smd-engine 0.6.0__tar.gz → 0.6.2__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (177) hide show
  1. smd_engine-0.6.2/CHANGELOG.md +67 -0
  2. {smd_engine-0.6.0 → smd_engine-0.6.2}/PKG-INFO +2 -2
  3. {smd_engine-0.6.0 → smd_engine-0.6.2}/README.md +1 -1
  4. {smd_engine-0.6.0 → smd_engine-0.6.2}/README_ru.md +1 -1
  5. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/bakeries-c1.board.yaml +10 -0
  6. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/bakeries-c2.board.yaml +22 -1
  7. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/golden/README.md +2 -0
  8. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/two-boards-c1.board.yaml +5 -1
  9. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example_ru/pekarni-c1.board.yaml +10 -0
  10. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example_ru/pekarni-c2.board.yaml +22 -1
  11. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/README.md +1 -0
  12. smd_engine-0.6.0/CHANGELOG.md → smd_engine-0.6.2/docs/changelog-0.2-0.5.md +3 -37
  13. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/checks.md +8 -3
  14. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/commands.md +18 -17
  15. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/data-model.md +8 -6
  16. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/first-board.md +1 -1
  17. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/roles-and-actors.md +4 -0
  18. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/profiles.md +2 -2
  19. smd_engine-0.6.2/docs/red-lines.md +54 -0
  20. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/stages.md +4 -0
  21. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/field-result.md +8 -2
  22. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/functional-plan.md +5 -3
  23. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/positional-construction.md +2 -2
  24. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/small-situation.md +3 -1
  25. {smd_engine-0.6.0 → smd_engine-0.6.2}/profiles/classic-two-boards.yaml +6 -0
  26. {smd_engine-0.6.0 → smd_engine-0.6.2}/profiles/management-four-boards.yaml +8 -1
  27. {smd_engine-0.6.0 → smd_engine-0.6.2}/profiles/program-three-boards.yaml +6 -0
  28. {smd_engine-0.6.0 → smd_engine-0.6.2}/pyproject.toml +1 -1
  29. smd_engine-0.6.2/src/smd_engine/nextstep.py +341 -0
  30. smd_engine-0.6.2/src/smd_engine/redline.py +180 -0
  31. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/index.py +10 -3
  32. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/markdown.py +2 -0
  33. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/sessionflow.py +27 -1
  34. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/text.py +159 -3
  35. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/scaffold.py +10 -4
  36. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/schemas/board_patch.schema.json +1 -1
  37. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/schemas/board_set.schema.json +34 -4
  38. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/status.py +44 -4
  39. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/__init__.py +14 -0
  40. smd_engine-0.6.2/src/smd_engine/validate/ai.py +95 -0
  41. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/context.py +7 -0
  42. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/graph.py +48 -12
  43. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/hypotheses.py +14 -3
  44. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/management.py +9 -6
  45. smd_engine-0.6.2/src/smd_engine/validate/redlines.py +215 -0
  46. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/structure.py +31 -0
  47. smd_engine-0.6.2/tests/test_ai_roles.py +312 -0
  48. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_apply.py +21 -1
  49. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_beer_layer.py +42 -3
  50. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_compare.py +1 -1
  51. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_export.py +43 -0
  52. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_hypothesis_standard.py +22 -0
  53. smd_engine-0.6.2/tests/test_nextstep.py +200 -0
  54. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_procedure_checks.py +118 -4
  55. smd_engine-0.6.2/tests/test_red_lines.py +463 -0
  56. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_render_and_cli.py +2 -0
  57. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_scaffold.py +6 -1
  58. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_status.py +1 -1
  59. {smd_engine-0.6.0 → smd_engine-0.6.2}/uv.lock +1 -1
  60. smd_engine-0.6.0/src/smd_engine/nextstep.py +0 -277
  61. smd_engine-0.6.0/tests/test_nextstep.py +0 -108
  62. {smd_engine-0.6.0 → smd_engine-0.6.2}/.claude/agents/smd-game-technician.md +0 -0
  63. {smd_engine-0.6.0 → smd_engine-0.6.2}/.claude/agents/smd-methodologist.md +0 -0
  64. {smd_engine-0.6.0 → smd_engine-0.6.2}/.claude/agents/smd-visionary.md +0 -0
  65. {smd_engine-0.6.0 → smd_engine-0.6.2}/.claude/commands/procedure.md +0 -0
  66. {smd_engine-0.6.0 → smd_engine-0.6.2}/.flake8 +0 -0
  67. {smd_engine-0.6.0 → smd_engine-0.6.2}/.github/workflows/ci.yml +0 -0
  68. {smd_engine-0.6.0 → smd_engine-0.6.2}/.github/workflows/release.yml +0 -0
  69. {smd_engine-0.6.0 → smd_engine-0.6.2}/.gitignore +0 -0
  70. {smd_engine-0.6.0 → smd_engine-0.6.2}/.python-version +0 -0
  71. {smd_engine-0.6.0 → smd_engine-0.6.2}/CONTRIBUTING.md +0 -0
  72. {smd_engine-0.6.0 → smd_engine-0.6.2}/LICENSE +0 -0
  73. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/golden/bakeries-c2-expected.board.yaml +0 -0
  74. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/golden/s3_transcript.txt +0 -0
  75. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/patches/bakeries-c2-1.patch.yaml +0 -0
  76. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example/sources.yaml +0 -0
  77. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example_ru/pekarni-oru-c1.board.yaml +0 -0
  78. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example_ru/pekarnya-holding-c1.board.yaml +0 -0
  79. {smd_engine-0.6.0 → smd_engine-0.6.2}/cases/example_ru/sources.yaml +0 -0
  80. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/glossary.md +0 -0
  81. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/README.md +0 -0
  82. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/encodable.md +0 -0
  83. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/odi-problematization.md +0 -0
  84. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/reconciliation.md +0 -0
  85. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/smd-classic.md +0 -0
  86. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/sources.md +0 -0
  87. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/structural-moves.md +0 -0
  88. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/method/triz-systems.md +0 -0
  89. {smd_engine-0.6.0 → smd_engine-0.6.2}/docs/plan.md +0 -0
  90. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/README.md +0 -0
  91. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/beer-viable-system.md +0 -0
  92. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/design-chain.md +0 -0
  93. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/fliamer-diamond.md +0 -0
  94. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/khaprov-levels.md +0 -0
  95. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/khaprov-ontology.md +0 -0
  96. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/scamper-check.md +0 -0
  97. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/schukin-boundaries.md +0 -0
  98. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/schukin-interface-tax.md +0 -0
  99. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/three-positions.md +0 -0
  100. {smd_engine-0.6.0 → smd_engine-0.6.2}/lenses/triz-contradictions.md +0 -0
  101. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/README.md +0 -0
  102. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/ingest-transcript.md +0 -0
  103. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/problematization.md +0 -0
  104. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/qualification-pass.md +0 -0
  105. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/reflexive-protocol.md +0 -0
  106. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/solo-session.md +0 -0
  107. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/stakeholder-ontology.md +0 -0
  108. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/theses-to-board.md +0 -0
  109. {smd_engine-0.6.0 → smd_engine-0.6.2}/procedures/visionary-tract.md +0 -0
  110. {smd_engine-0.6.0 → smd_engine-0.6.2}/roles/README.md +0 -0
  111. {smd_engine-0.6.0 → smd_engine-0.6.2}/roles/actors.md +0 -0
  112. {smd_engine-0.6.0 → smd_engine-0.6.2}/roles/game-technician.md +0 -0
  113. {smd_engine-0.6.0 → smd_engine-0.6.2}/roles/methodologist.md +0 -0
  114. {smd_engine-0.6.0 → smd_engine-0.6.2}/roles/researcher.md +0 -0
  115. {smd_engine-0.6.0 → smd_engine-0.6.2}/roles/visionary.md +0 -0
  116. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/__init__.py +0 -0
  117. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/agree.py +0 -0
  118. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/apply.py +0 -0
  119. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/cli.py +0 -0
