Tutorial > A first attempt at DSPy Agents from scratch learnbybuilding.ai/tutorials/dspy…

Notebook > Multi-Agent DSPy Programs: Bootstrapping & Aggregating Multiple ReActAgents github.com/stanfordnlp/ds…?

Someone's asked me, what optimization #DSPy framework does under the hood (youtu.be/NXI2l0wJNBY). I think a good explanation of that is in the paper and some clues can be found in code (github.com/stanfordnlp/ds…).

Whenever you compile a program with BootstrapFewShot as an

It’s happening, over 500+ signups? A packed venue at Edge, amazing partnerships with Arize AI cohere and of course Weaviate • vector database. Excited to learn more about DSPy from Omar Khattab!

What is the effect of quantizing the score in DSPy?
It's not what I expected:

In DSPy, increasing the number of candidate programs seems more helpful than increasing the number of bootstrapped demos

Starting a new module on how to build a Retrieval-Augmented Generation system using FastAPI, OpenAI, PostgreSQL, and pgvector. Frameworks like DSPy, LangChain, Llama-index, or even EmbedChain are good for fast prototyping but can be cumbersome when you want to build a system you

I struggled to pick up #DSPy so now that I feel semi-proficient I published Part 1 a 'DSPy User Guide' that contains some simple code snippets like this that I hope will help others who also prefer to learn with their fingers.

Interesting to see more information coming out about fine tuning (link in the comments)

DSPy, DSPy, DSPy - if you're using examples / in context learning, you might be on the right track

Not sure any of this is new info, but interesting to see the research!

Thanks for waiting, here we go.

dSPY and dTSLA now LIVE on app.seahorsefi.xyz

Collateral factor is 18%, more yield strategies now made possible.

$DFI #Defichain

DSPy allows you to nest bootstrapping, meaning you can use the bootstrapped program as a teacher:

DSPy Is All You Need ✨

Programming over prompting philosophy of DSPy ✨

Within minutes of compiling, a few lines of DSPy code allow GPT-3.5 and llama2-13b-chat to self-bootstrap pipelines that outperform standard few-shot prompting (generally by over 25% and 65%, respectively)

DSPy talk by Omar Khattab :

Learn to program optimized prompts

cohere Stanford NLP Group Arize AI Weaviate • vector database

Learn how to build a financial analyst copilot that will help you decide which stocks to buy or sell. Powered by DSPy, OpenAI, and a Weaviate database, this tool enables your agent to remember past information.

What is amazing about this agent is that you can actually access its

We programmed cohere Command R+ with Stanford NLP Group DSPy, Weaviate • vector database Vector store and crewAI agents to create a matchmaking application.

By using DSPy's declarative structured programming, we've built a highly modular and scalable AI system that effortlessly connects

Is #DSPy still hot? I hope it is, cause my new video is out where I'm explaining the approach providing a self-improving business case on company valuation

Imagine you can prompt Perplexity for company valuation and then go check the sources. If you don't like the result,

got a basic wandb setup for my dspy experiments. have to take a dpsy break now but can't wait to integrate this with a configuration entry point at the top and launcher+sweeps so I literally don't have to do any work, 'just' gather data

Tried out using a compiled DSPy program to optimize this same program again and it helps quite a bit

'Turning chicken shit into chicken salad'.. never heard that one until Connor Shorten mentioned at DSPy event

Lately we've been exploring several open-source frameworks, in addition to building our own in-house abstractions to create more robust AI systems. Take a look at our learnings and deep dive into DSPy.

🔗 gradient.ai/blog/achieving…
✅ GPT-4 Performance at 10x Lower Cost
✅