Releases·Tool Release
Imp is a full port of the DSPy LLM framework to the BEAM, published on hex.pm
Imp, a full port of DSPy to the BEAM, surfaces on Hex
104 engagement1 sourceSun, Sep 27, 2026, 07:28 PM
AI Overview
- Imp ports DSPy, the framework for programming language models, to the BEAM as a full port
- Imp is published as a package on hex.pm for the Elixir/Erlang ecosystem
- The port was shared on Hacker News in a link post pointing to the hex.pm package and dspy.ai docs
Related
Social·Community
Serj Smorodinsky and Brett Kennedy run Oct 3 hands-on workshop on DSPy optimization over manual prompt tweaking
Serj Smorodinsky and Brett Kennedy, co-authors of a book on LLM applications, are running a 3-hour hands-on workshop on October 3 covering systematic prompt optimization with DSPy. The session frames prompting as engineering: participants build a baseline classifier, construct an eval dataset with task-specific metrics, then apply DSPy's few-shot and instruction-level optimization rather than hand-tuning prompts.
Serj Smorodinsky and Brett Kennedy, co-authors of a book on LLM applications, are running a 3-hour hands-on workshop on October 3 covering systematic prompt optimization with DSPy. The session frames prompting as engineering: participants build a baseline classifier, construct an eval dataset with task-specific metrics, then apply DSPy's few-shot and instruction-level optimization rather than hand-tuning prompts.
7 engagement1 sourceredditMon, Sep 28, 2026, 06:58 PM
Releases·Tool Release
DSPy 3.5 released, DSPy 4.0 previewed: separating task signatures from model implementations
Maxime Rivest presented DSPy 3.5 and previewed DSPy 4.0, emphasizing the concept of separating task definitions (Signatures) from model implementations. This approach allows AI engineers to define tasks purely by inputs and outputs, then flexibly experiment with models, settings, and templates without altering the workflow. The talk highlights DSPy's evolution toward higher-level AI engineering abstractions.
Maxime Rivest presented DSPy 3.5 and previewed DSPy 4.0, emphasizing the concept of separating task definitions (Signatures) from model implementations. This approach allows AI engineers to define tasks purely by inputs and outputs, then flexibly experiment with models, settings, and templates without altering the workflow. The talk highlights DSPy's evolution toward higher-level AI engineering abstractions.
Updated Thu, Jul 23, 2026, 05:45 PM — DSPy 4.0 previewed, building on the Signature abstraction introduced in 3.5.
63 engagement1 sourceyoutubeThu, Jul 23, 2026, 05:45 PM
Developments·Paper
ClusterFewshot: semantic clustering + utility scoring for few-shot LLM optimization, tested in DSPy pipelines
An arXiv paper introduces ClusterFewshot, a strategy combining semantic clustering with utility-aware scoring to select few-shot demonstrations for LLM workflows. Evaluated in DSPy-based pipelines, it reduces optimization cost while improving accuracy over bootstrap-based methods in prompt tuning and hybrid prompt-weight optimization.
An arXiv paper introduces ClusterFewshot, a strategy combining semantic clustering with utility-aware scoring to select few-shot demonstrations for LLM workflows. Evaluated in DSPy-based pipelines, it reduces optimization cost while improving accuracy over bootstrap-based methods in prompt tuning and hybrid prompt-weight optimization.
0 engagement1 sourcearxivTue, Sep 22, 2026, 09:45 AM
Products·building
DSPy-based RAG optimization framework with automated prompt tuning
A framework that uses DSPy to automate prompt tuning and few-shot selection for RAG pipelines, supporting multi-stage retrieval with query rewriting, sub-query decomposition, metadata extraction, and hybrid retrieval via Weaviate. It solves the problem of manually optimizing RAG prompts for developers building retrieval-augmented generation systems.
A framework that uses DSPy to automate prompt tuning and few-shot selection for RAG pipelines, supporting multi-stage retrieval with query rewriting, sub-query decomposition, metadata extraction, and hybrid retrieval via Weaviate. It solves the problem of manually optimizing RAG prompts for developers building retrieval-augmented generation systems.
1 engagement1 sourceredditFri, Jul 31, 2026, 05:53 AM
Social·Community
DSPy – Program, don't prompt, your LLMs
Hacker News discusses DSPy's 'program, don't prompt' approach to LLM development
Hacker News discusses DSPy's 'program, don't prompt' approach to LLM development
7 engagement1 sourcehackernewsSun, Sep 27, 2026, 07:38 PM