We introduced the ParentDocumentRetriever last week to strike a balance between: – using small chunks during indexing
– passing larger chunks to the LLM Great overview and diagram by @clusteredbytes
@hwchase17
-
ParentDocumentRetriever balances chunk size for LLM optimization
By
–
-
Apps Embracing Open Source Models for Better Integration
By
–
Love to see more apps adding support for open source models!
-
Speaking at AWS Gen AI Day on Platform Integrations
By
–
Excited to be speaking at AWS Gen AI day! Lot's of awesome integrations with AWS platform, was fun to highlight them all https://
genaiday.virtual.awsevents.com/register?trk=5
7dbe57a-49c3-432b-a93b-0988d5011812&sc_channel=sm
… -
GPT Researcher Integrates Multiple LLM Providers and LangSmith
By
–
Excited to collaborate with @assaf_elovic to bring multiple LLM providers + an integration with LangSmith to GPT Researcher – the best research agent I know of! I'd highly encourage folks to try it out – it's very impressive
-
XML Tags as Prompt Engineering Technique for AI Applications
By
–
The exact xml tags you use should probably be pretty application specific, and should be treated as prompt engineering This is just a way to get started/inspiration
-
Generative Agents Simulation Logs and Complex Prompts Visibility
By
–
Really cool to be able to see the logs & complex prompts that went into the Generative Agents simulation! https://
smith.langchain.com/public/a4b09aa
f-99ec-4ad0-a58c-2c07741f611b/r
… -
LangChain Expression Language enables parallel operations optimization
By
–
If you do these operations sequentially, they can start to add up That's why it's important to do them in parallel One nice attribute of the LangChain Expression Language is it tries to do things in parallel where ever possible
-
LangSmith Parallel Execution Optimization in Runnable Maps
By
–
If we look at the LangSmith trace, we can see that the overall RunnableMap step is less than the sum of its components (because those components are run in parallel
-
LangChain RunnableMap Parallel Operations Guide
By
–
In this case, we have a `RunnableMap` (the dictionary) For these operations, they are run in parallel See our guide on that here: https://
python.langchain.com/docs/guides/ex
pression_language/interface#parallelism
… -
Multi-source retrieval API calls optimization challenges
By
–
Each retrieval step requires it own API calls: – For SQL, you need to call an LLM to generate a query
– For vectorstore, you need to create an embedding And this is just for two retrieval sources – but it's easy to imagine extending this to more.