Closed paul-paliychuk closed 3 weeks ago
@tmc Hi, Zep Engineer here. we currently support TS and Python LangChain integrations, and have recently released our go-sdk. Excited to deliver LangChain integration for our golang users too! Please let me know if you have any questions/suggestions about the implementation. Would love to hear how langchaingo users might prefer to use our memory.
For now I set up an example using ConversationChain, but we have a bunch more stuff we'd like to port to Go.
Also I notice that the build examples is failing, this is due to /memory/zep import not being available yet.
Also I notice that the build examples is failing, this is due to /memory/zep import not being available yet.
for example don't need use go.mod , can import only package
@devalexandre Hi, removed go.mod from example project, also upgraded zep sdk version with more straightforward initialization.
@tmc could review it?
@paul-paliychuk why update to draft?
@paul-paliychuk LMK when this is ready for review (or is itnow?)
@tmc apologies, we were refactoring our go sdk api recently, will update PR components soon and mark as ready to review. Thanks!
@tmc @devalexandre Ready for review
Sorry about the delay here, this looks great! A small comment but I think this can go in as-is and be fixed separately.
@tmc Appreciate the review, will open a docs PR this week
PR Checklist
memory: add interfaces for X, Y
orutil: add whizzbang helpers
).Fixes #123
).golangci-lint
checks.Creates a zep package that aims to integrate with zep cloud memory, link to our docs. I set up zep.ChatMessageHistory and zep.Memory following your BufferMemory implementation. This can be used by users who want access to long term memory and provide good context to their llm prompts. After each message (or memory) is added to a session, we generate summaries and extract facts from the conversation. When we get memory, we inject the most relevant summary and conversation facts into a system prompt.