  120. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/compare.py +0 -0
  121. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/eval.py +0 -0
  122. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/leverage.py +0 -0
  123. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/migrate.py +0 -0
  124. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/pack.py +0 -0
  125. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/profile.py +0 -0
  126. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/records.py +0 -0
  127. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/__init__.py +0 -0
  128. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/assets/frames.css +0 -0
  129. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/assets/frames.js +0 -0
  130. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/assets/graph.css +0 -0
  131. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/assets/graph.js +0 -0
  132. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/frames.py +0 -0
  133. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/frames_page.py +0 -0
  134. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/frames_text.py +0 -0
  135. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/graph.py +0 -0
  136. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/graph_ui.py +0 -0
  137. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/html.py +0 -0
  138. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/layout.py +0 -0
  139. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/light.py +0 -0
  140. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/mermaid.py +0 -0
  141. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/svg.py +0 -0
  142. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/render/vault.py +0 -0
  143. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/revise.py +0 -0
  144. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/schemas/__init__.py +0 -0
  145. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/schemas/source_manifest.schema.json +0 -0
  146. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/schemas/taxonomies.schema.json +0 -0
  147. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/screen.py +0 -0
  148. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/sources.py +0 -0
  149. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/hints.py +0 -0
  150. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/manifest.py +0 -0
  151. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/ontology.py +0 -0
  152. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/signoffs.py +0 -0
  153. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/taxonomies.py +0 -0
  154. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/theses.py +0 -0
  155. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/validate/tract.py +0 -0
  156. {smd_engine-0.6.0 → smd_engine-0.6.2}/src/smd_engine/verdict.py +0 -0
  157. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/conftest.py +0 -0
  158. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_action_program.py +0 -0
  159. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_agree.py +0 -0
  160. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_beer_render.py +0 -0
  161. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_eval.py +0 -0
  162. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_frames.py +0 -0
  163. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_graph.py +0 -0
  164. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_ingestion.py +0 -0
  165. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_lenses.py +0 -0
  166. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_management.py +0 -0
  167. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_migrate.py +0 -0
  168. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_ontology_continuum.py +0 -0
  169. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_procedures.py +0 -0
  170. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_profile.py +0 -0
  171. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_publish.py +0 -0
  172. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_record_boards.py +0 -0
  173. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_render_fields.py +0 -0
  174. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_render_ru_pages.py +0 -0
  175. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_render_ru_values.py +0 -0
  176. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_revise.py +0 -0
  177. {smd_engine-0.6.0 → smd_engine-0.6.2}/tests/test_validate.py +0 -0
@@ -0,0 +1,67 @@
1
+ # Changelog
2
+
3
+ Dates are those of the tag. All tags v0.2.1 to v0.3.5 were made on 2026-10-29. The version in `pyproject.toml` stayed at 0.2.0 through v0.2.4 and at 0.3.0 from v0.3.0 to v0.3.5; 0.3.6 is the first release whose package version equals its tag.
4
+
5
+ ## Unreleased
6
+
7
+ ## 0.6.2 - 2026-10-31
8
+
9
+ Gate ОТП1 В (the case owner, 31.10): a board carries the parts of a split hypothesis, as the group's hypothesis matrix does. Schema 0.3 stays; old records stay valid.
10
+
11
+ - A hypothesis split into parts keeps its number and adds one lowercase Latin letter: `H2a`, `H2b` (gate ОТП1 В, the case owner 31.10: the group's hypothesis matrix splits composite hypotheses into parts, each with one type). The board schema accepts `^H[0-9]+[a-z]?$` for a hypothesis `id` and an ontology record's `test`, and the patch schema accepts it for an item id, so a patch can remove `H2` and add `H2a` and `H2b`. Old records stay valid. `docs/data-model.md` gives the rule.
12
+ - The changelog of 0.2.0-0.5.9 moved to `docs/changelog-0.2-0.5.md` (this file had passed 300 lines).
13
+
14
+ ## 0.6.1 - 2026-10-31
15
+
16
+ Follow-ups of 0.6.0 and two gates of 31.10: AI on roles and functions (ИИ1 Б) and structured red lines with observed values (КЛ1 В, with dates as decided 31.10). Schema 0.3 stays; every new field is optional and old records stay valid.
17
+
18
+ Follow-ups of the 0.6.0 reviews. The schema gains one optional field, `carries` on a `flow` (additive: old records stay valid), and `function-flows` reads it; a record that sets none reports as before. Gate ИИ1 Б adds two optional element fields, `ai` and `red_line`, and gate КЛ1 В two more, `line` and `observed` (additive, schema stays 0.3).
19
+
20
+ - Red lines with a structured value (gate КЛ1 В, the case owner 31.10: «should the red_line threshold be a period (min;max)?», answered В; dates instead of a term in years). An element with `red_line: true`, or an `algedonic_channel`, takes `line`: `metric`, `unit`, a corridor (`min` and/or `max`, one-sided allowed) or a yes/no (`expect`), `since` (optional; the record's `date` when absent), `until` (required) and `recovery_days`. Errors: both a corridor and `expect`, or neither; `min` not below `max`; `until` not after the start; `recovery_days` and the channel's `recovery_time` both set and different; `line` on an element that is no red line. The free-text `threshold` stays as a fallback; `red-line-without-value` accepts either. Observations sit on the same element, `observed: [{value, date, source}]`: a value of the wrong type for the line is an error, a source not of kind `field_artefact` a warning (with a manifest). New `src/smd_engine/redline.py` holds the deterministic breach rule (observations from the line's start to `until`, sorted by date; a run of outside values spans to the next inside value, or, still open, to the later of the last observation and the record's `date`, never past `until`; a breach when longer than `recovery_days`, else `recovery_time`, any run when neither is set). Warnings: observations before the line takes force, and two on one date. The schema requires `metric` with `expect`. Numbers on the pages are plain digits (in Russian, a thin space between thousands and a decimal comma). New hints, all profiles: `red-line-breach` and `red-line-expired` (observations or the record dated after `until`; → field-result.md §5) and `red-line-outlives-pace` (the line's span in years longer than the `pace` of an element it is linked to, the analogue of `ontology-outlives-pace`; the case owner 26.10, H24; → functional-plan.md §4). New `validate/redlines.py`; the linking used by `ai-decision-unanswered` moved there and now reads `in` and `performed_by` both ways. The management hint `line-without-recovery` reads `line` too: a channel has a value when it has `threshold` or `line`, and a recovery time when it has `recovery_time` or `line.recovery_days`. `smd status` lists the red lines with a breach. The page, the digest, the index, the graph details and the frames details show the line in words («доля ФОТ от 18 до 25 %, с 01.01.2027 по 31.12.2031, восстановление 90 дней») and the last observation («27 % на 01.03.2027, вне коридора»). `bakeries-c2` and `pekarni-c2` carry the owner's red line `rl-unsold` (unsold dough at most 5 % of the night's volume, 01.03.2026 to 28.02.2027, recovery 14 days), with no observations; `--strict` stays 0/0. New `docs/red-lines.md` (the fields, the rule, the words), linked from `docs/data-model.md` and `docs/README.md`; `field-result` 0.3.2 gains section 5 (red lines from observations); `functional-plan` section 4, `docs/checks.md`, `docs/commands.md`.
