There are many more types of memory we hope to add! @sjwhitmore has been pioneering this effort so far, and we hope a lot of these come from the community. Check out this idea from @BrennanErbz for inspiration
CODE
-
Contribute Memory Implementations to LangChain Core Library
By
–
If you do have a memory implementation you think is generally useful, we'd LOVE to have it contributed back into the core library! We are super excited to see what people build with this https://
github.com/hwchase17/lang
chain
… -
Getting Started with Memory in Conversation Chains
By
–
Here is a notebook for getting started with memory. It walks through maybe the most clear application of memory: a conversation chain. It showcases short term memory – remembering the conversation to date https://
langchain.readthedocs.io/en/latest/gett
ing_started/memory.html
… -
Add Memory to Any LLM Chain and Agent
By
–
But you don't just have to use the predefined conversation chain! Here's how you can add memory to any LLM chain: https://
langchain.readthedocs.io/en/latest/exam
ples/memory/adding_memory.html
… And here's how you can add memory to an agent: https://
langchain.readthedocs.io/en/latest/exam
ples/memory/agent_with_memory.html
… -
Types of Memory in LangChain: Short-term and Long-term
By
–
There are different types of memory, eg: Short term memory, like what the previous user input was Long term memory, like what the user previous said about a particular person or place Both of these can fit into this memory abstraction. Let's take a closer look at some
-
LangChain Chains Now Support Stateful Conversations and Memory
By
–
So far all the chains/agents in LangChain have been stateless, meaning each interaction is independent This is far from ideal for some real world settings (eg chatbots) where you want the chain/agent to remember previous conversations
-
Memory Module: Loading and Saving Data in Chains
By
–
To achieve this, we added the the concept of a memory module This module: Loads things from memory before passing user input to the chain Saves things to memory after the chain is finished
-
ConversationChain and Memory Module in LangChain
By
–
This past Friday @sjwhitmore added the very cool ConversationChain As part of doing this, she also added the concept of a Memory module. Here's what that is and why it's useful, :
-
Whisper Fine Tuning Event: Architecture and VoxPopuli Dataset
By
–
Monday just got a whole lot more interesting! We're excited to announce 2 talks as part of the upcoming Whisper Fine Tuning Event. 1. @OpenAI
's @_jongwook_kim on the Whisper architecture.
2. @MetaAI
's @ChanghanWang on the VoxPopuli dataset. https://
youtube.com/watch?v=fZMiD8
sDzzg
… -
LangChain Version 0.0.23 Release with New Features and Bug Fixes
By
–
LangChain Version 0.0.23 new contributors! Unicode fix for windows (
@mark_k
) Self-ask w/ search bug fix (h/t @Raza_Habib496 for finding it) Readme update (
@dillchen
) @huggingface Hub embedding support (
@baga_tur
) Save prompts to disk (
@AkashSamant4
)