Skip to content

CHANGELOG.md

Latest commit

 

History

History
147 lines (99 loc) · 6.07 KB

CHANGELOG.md

File metadata and controls

147 lines (99 loc) · 6.07 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

0.6.0 - 2026-06-16

Added

  • New skill /rq-problematization — generate impactful research questions via the Alvesson & Sandberg (2011) problematization framework: surface taken-for-granted assumptions across five types and challenge them with a devil's-advocate critic subagent (assumption-challenging over gap-spotting)
    • Positioned upstream of /analysis-framing: rq-problematization (theory-driven RQ) → analysis-framing (data grounding) → analysis-design. Does not chain directly to analysis-design — framing is the data-grounding gate
    • Re-generation path from /analysis-reflection: when a hypothesis is rejected/inconclusive, question the underlying assumption instead of only seeking a new angle
    • Bundled evals/evals.json — the repository's first eval suite, focused on gap-spotting regression and citation-fabrication checks

0.5.1 - 2026-04-21

Fixed

  • Remove $schema field from .claude-plugin/marketplace.json to restore plugin skill loading (#123)
    • Claude Code's marketplace validator rejected $schema as an Unrecognized key, causing the runtime to treat the plugin as disabled even when enabled in settings
    • This blocked all insight-blueprint skills (analysis-design, analysis-framing, catalog-register, premortem, etc.) from loading into the available skills list

0.5.0 - 2026-04-21

Added

  • New skill /premortem — pre-flight risk evaluation with approval token issuance (#117)
  • spec-workflow MCP server registered in .mcp.json (#118)

Changed

  • Batch analysis harness hardened for overnight stability (#117)
    • Flat manifest status contract with completed state (was nested execution.status)
    • 2-stage design hash verification for safe resume after crash
    • Launcher-owned finalize_run for consistent lifecycle termination
    • Hash mismatch detection in claude step + stub env toggles
    • User-facing documentation for risk levels, .insight/config.example.yaml, crash recovery

Fixed

  • uv.lock self-reference version out of sync with pyproject.toml (#119)

Chore

  • Ignore per-clone runtime stubs under .insight/ (#120)

[0.4.2] - [0.4.4]

Version bumps and maintenance releases. See git history for details: git log v0.4.1..v0.4.4 --oneline.

0.4.1 - 2026-04-06

Added

  • /batch-analysis skill for overnight batch execution of queued analysis designs
  • batch-prompt.md headless orchestration prompt with 8-cell notebook contract
  • next_action field convention for queue management (batch_execute type)
  • Claude Code Plugin distribution format (.claude-plugin/plugin.json, .mcp.json)
  • Plugin validation CI job (plugin-validate)
  • data-lineage Python package prerequisites check
  • Migration guide in README
  • Custom marketplace for plugin distribution

Changed

  • Skills moved from src/insight_blueprint/_skills/ to skills/ (repository root) for Plugin auto-discovery
  • Rules (_rules/) integrated into SKILL.md definitions
  • Extension policy and optional package note added to CLAUDE.md managed section

Removed

  • upgrade-templates CLI subcommand
  • Skill/rule copy logic from storage/project.py (~400 lines)

0.4.0 - 2026-03-20

Added

  • --mode server for team deployment: serves WebUI + MCP SSE on the same HTTP port
  • --mode headless for lightweight deployment: MCP SSE only (no WebUI)
  • --host and --port options for server/headless mode (default: 0.0.0.0:4000)
  • --no-browser flag to suppress browser auto-open in full mode
  • Concurrent write safety with threading.Lock on write_yaml

Deprecated

  • --headless flag: use --no-browser instead (flag still works with deprecation warning)

0.3.0 - 2026-03-17

Added

  • get_review_comments MCP tool for reading review batches (closes read/write asymmetry in review workflow)
  • /analysis-revision skill for structured revision of review comments with per-comment tracking

Changed

  • Added methodology and analysis_intent as reviewable sections in WebUI inline comments

0.2.0 - 2026-03-11

Added

  • Analysis Framing skill for structuring analysis scope and approach
  • Unified 6-skill chaining: analysis-design → analysis-framing → analysis-journal → analysis-reflection → catalog-register → data-lineage
  • --version flag for CLI version display

Changed

  • Upgraded shadcn from 4.0.0 to 4.0.2
  • Restructured README to prioritize user experience

CI

  • Added Dependabot auto-merge for patch updates

0.1.0 - 2026-03-10

Added

  • Analysis Design management with hypothesis-driven workflow
  • Data Catalog for domain knowledge and caution registration
  • Review workflow with status transitions (draft → reviewing → approved/revised)
  • Domain Knowledge suggestion for analysis designs (4 matching strategies)
  • Data Lineage tracking for data transformations
  • WebUI Dashboard with 2-tab navigation (Designs / Catalog)
  • Bundled Skills: analysis-design, analysis-journal, analysis-reflection, catalog-register, data-lineage
  • Typed verification models: ExplanatoryVariable, Metric, ChartSpec, Methodology
  • YAML direct edit resilience (extra field preservation + corrupt file isolation)
  • SQLite FTS5 full-text search index