Open source question! Any tips on managing PRs? Some Qs… – Any standards on deciding what/when to merge? How to communicate?
– Can you… just take parts of a merge?
– Any common mistakes to avoid?
– How do you balance complexity (can do more) vs simplicity (easier to build on
CODE
-
Managing Pull Requests in Open Source Projects
By
–
-
Rasa Framework Updates on Conversational AI Agents
By
–
@alanmnichol from Rasa has posted about this I think
-
Task Serialization and Parallel Execution Strategies
By
–
I want to set up task serialization (only do this task after this task), which can then lead to parallel tasks that don’t rely on each other. Curious about your suggested approach
-
Removing Langchain for Better AI Architecture Design
By
–
Thanks!! I stripped out Langchain for two reasons (1) I didn’t think it was set up correctly and I wanted some time to think on how to do it right, and (2) it felt a little opinionated and wanted to share a barebones structure first to make it easy for people to iterate on.
-
Exploring Llama and LlamaIndex AI frameworks
By
–
Ohhh sorry I was confusing w Llamaindex. I haven’t tried llama, I should dig in
-
LangChain Executor Agent Implementation Cleanup
By
–
I have a version w Langchain as executor agent, just need to clean up a bit
-
GPT-4 Replit Integration Automates Code Fixing
By
–
@Replit and @openai #GPT4 combine to make it insanely easy for morons like me to play with code. See here how http://
wolverine.py by @biobootloader uses GPT4 to fix code right in the repl. You can actually see your code getting fixed…amazing. https://
loom.com/share/6dbf0e74
97f24c7789583dc52c67d556
… -
Hybrid Search Strategy for LLM Date Filtering and Vector Queries
By
–
So I think this is a case where you might to use hybrid search. You can store the dates as metadata, and have the LLM create an array with the date to filter by, and then a query for the vector search.
-
Language Models can Solve Computer Tasks
By
–
A computer agent that can understand natural language and perform tasks on a computer. They used a large pre-trained language model and a simple prompting scheme that helps the agent improve its output. https://arxiv.org/abs/2303.17491