21
+
22
+ - AI on the boards (gate ИИ1 Б, the case owner 31.10: "should flows include 'intelligence'? when AI is used", answered Б). AI is not a new `carries` value: an AI agent is an actor in `fillers`, and what it produces moves as information or product. Roles (kinds a profile flags `role`; a gap only with `of: position`) and functions take an optional `ai`: `assists` («ИИ помогает») or `performs` («выполняет ИИ»); anywhere else it is an error. A function's effective `ai` is its own, else that of the role in its `performed_by`. Elements take an optional `red_line: true` (agreed by the case owner, 31.10; refused on kinds flagged `light_off`: notes, and facts and signals on the default profile); its value, a threshold or yes/no, goes in `threshold`, now allowed on any red-line element, and the new hint `red-line-without-value` (all profiles, → functional-plan.md §4) names one without. An `algedonic_channel` with a `threshold` is a red line without it. `escalates` gains a second use: from a function to the role that answers for it at a red line. New hint `ai-decision-unanswered` (all three profiles, → functional-plan.md §3; new `validate/ai.py`): a function AI performs, linked to a red line (itself one, `in` one, performed by a red-line role, or joined to one by a relation of any kind), for which no role without `ai: performs` answers by `performed_by`, `subordinates` (role to function), `escalates` or a `flow` with `carries: information` (function to role); the case behind it is the MIPT case's H64. `smd status` (counts) and `smd index` (top of the page) print «функции: выполняет ИИ n, ИИ помогает k, из m» when any element has `ai`, nothing otherwise. The page, the digest, the index, the graph details and the frames details show `ai` and `red_line` in words (`LABELS`). `fn-dough` in `bakeries-c2` and `pekarni-c2` has `ai: assists` (the night's volume planned with an AI demand forecast; the organiser decides); `--strict` stays 0/0. The MIPT case has no `ai` and gets no new finding (0 errors, 150 warnings). `functional-plan` 0.2.7 (section 3) and `positional-construction` 0.2.8 (section 2) say where `ai` goes; `docs/method/roles-and-actors.md` gains "AI as an actor"; `docs/data-model.md`, `docs/checks.md`, `docs/commands.md`.
23
+
24
+ - Example cases draw a clean functional plan. The hint `function-flows` fired on every function of `f-new` in `bakeries-c1`, `bakeries-c2`, `two-boards-c1`, `pekarni-c1`, `pekarni-c2` and the golden record `bakeries-c2-expected` (found by running `smd validate --hints` on every record under `cases/`; the two management-profile records do not enable it). `bakeries-c2` and `pekarni-c2` now draw the `flow` relations the story states (mills supply flour into the dough, the owner pays per tonne and receives the dough, dough goes to logistics, the owner pays inside the dough price for now and receives it before four in the morning); `two-boards-c1` draws flour, the dough and the delivery to the owner and links the open `gap-payer` to both functions. `bakeries-c1` and `pekarni-c1` are cycle 1, where K2 says nobody yet performs or pays: they draw the flour and the chain dough, quality standard, logistics, and leave the payer and the receiver as gaps (`gap-payer`, and a new `gap-receiver`, `of: position`) linked to each function by `refers`. `function-flows` and `unlinked-functions` are silent on all five; the hints left in cycle 1 (`performed_by`, `idle-positions`) are K2 itself. The golden record is unchanged, and its README says why. `tests/test_procedure_checks.py` checks the five records.
25
+ - Relations of kind `flow` take an optional `carries` (`money`, `material`, `product`, `information`; schema additive, old records stay valid; words in both languages in `LABELS`, shown in the Markdown digest). `function-flows` reads a payer as an incoming flow of `money` from a position or place, an input as an incoming flow of anything but money, and a receiver as an outgoing flow of anything but money to a position or place, so a supplier (mills with flour) is no longer a payer. A receiver is also a non-money flow to a later function of the same frame that has one (followed along the chain, cycles walked once), so a middle function needs no direct line; on the MIPT case's cycle 5 the function-flow hints go from 10 to 7. A flow with no `carries` is read as before for payer and input; the hint's wording names `carries: money`. All flows of the example cases set `carries`. `functional-plan` 0.2.6 asks for it (section 4). `docs/checks.md`, `docs/data-model.md`.
26
+ - `bakeries-c2` and `pekarni-c2` draw the owner's and the organiser's `flow` to and from `coop-plant` (what each gives or takes), so the `cooperation-flows` hint is silent on the examples (the plant pays the organiser: `carries: money`, "paid by use against the pledged volume"). The direct `fn-dough` to owner lines are dropped from `bakeries-c2`, `pekarni-c2` and `two-boards-c1`: the dough reaches the owner through logistics.
27
+ - An `untested` hypothesis with `observed` gets no verdict proposal, in `smd validate`, `smd status` or the SessionFlow export; the hint `verdict` says "observed data recorded but status is untested: set testing". Only a `testing` one is read by its rule.
28
+ - `smd next` reads the hint pointers. Every hint finding contributes its procedure and section (`HINT_PROCEDURES`) as a candidate with the hint's message as the reason; a hint on a document (`board-dense`) is printed as a note. The conditions that no hint reports stay rules (`RULES`: gate 0, unplaced theses when no hint covers them, no K, a first cycle with no theses, a hypothesis in testing, closing the cycle when no hint of steps 1 to 6 is open). All candidates are sorted by one documented order (`ORDER`: gate 0; then the step-2 procedures `ingest-transcript`, `theses-to-board`, `field-result`, which lay information on the board; then the steps of `solo-session` section 3; then `reflexive-protocol` and `solo-session`). The closing rule waits for a hint at any procedure before `reflexive-protocol`, and the unplaced-theses rule is silenced only by a hint at `theses-to-board` section 2. A first cycle with no theses therefore gets `ingest-transcript` before `small-situation`, each procedure once with the sections of all its candidates; the output prints them (`functional-plan §3, §4`). The rules that re-derived hints (empty halves, incomplete positional set, new frame, gaps without a scheme block, closing by hint) are gone; the hints say the same, with the right section. `docs/commands.md` gives the table and the order.
29
+ - The SessionFlow export shows the hypothesis fields of 0.6.0 under each hypothesis of section 1, in Russian: `type`, `rule` (metric, operator, threshold, `n_min`, basis), `observed` (k/n and the artefact) and, for an untested or a testing hypothesis, the verdict the rule proposes (`verdict.py`; the group signs).
30
+ - `docs/commands.md`: `sources --refresh` writes `lines` only into an entry that already has a `lines:` line; its report prints the file's line count for every refreshed entry. The doc said the report carried a count only where `lines` is present; the behaviour (a count on every line of the report, the manifest touched only where `lines` exists) is kept and now described.
31
+ - `docs/release.md`: the tag example is `v<version>`, not `v0.3.6`; the public commit is authored as the organisation's GitHub noreply identity with `git -c user.name=... -c user.email=... commit` and checked with `git log -1 --format='%an <%ae>'` before the push.
32
+ - `small-situation` 0.2.4 and `field-result` 0.3.1 gain the "Record the run" line, in the words of the other procedures.
33
+ - `voice` of a K, a hypothesis, a relation and a transition is free text, and `AI`, `ai` and `ИИ` already read as an AI voice everywhere (`is_ai`). `smd migrate` and the validator leave them as written: a warning would add one per item (the MIPT case's cycle 5 has 158 `voice: AI` values, 7 of them on hypotheses) for a spelling that changes no result. `docs/data-model.md` and the schema description say so.
34
+
35
+ ## 0.6.0 - 2026-10-31
36
+
37
+ Method review of 31.10 (gates М1 В, Ф В, У В): reproducible runs, deterministic checks for rules the procedures asked a model to check by hand, a typed standard for hypotheses with a pre-registered decision rule, and a shorter path for new users. Schema 0.3 stays; every new field is optional and old records stay valid.
