Stay Updated with DSPyWeekly!
đź“š Articles
Engineering VP Josh Clemm on how we use knowledge graphs, MCP, and DSPy in Dash - Dropbox
Josh Clemm, VP of Engineering at Dropbox, details the technical architecture behind Dropbox Dash, a universal search tool designed to navigate proprietary work content by leveraging a robust "context engine" rather than simple federated retrieval. Dash prioritizes index-based retrieval to ensure speed and company-wide access, utilizing knowledge graphs that model cross-app relationships—such as connecting meetings to attendees and documents—via "knowledge bundles" integrated into standard indices rather than high-latency graph databases. To mitigate the context window and speed limitations of the Model Context Protocol (MCP), the team employs "super tools" and specialized sub-agents. Furthermore, Clemm highlights the use of "LLM as a judge" to rigorously evaluate retrieval relevance and the adoption of DSPy for programmatic prompt optimization, which facilitates efficient prompt management and seamless model switching at scale.
dspy-evaluation-suite Agent Skills by OmidZamani /OmidZamani/dspy-skills | AgentSkillsRepo
This skill should be used when the user asks to "evaluate a DSPy program", "test my DSPy module", "measure performance", "create evaluation metrics", "use answer_exact_match or SemanticF1", mentions "Evaluate class", "comparing programs", "establishing baselines", or needs to systematically test and measure DSPy program quality with custom or built-in metrics.
BigSpin ( Startup using DSPy )
The technical foundation of Bigspin is the open-source DSPy project
PAPER - Unpacking Generative AI in Education: Computational Modeling of Teacher and Student Perspectives in Social Media Discourse
This paper, "Unpacking Generative AI in Education," employs computational modeling to analyze the divergent perspectives of teachers and students regarding Generative AI (GenAI) in educational settings by examining social media discourse on Reddit. Using Natural Language Processing (NLP) techniques such as sentiment analysis and topic modeling (including BERTopic, Top2Vec, and LDA), the study reveals a distinct "AI divide": educators primarily focus on challenges related to academic integrity, plagiarism detection, and the need for robust policies, often expressing negative sentiment. DSPy is being used here.
🎥 Video
DSPy Introduction - (Korean Language)
Taught by a Professor https://github.com/junji64/DSPy/blob/main/DSPy.ipynb
DSPy Explained (Databricks Demo): Build Model-Agnostic Agents + Auto Prompt Optimization (GEPA)
Prompt engineering doesn't scale—especially when models change, prompts drift, and your “logic” lives inside a giant string. | Channel: VectorLab
🚀 Projects
Microcode - TUI based agent for RLM
Open sourcing Microcode! Microcode is a context-efficient, general purpose terminal agent fully powered by a packaged `dspy.RLM()` program. Set your own OpenRouter API key via and choice of models. Supports MCP servers too with @MaximeRivest mcp2py.
karthikscale3/dspyground
A tool kit for generating high quality prompts using DSPy GEPA optimizer | Language: TypeScript | License: Other
PaulLockett/Storyhost
A very baseline version of an AI assitant in Minecraft | Language: Python
leockl/vidspy
DSPy-style framework for optimizing text-to-video generation using VBench metric feedback | Language: Python | License: MIT License
đź’¬ Discussion
DSPy + Claude as skill
Raveesh put together a skill that allows folks to try DSPy and GEPA and get a feel of how it works.
Eito Miyamura Long Form Twitter Artcile
This is a mock summary for the article at https://x.com/Eito_Miyamura/status/2014757193766093069.
Thanks for reading DSPyWeekly!