spp
  • Home
  • Why spp
  • Methodology
  • Installation
  • Usage
  • Phases
  • Examples
  • Scope
  • Changelog
  • Benchmark

On this page

  • Unreleased
  • Releases
  • Versioning posture

Changelog

Release history. spp follows semantic versioning; the methodology froze at v1.0.

The detailed, canonical changelog (Keep a Changelog format) lives in the repo: CHANGELOG.md. This is the summary.

Unreleased

  • Added the public-benchmark example — the first backed by a real, reproducible public-data run — plus links to the live benchmark.
  • Fixed the runner to support OpenAI reasoning models (gpt-5*, o-series).

Documentation and a worked artifact set only; the frozen v1.0 contract is untouched.

Releases

Version Date Summary
1.0.0 2026-06-08 Stabilization — the contract freeze. The methodology is frozen; new capability is now v2.0.
0.11.0 2026-06-08 Prompt decomposition — a managed linear pipeline of node-local tasks.
0.10.0 2026-06-08 Structured extraction as a designer-agent mode (span / NER).
0.9.0 2026-06-08 Prompt-structure advisor with batched I/O.
0.8.0 2026-06-04 Operational hardening — per-step loop resumption and the sacred-test-set guard hook.
0.7.0 2026-06-02 Judge-panel-assisted baseline labeling.
0.6.0 2026-06-02 Input preprocessing and multilingual data.
0.5.0 2026-05-31 Failure-driven prompting-technique suggestions.
0.4.0 2026-05-30 The K>1 multi-field runner.
0.3.0 2026-05-29 Finalize-layer statistics — bootstrap CIs and permutation tests.
0.2.0 2026-05-14 Bookkeeping generalization for broader output shapes (the schema layer).
0.1.0 2026-05-06 First tagged release — the core four-phase methodology.

Versioning posture

v1.0.0 froze the methodology. Near-term releases are patch-only (v1.0.x): bug fixes, documentation corrections, and advisor-catalog entries that conform to the frozen contract. New capability that changes the front-door surface waits for a dedicated v2.0 arc — see Scope & roadmap.

spp — Supervised Prompt Producer · MIT

 

Built with Quarto