38
+
39
+ - Hypothesis standard (method review 31.10, gates М1 В and Ф В, formal audit proposal 6). Three optional hypothesis fields, additive (schema stays 0.3, old records stay valid): `type` (`core`, `product`, `claim`, `estimate`, `probe`, `design`: how the hypothesis is read), `rule` (`{metric, op, threshold, n_min, basis}`: the pre-registered decision rule, a structured `revise_if`; `op` states the refutation condition) and `observed` (`{k, n, artefact}`). New `src/smd_engine/verdict.py`: the 95% Wilson score interval of k/n (Wilson 1927; reproduces 0.2553-0.3662 for 81/263, Newcombe 1998) and a proposed verdict for an untested or testing hypothesis when the threshold is a share strictly between 0 and 1 (`op` is `<`, `<=`, `>` or `>=`): too few (n below `n_min`), refuted, confirmed, or inconclusive (the interval straddles the threshold). The proposal is a hint (`verdict`) and a section of `smd status`; the sign-off stays human. New error: k greater than n; new warnings: a recorded result opposite to the proposal, and (with a manifest) an `observed.artefact` that is not of kind `field_artefact`; `validate_record` takes the manifest's kinds as `source_kinds`. New hints `hypothesis-type`, `hypothesis-rule`, `rule-basis`, `rule-n-min`, `type-status` (`docs/checks.md`), enabled on all three profiles; `revise-if` accepts a `rule` and leaves out claims and estimates (they get `hypothesis-rule`) and cores, designs and probes. The MIPT case has no typed hypothesis and gets no new finding. Labels of the fields and of the six types in both languages; the page, the digest, the index and the graph show them under the hypothesis. `field-result` 0.3.0 gains section 0 (the six types and how each is read; the decision rule and the interval) and reads the proposal in section 3; `small-situation` 0.2.3 (section 4) gives each new H a `type`, and a claim or estimate a `rule` with its `basis`.
40
+ - `smd agree <record> <record> [...] [--fields ...] [--json]` (the audit of 31.10, proposal 4, gate Ф В): how stable are the labels a procedure writes. Takes N runs of one procedure on one input; warns, and goes on, when `run.procedure`, `run.pack_sha256` or `run.inputs` differ or are missing. Theses are matched across runs by the address they cite (the matching of `smd eval`, generalised to N runs); elements of the first record are matched in the other runs as `smd compare` matches them. For `belt`, `kind`, `status`, `voice`, `of` and the move `op` of theses, and `kind`, `of` and `op` of elements, it prints Krippendorff's alpha (nominal, any number of runs, missing values allowed; written in the new `agree.py` from the coincidence matrix, no new dependency; `n/a` with the reason when only one value occurs), the share of cite addresses found by all, some or one run with the one-run-only addresses, and a reading by Krippendorff's conventional cut-offs (0.800, 0.667), stated as a convention; a field with fewer than five pairable units shows its alpha with "few units" and is left out of the reading line. The alpha code reproduces the 0.743 of Krippendorff's 2011 worked example. `eval` is unchanged. `docs/commands.md` and the golden README describe it.
41
+ - `smd new <folder> [--profile] [--lang] [--case] [--title]` (UX audit of 31.10, gate У В): scaffolds a case folder that validates at once, `smd validate --strict` giving 0 errors and 0 warnings on every profile: `<case>-c1.board.yaml` (required fields, `profile`, an entry where the profile has the kind, one position to rename, empty lists), a `sources.yaml` skeleton with one entry, and a README naming the first procedure (`small-situation` on the default profile, `positional-construction` on the others). It refuses, writing nothing, if a file exists. `docs/first-board.md` starts from it.
42
+ - `smd status <record> [--lang] [-o file.md]`: the core view of a board set on one page of Markdown (46 lines for the MIPT case's cycle 5): state line, K not closed, hypotheses in testing and untested ones in leverage order, AI-voiced items waiting for a sign-off, transitions, counts per board and kind; lists are cut with "and N more". New `LABELS[...]["core"]` words in both languages.
43
+ - `smd next <record>`: the procedure to run next with one line why and at most two alternatives, by the rule table `RULES` in `src/smd_engine/nextstep.py` (written out in `docs/commands.md`); procedures of another profile are skipped; "no rule fires" is an answer.
44
+ - Hypothesis leverage (formal audit, section 4 row 5): for each untested hypothesis the size of what a refutation would reach (targets, K, ontology records testing it, theses on its targets, elements one relation away); a proposal, the group chooses. Listed by `smd status`, and a column "reach (proposal)" in the hypotheses table of `smd index`.
45
+ - Run stamp v2 (the audit of 31.10, formal section 3): the record's `run` takes four optional fields, `pack_sha256` (sha256 of the `smd pack` output the run used), `inputs` (source id to the sha256 of the input file the run read), `temperature` (a number) and `sample` (the run index, when a procedure is run several times on one input). Additive: the schema stays 0.3 and old records stay valid. The schema checks the types (a 64-character lowercase hex for a hash, an integer from 1 for `sample`); `docs/data-model.md`, `run`, describes the fields.
46
+ - `smd pack` prints `pack sha256: <hex>` to stderr, the hash of the bytes it wrote (the file with `-o`, the text on stdout otherwise), so that a run can record it as `run.pack_sha256`.
47
+ - `smd revise` no longer copies the previous cycle's `run` into the next cycle: the new record is not the product of that run, and a carried `pack_sha256` would name a pack it was not made from. The next procedure writes its own stamp.
48
+ - `smd eval` prints the whole `run` stamp of both records, `inputs` as `{id: sha256, ...}`; until now a mapping showed as a Python dict.
49
+ - The "Record the run" line of `functional-plan` 0.2.5, `ingest-transcript` 0.3.1, `positional-construction` 0.2.7, `problematization` 0.2.3, `qualification-pass` 0.4.1, `reflexive-protocol` 0.3.2, `solo-session` 0.2.4, `stakeholder-ontology` 0.1.1, `theses-to-board` 0.3.6 and `visionary-tract` 0.2.3 names the new fields, in the same words: write what you know, omit the rest, invent nothing. `small-situation` and `field-result` have no such line and are unchanged.
50
+ - `docs/first-board.md`: a walkthrough from an empty folder to a first validated and drawn board (an invented family bakery handing over to a manager) and on to the first procedure. It lists the required top-level fields (`title` and `status` were not named in the README), which profile to choose when, which procedure comes first and next, and what to bring. Its commands were run as shown.
51
+ - `docs/glossary.md`: the terms a first-time user meets, one line each; linked from the README, `README_ru.md`, `docs/README.md` and the walkthrough.
52
+ - `README.md` and `README_ru.md`: the stale "Version 0.3.6" line and the `uvx` pin in the CI example now say 0.5.9 (schema 0.3); the two files link the walkthrough and the glossary near the top; the data-model table names the required top-level fields of a board set.
53
+ - Procedure `stakeholder-ontology` 0.1.0 (step 3, role игротехник, any profile): elicits the ontology of one stakeholder position, by default the owner, the tier-1 stakeholder (its stakeholder ontology is the Owner Ontology), in the position's own words, in a solo session with the holder or from an interview transcript; in proposal mode only from registered sources in which the holder speaks. Six parts with question forms: entities, definitions, absences, topology, time, causality, plus Khaprov's ontology test for `competes` and a question for the ontology-dependent records. Product: an ontology record with `valid_for` that one position, object and subject quoted with addresses, `tags` tier 1 for the owner (`{tier: [intent]}` in a case whose taxonomy names tier 1 so), and a `test` H (by default: does the shareable part work as a reading key); the shareable part (verbatim glossary of the position's terms as used in records, list of ontology-dependent records) as a separate file registered in the manifest and cited by the record; the private part (deliberate absences, causal beliefs, sources of the edge, values, judgement of people) kept off the board, the case repository and the registered log, the record's `note` saying only that it exists and who keeps it. The translation into the six-tier ontology is a separate step: an AI-voiced `concept` element whose `detail` table names, per term, the tier and what the translation loses. Typical errors include an ontology without a position, the analyst's words in the holder's mouth, the private part on the board, the owner set apart from the stakeholders, and a seventh tier. No schema or code change.
