Skip to content

armgit / dspy

Public Imported from GitHub
0

DSPy: The framework for programming—not prompting—language models

Python

docs build fix

Herumb Shandilyad677e4623061
diff --git a/docs/docs/dspy-usecases.md b/docs/docs/dspy-usecases.md
index a0b8f645d..212dc08f5 100644
--- a/docs/docs/dspy-usecases.md
+++ b/docs/docs/dspy-usecases.md
@@ -115,4 +115,4 @@ TODO: This list in particular is highly incomplete. There are dozens of other go
### Weaviate has a directory of 10 amazing notebooks and 6 podcasts!
-Huge shoutout to them for the massive support <3. See the [Weaviate DSPy directory](https://weaviate.io/developers/weaviate/more-resources/dspy).
+Huge shoutout to them for the massive support ❤️. See the [Weaviate DSPy directory](https://weaviate.io/developers/weaviate/more-resources/dspy).
diff --git a/docs/docs/roadmap.md b/docs/docs/roadmap.md
index 623d40964..9596506c8 100644
--- a/docs/docs/roadmap.md
+++ b/docs/docs/roadmap.md
@@ -1,3 +1,8 @@
+---
+draft: true
+---
+
+
# Roadmap Sketch: DSPy 2.5+
It’s been a year since DSPy evolved out of Demonstrate–Search–Predict (DSP), whose research started at Stanford NLP all the way back in February 2022. Thanks to 200 wonderful contributors, DSPy has introduced tens of thousands of people to building modular LM programs and optimizing their prompts and weights automatically. In this time, DSPy has grown to 160,000 monthly downloads and 16,000 stars on GitHub, becoming synonymous with prompt optimization in many circles and inspiring at least a half-dozen cool new libraries.
@@ -21,7 +26,7 @@ To a first approximation, DSPy’s current user-facing language has the minimum
## Team & Organization
-DSPy is fairly unusual in its technical objectives, contributors, and audience. Though DSPy takes inspiration from PyTorch, a library for building and optimizing DNNs, there is one major difference: PyTorch was introduced well after DNNs were mature ML concepts, but DSPy seeks to establish and advance core LM Programs research: the framework is propelled by constant academic research from programming abstractions (like the original [Demonstrate–Search–Predict]() concepts, DSPy [Signatures](), or [LM Assertions]()) to NLP systems (like [STORM](), [PATH](), and [IReRa]()) to prompt optimizers (like [MIPRO]()) and RL (like [BetterTogether]()), among many other related directions.
+DSPy is fairly unusual in its technical objectives, contributors, and audience. Though DSPy takes inspiration from PyTorch, a library for building and optimizing DNNs, there is one major difference: PyTorch was introduced well after DNNs were mature ML concepts, but DSPy seeks to establish and advance core LM Programs research: the framework is propelled by constant academic research from programming abstractions (like the original **Demonstrate–Search–Predict** concepts, DSPy **Signatures**, or **LM Assertions**) to NLP systems (like **STORM**, **PATH**, and **IReRa**) to prompt optimizers (like **MIPRO**) and RL (like **BetterTogether**), among many other related directions.
This research all composes into a concrete, practical library, thanks to dozens of industry contributors, many of whom are deploying apps in production using DSPy. Because of this, DSPy reaches not only of grad students and ML engineers, but also many non-ML engineers, from early adopter SWEs to hobbyists exploring new ways of using LMs. The following team, with help from many folks in the OSS community, is working towards the objectives in this Roadmap.