DSPyWeekly Issue No #11

Self-Evolving Agents - A Cookbook for Autonomous Agent Retraining

Agentic systems often reach a plateau after proof-of-concept because they depend on humans to diagnose edge cases and correct failures. This cookbook introduces a repeatable retraining loop that captures those issues, learns from the feedback, and promotes improvements back into production-like workflows. We ground the approach in a regulated healthcare documentation task, but the patterns generalize to any domain that demands accuracy, auditability, and rapid iteration.

Teaching Local Models to Call Tools Like Claude

Distillation works like a tutor training a student : a large model teaches a smaller one.1 As we’ve shifted from knowledge retrieval to agentic systems, we wondered if there was a parallel technique for tool calling. Could a large model teach a smaller one to call the right tools?

DSPy Neo4j Integration

DSPy is a framework for algorithmically optimizing LM prompts and weights, especially when LMs are used one or more times within a pipeline. The Neo4j integration allows for vector search.

Track DSPy Events

Excited to launch a brand new "Events" section on DSPyWeekly.com! This new page is your central hub for finding and sharing DSPy-related events happening around the world. Whether it's a local meetup, a virtual workshop, or a major conference talk. Also has an RSS feed you can subscribe to.

PAPER: Feedback Descent: Open-Ended Text Optimization via Pairwise Comparison

The text introduces Feedback Descent, a framework for optimizing text-based items (like prompts, code, or molecules) using detailed, structured textual feedback instead of simple scalar rewards (like a "good/bad" score).99.9 Uses DSPy. th percentile of a database with more than 260 , 000 compounds across six protein targets.

DSPy package registry, tweet

Codex-Agent is a module that wraps the OpenAI Codex SDK in DSPy signatures for type-safe, stateful coding agents. Each instance maintains its own conversation thread with full execution visibility. Now packaged and customizable to run in a few lines of code

DSPy in Production - Customer Case Study

Koantek’s AscendAI Agent Factory redefines how enterprises build and deploy AI agents on Databricks. Powered by Agent Bricks and DSPy’s “programming over prompting” paradigm, it enables businesses to generate complete agent projects using pre-built templates such as Text Classifier, Data Extraction, or SQL Generator, or through custom specifications.

vedant007-v/codex_dspy

🤖 Simplify multi-turn conversations with CodexAgent, a DSPy module for OpenAI Codex, offering stateful threads and rich, typed outputs. | Language: Python | License: MIT License

nafew-azim/AUTODSPy

RL-driven framework that composes modular DSPy pipelines and teleprompters to improve LLM reasoning (experiments use GPT-2). | Language: Jupyter Notebook | License: MIT License

Philosophy behind DSPy

DSPy is a research paradigm. We're not trying to building whatever reaches the largest userbase. We're trying to realize a very specific vision and design better algorithms over time to demonstrate that it's possible to systematically engineer AI systems.

Stay Updated with DSPyWeekly!