54
+ - `positional-construction` 0.2.6 (section 5) and `solo-session` 0.2.3 (section 3) point to it; `stakeholder-ontology` uses the full form "six-tier ontology of a business" at first use.
55
+ - `procedures/README.md`, `lenses/README.md` (selection table: no lens offered) and `docs/method/roles-and-actors.md` (where the role rule is applied) list the new procedure.
56
+
57
+ - Method review 31.10.2026 (gates Ф В and У В): rules the procedures asked the model to check by hand, which are exact graph or set conditions on the record, are now checks of the validator (`docs/checks.md`).
58
+ - New errors: a cycle of `belongs`, of `subordinates`, or of move sources (`move.from` of elements and theses). A cycle of `flow` is legitimate and is not checked.
59
+ - New warning: an ontology position of `valid_for` with no line in `subjects` (the reverse of the existing warning).
60
+ - New hints of the functional plan (functional-plan.md §4): `function-flows` (a function in a new frame with no payer, input or receiver `flow`, unless a gap with the matching `of` is linked to it), `unlinked-functions`, `idle-positions`.
61
+ - New hints of the typical errors of a group screen on `program-three-boards` (qualification-pass.md §3.2): `positions-unlinked` (E2), `object-isolated` (E3), `isolated-facts` (E6, E11), `jump-to-frame` (E12), `transition-unanchored` (E14). The hint `belts` also names theses that are all мД ("mechanical reproduction").
62
+ - New hints of the procedures' self-checks: `ai-cooperation` and `cooperation-flows` (positional-construction.md §5), `translation-loss` (stakeholder-ontology.md §5), `garland`, `bridge-from` and `move-via` (visionary-tract.md §1, §3, §5; `program-three-boards` only).
63
+ - New hint `board-dense`: a board with more elements than the profile's new optional key `board_dense` (default 15) gets advice to fold or split it.
64
+ - `smd validate --hints` prints after each hint the procedure and section that act on it, "→ procedures/functional-plan.md §4" (`Finding.procedure`, filled from `HINT_PROCEDURES` in `validate/context.py`; the field is not compared, so findings stay equal).
65
+ - The three profiles enable the new hints that their kinds allow. `functional-plan` 0.2.5, `qualification-pass` 0.4.1 and `visionary-tract` 0.2.3, with `positional-construction` and `stakeholder-ontology` in their unreleased versions, say which of their checks the validator now reports.
66
+
67
+ Older releases (0.2.0 to 0.5.9): [docs/changelog-0.2-0.5.md](docs/changelog-0.2-0.5.md).
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.5
2
2
  Name: smd-engine
3
- Version: 0.6.0
3
+ Version: 0.6.2
4
4
  Summary: Records, checks and renderings for SMD-style thinking sessions (three boards, cycles, qualifications)
5
5
  Project-URL: Homepage, https://github.com/viberesearch/smd-engine
6
6
  Project-URL: Repository, https://github.com/viberesearch/smd-engine
@@ -98,7 +98,7 @@ Examples with options are in `docs/commands.md`; `smd <command> --help` lists th
98
98
  | `apply` | Applies a patch (add, set and remove by id or by transition number `n`, `board_notes`, state) to a board set as its next sub-version, then validates |
99
99
  | `pack` | One Markdown file with a procedure, its roles, lenses, the data model, the cells to fill and the records, for any model |
100
100
  | `status` | The core view of a board set on one page: state line, open K, hypotheses (untested ones in leverage order, a proposal), AI items waiting for a sign-off, transitions, counts |
101
- | `next` | The procedure to run next, with one line why and at most two alternatives, by a rule table (`docs/commands.md`) |
101
+ | `next` | The procedure to run next, with one line why and at most two alternatives, from the hint pointers and a few explicit rules, in the solo-session order (`docs/commands.md`) |
102
102
 
103
103
  ## Data model on one screen
104
104
 
@@ -73,7 +73,7 @@ Examples with options are in `docs/commands.md`; `smd <command> --help` lists th
73
73
  | `apply` | Applies a patch (add, set and remove by id or by transition number `n`, `board_notes`, state) to a board set as its next sub-version, then validates |
74
74
  | `pack` | One Markdown file with a procedure, its roles, lenses, the data model, the cells to fill and the records, for any model |
75
75
  | `status` | The core view of a board set on one page: state line, open K, hypotheses (untested ones in leverage order, a proposal), AI items waiting for a sign-off, transitions, counts |
76
- | `next` | The procedure to run next, with one line why and at most two alternatives, by a rule table (`docs/commands.md`) |
76
+ | `next` | The procedure to run next, with one line why and at most two alternatives, from the hint pointers and a few explicit rules, in the solo-session order (`docs/commands.md`) |
77
77
 
78
78
  ## Data model on one screen
79
79
 
@@ -75,7 +75,7 @@ uv run smd frames cases/example_ru/pekarni-c2.board.yaml -o out/frames.html
75
75
  | `apply` | Применяет правку (add, set и remove по id или по номеру перехода `n`, `board_notes`, state) к записи как её следующую подверсию и проверяет результат |
76
76
  | `pack` | Один файл Markdown с процедурой, её ролями, линзами, моделью данных, ячейками для заполнения и записями, для любой модели |
77
77
  | `status` | Ядро записи на одной странице: строка состояния, открытые K, гипотезы (непроверенные по охвату, это предложение), элементы от ИИ, ждущие подписи, переходы, счёт |
78
- | `next` | Какую процедуру запускать дальше, с одной строкой «почему» и не более чем двумя запасными, по таблице правил (`docs/commands.md`) |
78
+ | `next` | Какую процедуру запускать дальше, с одной строкой «почему» и не более чем двумя запасными, по указателям подсказок валидатора и нескольким явным правилам, в порядке solo-session (`docs/commands.md`) |
79
79
 
80
80
  ## Модель данных на одном экране
81
81
 
@@ -43,8 +43,18 @@ elements:
43
43
  - {id: s-big-signal, board: s, half: big, kind: signal,
44
44
  label: "chains move to central production", seen_from: p-organiser}
45
45
  - {id: gap-payer, board: od, kind: gap, label: "who pays for the shared plant?", of: position}
46
+ - {id: gap-receiver, board: od, kind: gap, label: "who receives the dough? no bakeries on the screen", of: position}
46
47
  relations:
47
48
  - {from: p-owner, to: p-organiser, kind: shift, label: "other position"}
49
+ - {from: place-mills, to: fn-dough, kind: flow, carries: material, label: "flour"}
50
+ - {from: fn-dough, to: fn-quality, kind: flow, carries: product, label: "dough to check"}
51
+ - {from: fn-quality, to: fn-logistics, kind: flow, carries: product, label: "dough cleared for delivery"}
52
+ - {from: gap-payer, to: fn-dough, kind: refers, label: "payer missing"}
53
+ - {from: gap-payer, to: fn-quality, kind: refers, label: "payer missing"}
54
+ - {from: gap-payer, to: fn-logistics, kind: refers, label: "payer missing"}
55
+ - {from: gap-receiver, to: fn-dough, kind: refers, label: "receiver missing"}
56
+ - {from: gap-receiver, to: fn-quality, kind: refers, label: "receiver missing"}
57
+ - {from: gap-receiver, to: fn-logistics, kind: refers, label: "receiver missing"}
48
58
  transitions:
49
59
  - {n: 1, from: p-owner, to: s-small, cites: ["s1#t=00:05:00"]}
50
60
  - {n: 2, from: s-small, to: s-big}
@@ -25,7 +25,8 @@ elements:
25
25
  label: "new frame: shared production for independent bakeries", seen_from: p-organiser,
26
26
  cites: ["s1#t=00:31:00"]}
27
27
  - {id: fn-dough, board: oo, kind: function, label: "dough production, paid per tonne by members",
28
- in: [f-new], seen_from: p-organiser, performed_by: p-organiser, cites: ["s3#t=00:14:20"]}
28
+ in: [f-new], seen_from: p-organiser, performed_by: p-organiser, cites: ["s3#t=00:14:20"],
29
+ ai: assists, label_note: "the night's volume is planned from members' orders with an AI demand forecast; the organiser decides"}
29
30
  - {id: fn-logistics, board: oo, kind: function, label: "night logistics, run by the plant",
30
31
  in: [f-new], seen_from: p-organiser, performed_by: p-organiser}
31
32
  - id: m-coop
@@ -42,9 +43,29 @@ elements:
42
43
  - {id: s-big, board: s, half: big, kind: situation, label: "independent bakeries close",
43
44
  seen_from: p-organiser}
44
45
  - {id: gap-risk, board: oo, kind: gap, label: "who carries unsold dough?", of: position}
46
+ # A red line the owner sets for the pilot season (gate КЛ1 В, docs/red-lines.md); no
47
+ # observations yet: the nightly reports start with the plant.
