Skip to content

Changelog

v0.8.3 — Studio React style fix

Fixes

  • Studio Quick-start footer used HTML string style attrs → React error #62 crash
  • make lint gofmt ignores .slmcode/ workspace artifacts

v0.7.3 — Incomplete finalize recovery

Highlights

  • Detect empty finalize and synthetic model ended on a tool call blocked JSON
  • Up to two finish-steer corrector passes (demand status JSON, stop tool chains)
  • Provisional done from disk/tool evidence when finalize still fails
  • Knowledge cards: Python / Go project bars for language expectations

v0.7.2 — Escalate timeout → SLM arbitrator

Highlights

  • On escalate HITL timeout, dedicated @escalate agent decides retry / re_scope / abort / mark_done (override via escalate_timeout_agent)
  • Heuristic fallback when the LLM is unavailable (stubs → retry, vague → re_scope)
  • Docs: Studio / TUI / Agents / Architecture / Config cover escalate + runnable QA bar
  • CI: trailing-whitespace fix in docs/pipeline.md

v0.7.1 — Runnable quality bar + escalate HITL

Highlights

  • Greenfield Python QA defaults to pytest (fail closed), not compileall
  • Acceptance smoke — whitelisted commands from task acceptance run after each worker
  • Worker critique loops until smoke/static/acceptance green (bounded by max_retries)
  • Run success requires a strong QA gate — syntax-only cannot rubber-stamp incomplete boards
  • Escalate HITL — Studio modal + TUI /escalate; options: re-scope / retry / mark done / abort

API

  • GET / POST /api/escalate/pending|answer

v0.7.0 — Pipeline control + reference quality

Highlights

  • Config-driven pipeline (.slmcode/pipeline.yaml) — bind any agent to any phase, configure execute-loop reviewer/corrector, insert slots before/after/replace phases
  • Studio Pipeline tab — edit phases/slots live; progress header follows config dynamically
  • Reference-quality bar — project completeness gate blocks TestSLMs-style false success
  • Real-query eval suite — offline harness + optional live LangGraph / FastAPI / CLI cases
  • Loop feedback — SSE kind=loop with wave reasons + continue/abort/flag HITL
  • Placeholder polish — @placeholder pass + precise gap flagging
  • Custom agents accepted in specialist mode and board role dropdowns

API

  • GET / PUT /api/pipeline
  • POST /api/pipeline/reset
  • GET / POST /api/continue/pending|answer

Docs

  • New Pipeline guide; updates to Studio, Agents, Config, Architecture

v0.6.0 — SLM quality ports + eval

Harness gates, interventions, turn meter, slmcode eval, and Studio/TUI polish.