej Karpathy is, as usual, very good at sharing his knowledge.
Converting a LangChain App from OpenAI to OpenSource
Another great video from Sam Witteven walks us through building a retrieval QA application with Chroma and OpenAI, then converting it to open source models.
Testing LLM with GPT4All
🎥 Testing Free Large Language Models with Steaminlit, LangChain and GPT4All by Nicholas Renotte
Other News
Is LangChain the new LAMP stack? THENEWSTACK takes on the incredible growth of langChain: LangChain: The Trendiest Web Framework of 2023, Thanks to AI
The open source Vector Toolkit for Postgres, Use the Supabase client libraries to store, index, and query your vector embeddings at scale. https://supabase.com/vector
That's All for Now!
Thank you for being a part of this exciting journey with us. The LangChain ecosystem wouldn't be the same without its vibrant community. If you enjoyed this week's digest, please consider sharing it with others. With more people involved, our content becomes richer and more diverse.
Your input is invaluable to us. If we missed something or you'd like to suggest a project, please make a PR, add an issue on our GitHub repository, or email us at awesomelangchain[at]kyrolabs.com. We can't wait to hear from you!
Stay curious, stay awesome, and see you next week!
The team at Kyrolabs lovingly curates this newsletter.