48
+ - id: rl-unsold
49
+ board: oo
50
+ kind: standard
51
+ label: "unsold dough stays at most 5% of the night's volume"
52
+ seen_from: p-owner
53
+ red_line: true
54
+ line: {metric: "unsold dough, share of the night's volume", unit: "%", max: 5,
55
+ since: "2026-03-01", until: "2027-02-28", recovery_days: 14}
45
56
  relations:
46
57
  - {from: p-owner, to: p-organiser, kind: shift, label: "other position"}
47
58
  - {from: p-owner, to: f-old, kind: represents}
59
+ - {from: place-mills, to: fn-dough, kind: flow, carries: material, label: "flour"}
60
+ - {from: p-owner, to: fn-dough, kind: flow, carries: money, label: "pays per tonne"}
61
+ - {from: rl-unsold, to: fn-dough, kind: refers, label: "the night's volume is planned inside the line"}
62
+ - {from: fn-dough, to: fn-logistics, kind: flow, carries: product, label: "dough to deliver"}
63
+ - {from: p-owner, to: fn-logistics, kind: flow, carries: money, label: "pays inside the dough price, for now"}
64
+ - {from: fn-logistics, to: p-owner, kind: flow, carries: product, label: "dough before four in the morning"}
65
+ - {from: p-owner, to: coop-plant, kind: flow, carries: information, label: "orders dough, pledges a volume"}
66
+ - {from: coop-plant, to: p-owner, kind: flow, carries: product, label: "dough at a lower cost than the night shift"}
67
+ - {from: p-organiser, to: coop-plant, kind: flow, carries: product, label: "dough and delivery"}
68
+ - {from: coop-plant, to: p-organiser, kind: flow, carries: money, label: "paid by use against the pledged volume"}
48
69
  transitions:
49
70
  - {n: 1, from: p-reflect, to: f-new, label: "what session 1 left without a payer"}
50
71
  - {n: 2, from: f-new, to: s-big}
@@ -9,6 +9,8 @@ An invented check for the pair `ingest-transcript` then `theses-to-board`. It te
9
9
 
10
10
  The expected record is a hand-checked reference, not the only right answer. `smd eval` shows how far a run is from it; read the unmatched lines before you call a run worse.
11
11
 
12
+ It stops where the two procedures stop. The functional plan is drawn by `functional-plan`, so `smd validate --hints` on this record still reports the hints `function-flows`, `unlinked-functions` and `idle-positions`, which `../bakeries-c2.board.yaml` clears with `flow` relations. They are left here on purpose: `smd eval` compares elements, theses, K and H, not relations, and a record that carries a plan the transcript does not state would not be what the two procedures produce.
13
+
12
14
  ## Run it
13
15
 
14
16
  1. Pack the two procedures with the transcript's manifest and the previous cycle:
@@ -36,7 +36,11 @@ elements:
36
36
  relations:
37
37
  - {from: p-owner, to: p-organiser, kind: shift, label: "other position"}
38
38
  - {from: p-reflexive, to: p-owner, kind: refers, label: "looks back at"}
39
- - {from: fn-dough, to: fn-logistics, kind: flow, label: "dough"}
39
+ - {from: fn-dough, to: fn-logistics, kind: flow, carries: product, label: "dough"}
40
+ - {from: place-mills, to: fn-dough, kind: flow, carries: material, label: "flour"}
41
+ - {from: fn-logistics, to: p-owner, kind: flow, carries: product, label: "dough before four in the morning"}
42
+ - {from: gap-payer, to: fn-dough, kind: refers, label: "payer missing"}
43
+ - {from: gap-payer, to: fn-logistics, kind: refers, label: "payer missing"}
40
44
  transitions:
41
45
  - {n: 1, from: p-organiser, to: f-new, label: "frame of a new object"}
42
46
  qualifications:
@@ -43,8 +43,18 @@ elements:
43
43
  - {id: s-big-signal, board: s, half: big, kind: signal,
44
44
  label: "сети переходят на централизованное производство", seen_from: p-organiser}
45
45
  - {id: gap-payer, board: od, kind: gap, label: "кто платит за общий цех?", of: position}
46
+ - {id: gap-receiver, board: od, kind: gap, label: "кто получает тесто? пекарен на экране нет", of: position}
46
47
  relations:
47
48
  - {from: p-owner, to: p-organiser, kind: shift, label: "другая позиция"}
49
+ - {from: place-mills, to: fn-dough, kind: flow, carries: material, label: "мука"}
50
+ - {from: fn-dough, to: fn-quality, kind: flow, carries: product, label: "тесто на проверку"}
51
+ - {from: fn-quality, to: fn-logistics, kind: flow, carries: product, label: "тесто, допущенное к доставке"}
52
+ - {from: gap-payer, to: fn-dough, kind: refers, label: "нет плательщика"}
53
+ - {from: gap-payer, to: fn-quality, kind: refers, label: "нет плательщика"}
54
+ - {from: gap-payer, to: fn-logistics, kind: refers, label: "нет плательщика"}
55
+ - {from: gap-receiver, to: fn-dough, kind: refers, label: "нет получателя"}
56
+ - {from: gap-receiver, to: fn-quality, kind: refers, label: "нет получателя"}
57
+ - {from: gap-receiver, to: fn-logistics, kind: refers, label: "нет получателя"}
48
58
  transitions:
49
59
  - {n: 1, from: p-owner, to: s-small, cites: ["s1#t=00:05:00"]}
50
60
  - {n: 2, from: s-small, to: s-big}
@@ -25,7 +25,8 @@ elements:
25
25
  label: "новая рамка: общее производство для независимых пекарен", seen_from: p-organiser,
26
26
  cites: ["s1#t=00:31:00"]}
27
27
  - {id: fn-dough, board: oo, kind: function, label: "производство теста, участники платят за тонну",
28
- in: [f-new], seen_from: p-organiser, performed_by: p-organiser, cites: ["s3#t=00:14:20"]}
28
+ in: [f-new], seen_from: p-organiser, performed_by: p-organiser, cites: ["s3#t=00:14:20"],
29
+ ai: assists, label_note: "объём на ночь считают по заказам участников с прогнозом спроса от ИИ; решает организатор"}
29
30
  - {id: fn-logistics, board: oo, kind: function, label: "ночная логистика силами цеха",
30
31
  in: [f-new], seen_from: p-organiser, performed_by: p-organiser}
31
32
  - id: m-coop
@@ -42,9 +43,29 @@ elements:
42
43
  - {id: s-big, board: s, half: big, kind: situation, label: "независимые пекарни закрываются",
43
44
  seen_from: p-organiser}
44
45
  - {id: gap-risk, board: oo, kind: gap, label: "кто несёт риск непроданного теста?", of: position}
46
+ # Красная линия владельца на пилотный сезон (ворота КЛ1 В, docs/red-lines.md); наблюдений
47
+ # пока нет: ночные отчёты начнутся вместе с цехом.
