Repository navigation
3.5.0 #10866
isaacbmiller
announced in
Announcements
3.5.0
#10866
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
DSPy 3.5: A Lighter, Faster Foundation
DSPy 3.5 brings a lighter LM layer, graduated core modules, and new History and image capabilities for RLM.
Upgrading from 3.4.1
This release changes prompts, defaults and APIs. Read the upgrade guide for breaking changes, migration steps and compatibility limits, and re-evaluate optimized programs.
Highlights
RLM, Flex and GEPA graduate
dspy.RLM,dspy.Flexanddspy.GEPAare no longer experimental.ReActV2 now replaces ReAct
ReActV2 is now simply
dspy.ReAct, with structured History and a typedsubmittool; the old name is deprecated until 3.6.Rich tool results
Tools can return images, files and other DSPy content parts on supported provider routes.
RLM History — authored by Matt Mannucci (@mannuch)
RLM can save and reload conversation History for later calls; it does not restore live interpreter variables.
A lighter LM foundation — lm15 work by Maxime Rivest (@MaximeRivest), integration by Isaac Miller (@isaacbmiller)
LiteLLM and OpenAI’s Python library are now optional: DSPy calls supported providers through the lightweight lm15 engine.
Automatic API-family routing
model_type="auto"is the default, rather than explicitly selecting chat or responses.XMLAdapter is now the default Adapter
XMLAdapter now formats and parses predictor calls by default, changing prompts and the output contract even when your Signature stays the same.
Native reasoning and explicit plans
Prefer native reasoning; request an explicit plan when planning before acting helps the task or it needs a human-auditable plan.
Disk caches: secure by default
Restricted loading is now the default, and plain NumPy arrays are stored without pickle.
Image-capable RLM
RLM can crop images with Pillow and send mixed text/image subqueries from its Python REPL.
With DSPy imported, an image-capable LM configured and the default Deno runtime available:
Prewarmed Interpreters
Real DSPy now runs inside RLM and Flex interpreters, with a precompiled standard library, cached startup snapshots and
InterpreterPoolfor prewarming.Medians of nine fresh processes per path, with prepared runtime caches; pool readiness costs another 0.55 s, excluded from the pooled forward. First-use preparation can be slower.
Full PR List
All 99 PRs in the verified range appear once in the list below. Earlier intermediate bundled-lm15 and metadata updates are included for completeness; the cutoff contains bundled lm15 2.2.0 and metadata sourced from released LiteLLM 1.108.3.
Language Models, Adapters, and Reasoning
Agents, History, Tools, Images, and Interpreters
Optimization, Evaluation, Core APIs, and Caching
Documentation
Dependencies, Metadata, CI, Typing, and Release Engineering
Contributors
Thank you to @Aditya17-bot, @Blacksujit, @dbreunig, @divakantpandey, @emme1t, @fredheir, @isaacbmiller, @kausthubhk, @michaelisaac-dev, @SaiMouli3, @shaikn6, @shi-felix, @tomron, @Virtuoso461, @Zhuoxi2000, @Zsanz3, @mannuch and @MaximeRivest. Automation contributions: @dependabot and @github-actions.
Full Changelog: 3.4.1...471b3fb
This discussion was created from the release 3.5.0.
All reactions