hoping to get this in langchain soon! cc @aman_madaan
@langchain
-
LangChain Implements Conversational Agent Interface
By
–
Conversational Agent One pain point we heard was people trying to get agents to work in a conversational interface Although not perfect, we now have an implementation of this h/t @brunotorious for driving a lot of this, + @nfcampos & @andrewgleave https://
langchain.readthedocs.io/en/latest/modu
les/memory/examples/conversational_agent.html
… -
Tool Priority: Guide to LLM Tool Selection
By
–
Tool Priority First up, an example of how to create some sense of priority amongst tools Basically just tell the LLM (with language) when to use which tools Thank you @Yoooongtae for creating this excellent guide! https://
langchain.readthedocs.io/en/latest/modu
les/agents/examples/custom_tools.html#defining-the-priorities-among-tools
… -
LangChain Direct Return Flag for Tool Responses
By
–
Direct Return Sometimes, if a specific tool is chosen, you may just want to return the response of that directly to the user There is now a flag to easily allow for exactly that Thank you @andrewgleave @brunotorious for the suggestion! https://
langchain.readthedocs.io/en/latest/modu
les/agents/examples/custom_tools.html#using-tools-to-return-directly
… -
OpenAI API Update: Logprobs and Agent Tool Improvements
By
–
v0.0.58 Add generation reason and logprobs to @OpenAI response – @diwanksingh Add @OpenAI logit bias param – @benderville bug fixes (h/t @shivanshug11 , @brain_exe_ai ) And a bunch of Agent & Tool improvements! So many we can't list them all in this tweet ->
-
v0.0.57 Release: Documentation and OpenAI Wrapper Improvements
By
–
v0.0.57 Document improvements – @eltociear @rubensmau OpenAI wrapper improvements – @Yoooongtae Conda developer guide docs – @samcwl Different schemas for cache – Jason Gill
Fix bug in LLM caching – h/t @diwanksingh Link checking action – @HunterGerlach -
First Paper Using LangChain Framework Published
By
–
The first paper (that we know off) to use @langchain
! @johnjnay -
LangChain Contribution Opportunity for App Developers
By
–
A really cool opportunity for anyone thinking of contributing to (or building interesting apps on) LangChain!
-
LangChain Memory Human Prefix Parameter Now Customizable
By
–
Add human prefix for memory Previously, in memory the user input was hardcoded to "Human" Now, thanks to Andrew Huang, it is exposed as a parameter that can be controlled https://
langchain.readthedocs.io/en/latest/modu
les/memory/examples/conversational_customization.html#human-prefix
… -
LangChain Improves Agent Max Iterations Handling with Final LLM Pass
By
–
More graceful handling of agent max iterations Previously, when an agent reached max iterations, it just stopped with a fixed message Now it makes one final forward pass through the LLM to generate a best guess final answer h/t @johnjnay for idea https://
langchain.readthedocs.io/en/latest/modu
les/agents/examples/max_iterations.html
…