48
+ - id: rl-unsold
49
+ board: oo
50
+ kind: standard
51
+ label: "непроданное тесто — не больше 5 % объёма ночи"
52
+ seen_from: p-owner
53
+ red_line: true
54
+ line: {metric: "доля непроданного теста от объёма ночи", unit: "%", max: 5,
55
+ since: "2026-03-01", until: "2027-02-28", recovery_days: 14}
45
56
  relations:
46
57
  - {from: p-owner, to: p-organiser, kind: shift, label: "другая позиция"}
47
58
  - {from: p-owner, to: f-old, kind: represents}
59
+ - {from: place-mills, to: fn-dough, kind: flow, carries: material, label: "мука"}
60
+ - {from: p-owner, to: fn-dough, kind: flow, carries: money, label: "платит за тонну"}
61
+ - {from: rl-unsold, to: fn-dough, kind: refers, label: "объём ночи планируют в пределах линии"}
62
+ - {from: fn-dough, to: fn-logistics, kind: flow, carries: product, label: "тесто к доставке"}
63
+ - {from: p-owner, to: fn-logistics, kind: flow, carries: money, label: "платит в цене теста, пока так"}
64
+ - {from: fn-logistics, to: p-owner, kind: flow, carries: product, label: "тесто к четырём утра"}
65
+ - {from: p-owner, to: coop-plant, kind: flow, carries: information, label: "заказывает тесто, обещает объём"}
66
+ - {from: coop-plant, to: p-owner, kind: flow, carries: product, label: "тесто дешевле своей ночной смены"}
67
+ - {from: p-organiser, to: coop-plant, kind: flow, carries: product, label: "тесто и доставка"}
68
+ - {from: coop-plant, to: p-organiser, kind: flow, carries: money, label: "оплата по использованию против обещанного объёма"}
48
69
  transitions:
49
70
  - {n: 1, from: p-reflect, to: f-new, label: "что сессия 1 оставила без плательщика"}
50
71
  - {n: 2, from: f-new, to: s-big}
@@ -10,6 +10,7 @@ Start with `README.md` in the repository root. This folder holds the reference d
10
10
  | `glossary.md` | the terms a first-time user meets, one line each |
11
11
  | `data-model.md` | the record format: board sets, source manifests, taxonomies, patches, the `profile` key, migration |
12
12
  | `checks.md` | the checks the validator runs: errors, warnings and hints, and the profile and source of each hint |
13
+ | `red-lines.md` | red lines: `red_line`, the structured value `line`, `observed` values and the breach rule |
13
14
  | `profiles.md` | the profile format: boards and their regions, kinds, the screen layout, the ids every profile defines, the checks of a profile file |
14
15
  | `commands.md` | the commands of `smd` with examples, and the exit codes |
15
16
  | `plan.md` | the build plan approved 2026-10-25 and the architecture of the repository |
@@ -1,40 +1,6 @@
1
- # Changelog
2
-
3
- Dates are those of the tag. All tags v0.2.1 to v0.3.5 were made on 2026-10-29. The version in `pyproject.toml` stayed at 0.2.0 through v0.2.4 and at 0.3.0 from v0.3.0 to v0.3.5; 0.3.6 is the first release whose package version equals its tag.
4
-
5
- ## Unreleased
6
-
7
- ## 0.6.0 - 2026-10-31
8
-
9
- Method review of 31.10 (gates М1 В, Ф В, У В): reproducible runs, deterministic checks for rules the procedures asked a model to check by hand, a typed standard for hypotheses with a pre-registered decision rule, and a shorter path for new users. Schema 0.3 stays; every new field is optional and old records stay valid.
10
-
11
- - Hypothesis standard (method review 31.10, gates М1 В and Ф В, formal audit proposal 6). Three optional hypothesis fields, additive (schema stays 0.3, old records stay valid): `type` (`core`, `product`, `claim`, `estimate`, `probe`, `design`: how the hypothesis is read), `rule` (`{metric, op, threshold, n_min, basis}`: the pre-registered decision rule, a structured `revise_if`; `op` states the refutation condition) and `observed` (`{k, n, artefact}`). New `src/smd_engine/verdict.py`: the 95% Wilson score interval of k/n (Wilson 1927; reproduces 0.2553-0.3662 for 81/263, Newcombe 1998) and a proposed verdict for an untested or testing hypothesis when the threshold is a share strictly between 0 and 1 (`op` is `<`, `<=`, `>` or `>=`): too few (n below `n_min`), refuted, confirmed, or inconclusive (the interval straddles the threshold). The proposal is a hint (`verdict`) and a section of `smd status`; the sign-off stays human. New error: k greater than n; new warnings: a recorded result opposite to the proposal, and (with a manifest) an `observed.artefact` that is not of kind `field_artefact`; `validate_record` takes the manifest's kinds as `source_kinds`. New hints `hypothesis-type`, `hypothesis-rule`, `rule-basis`, `rule-n-min`, `type-status` (`docs/checks.md`), enabled on all three profiles; `revise-if` accepts a `rule` and leaves out claims and estimates (they get `hypothesis-rule`) and cores, designs and probes. The MIPT case has no typed hypothesis and gets no new finding. Labels of the fields and of the six types in both languages; the page, the digest, the index and the graph show them under the hypothesis. `field-result` 0.3.0 gains section 0 (the six types and how each is read; the decision rule and the interval) and reads the proposal in section 3; `small-situation` 0.2.3 (section 4) gives each new H a `type`, and a claim or estimate a `rule` with its `basis`.
12
- - `smd agree <record> <record> [...] [--fields ...] [--json]` (the audit of 31.10, proposal 4, gate Ф В): how stable are the labels a procedure writes. Takes N runs of one procedure on one input; warns, and goes on, when `run.procedure`, `run.pack_sha256` or `run.inputs` differ or are missing. Theses are matched across runs by the address they cite (the matching of `smd eval`, generalised to N runs); elements of the first record are matched in the other runs as `smd compare` matches them. For `belt`, `kind`, `status`, `voice`, `of` and the move `op` of theses, and `kind`, `of` and `op` of elements, it prints Krippendorff's alpha (nominal, any number of runs, missing values allowed; written in the new `agree.py` from the coincidence matrix, no new dependency; `n/a` with the reason when only one value occurs), the share of cite addresses found by all, some or one run with the one-run-only addresses, and a reading by Krippendorff's conventional cut-offs (0.800, 0.667), stated as a convention; a field with fewer than five pairable units shows its alpha with "few units" and is left out of the reading line. The alpha code reproduces the 0.743 of Krippendorff's 2011 worked example. `eval` is unchanged. `docs/commands.md` and the golden README describe it.
13
- - `smd new <folder> [--profile] [--lang] [--case] [--title]` (UX audit of 31.10, gate У В): scaffolds a case folder that validates at once, `smd validate --strict` giving 0 errors and 0 warnings on every profile: `<case>-c1.board.yaml` (required fields, `profile`, an entry where the profile has the kind, one position to rename, empty lists), a `sources.yaml` skeleton with one entry, and a README naming the first procedure (`small-situation` on the default profile, `positional-construction` on the others). It refuses, writing nothing, if a file exists. `docs/first-board.md` starts from it.
14
- - `smd status <record> [--lang] [-o file.md]`: the core view of a board set on one page of Markdown (46 lines for the MIPT case's cycle 5): state line, K not closed, hypotheses in testing and untested ones in leverage order, AI-voiced items waiting for a sign-off, transitions, counts per board and kind; lists are cut with "and N more". New `LABELS[...]["core"]` words in both languages.
15
- - `smd next <record>`: the procedure to run next with one line why and at most two alternatives, by the rule table `RULES` in `src/smd_engine/nextstep.py` (written out in `docs/commands.md`); procedures of another profile are skipped; "no rule fires" is an answer.
16
- - Hypothesis leverage (formal audit, section 4 row 5): for each untested hypothesis the size of what a refutation would reach (targets, K, ontology records testing it, theses on its targets, elements one relation away); a proposal, the group chooses. Listed by `smd status`, and a column "reach (proposal)" in the hypotheses table of `smd index`.
17
- - Run stamp v2 (the audit of 31.10, formal section 3): the record's `run` takes four optional fields, `pack_sha256` (sha256 of the `smd pack` output the run used), `inputs` (source id to the sha256 of the input file the run read), `temperature` (a number) and `sample` (the run index, when a procedure is run several times on one input). Additive: the schema stays 0.3 and old records stay valid. The schema checks the types (a 64-character lowercase hex for a hash, an integer from 1 for `sample`); `docs/data-model.md`, `run`, describes the fields.
18
- - `smd pack` prints `pack sha256: <hex>` to stderr, the hash of the bytes it wrote (the file with `-o`, the text on stdout otherwise), so that a run can record it as `run.pack_sha256`.
19
- - `smd revise` no longer copies the previous cycle's `run` into the next cycle: the new record is not the product of that run, and a carried `pack_sha256` would name a pack it was not made from. The next procedure writes its own stamp.
20
- - `smd eval` prints the whole `run` stamp of both records, `inputs` as `{id: sha256, ...}`; until now a mapping showed as a Python dict.
21
- - The "Record the run" line of `functional-plan` 0.2.5, `ingest-transcript` 0.3.1, `positional-construction` 0.2.7, `problematization` 0.2.3, `qualification-pass` 0.4.1, `reflexive-protocol` 0.3.2, `solo-session` 0.2.4, `stakeholder-ontology` 0.1.1, `theses-to-board` 0.3.6 and `visionary-tract` 0.2.3 names the new fields, in the same words: write what you know, omit the rest, invent nothing. `small-situation` and `field-result` have no such line and are unchanged.
22
- - `docs/first-board.md`: a walkthrough from an empty folder to a first validated and drawn board (an invented family bakery handing over to a manager) and on to the first procedure. It lists the required top-level fields (`title` and `status` were not named in the README), which profile to choose when, which procedure comes first and next, and what to bring. Its commands were run as shown.
23
- - `docs/glossary.md`: the terms a first-time user meets, one line each; linked from the README, `README_ru.md`, `docs/README.md` and the walkthrough.
24
- - `README.md` and `README_ru.md`: the stale "Version 0.3.6" line and the `uvx` pin in the CI example now say 0.5.9 (schema 0.3); the two files link the walkthrough and the glossary near the top; the data-model table names the required top-level fields of a board set.
25
- - Procedure `stakeholder-ontology` 0.1.0 (step 3, role игротехник, any profile): elicits the ontology of one stakeholder position, by default the owner, the tier-1 stakeholder (its stakeholder ontology is the Owner Ontology), in the position's own words, in a solo session with the holder or from an interview transcript; in proposal mode only from registered sources in which the holder speaks. Six parts with question forms: entities, definitions, absences, topology, time, causality, plus Khaprov's ontology test for `competes` and a question for the ontology-dependent records. Product: an ontology record with `valid_for` that one position, object and subject quoted with addresses, `tags` tier 1 for the owner (`{tier: [intent]}` in a case whose taxonomy names tier 1 so), and a `test` H (by default: does the shareable part work as a reading key); the shareable part (verbatim glossary of the position's terms as used in records, list of ontology-dependent records) as a separate file registered in the manifest and cited by the record; the private part (deliberate absences, causal beliefs, sources of the edge, values, judgement of people) kept off the board, the case repository and the registered log, the record's `note` saying only that it exists and who keeps it. The translation into the six-tier ontology is a separate step: an AI-voiced `concept` element whose `detail` table names, per term, the tier and what the translation loses. Typical errors include an ontology without a position, the analyst's words in the holder's mouth, the private part on the board, the owner set apart from the stakeholders, and a seventh tier. No schema or code change.
26
- - `positional-construction` 0.2.6 (section 5) and `solo-session` 0.2.3 (section 3) point to it; `stakeholder-ontology` uses the full form "six-tier ontology of a business" at first use.
27
- - `procedures/README.md`, `lenses/README.md` (selection table: no lens offered) and `docs/method/roles-and-actors.md` (where the role rule is applied) list the new procedure.
28
-
29
- - Method review 31.10.2026 (gates Ф В and У В): rules the procedures asked the model to check by hand, which are exact graph or set conditions on the record, are now checks of the validator (`docs/checks.md`).
30
- - New errors: a cycle of `belongs`, of `subordinates`, or of move sources (`move.from` of elements and theses). A cycle of `flow` is legitimate and is not checked.
31
- - New warning: an ontology position of `valid_for` with no line in `subjects` (the reverse of the existing warning).
32
- - New hints of the functional plan (functional-plan.md §4): `function-flows` (a function in a new frame with no payer, input or receiver `flow`, unless a gap with the matching `of` is linked to it), `unlinked-functions`, `idle-positions`.
33
- - New hints of the typical errors of a group screen on `program-three-boards` (qualification-pass.md §3.2): `positions-unlinked` (E2), `object-isolated` (E3), `isolated-facts` (E6, E11), `jump-to-frame` (E12), `transition-unanchored` (E14). The hint `belts` also names theses that are all мД ("mechanical reproduction").
34
- - New hints of the procedures' self-checks: `ai-cooperation` and `cooperation-flows` (positional-construction.md §5), `translation-loss` (stakeholder-ontology.md §5), `garland`, `bridge-from` and `move-via` (visionary-tract.md §1, §3, §5; `program-three-boards` only).
35
- - New hint `board-dense`: a board with more elements than the profile's new optional key `board_dense` (default 15) gets advice to fold or split it.
36
- - `smd validate --hints` prints after each hint the procedure and section that act on it, "→ procedures/functional-plan.md §4" (`Finding.procedure`, filled from `HINT_PROCEDURES` in `validate/context.py`; the field is not compared, so findings stay equal).
37
- - The three profiles enable the new hints that their kinds allow. `functional-plan` 0.2.5, `qualification-pass` 0.4.1 and `visionary-tract` 0.2.3, with `positional-construction` and `stakeholder-ontology` in their unreleased versions, say which of their checks the validator now reports.
1
+ # Changelog, releases 0.2.0 to 0.5.9
2
+
3
+ Moved from `CHANGELOG.md` on 2026-10-31; newer releases are there. Dates are those of the tag. All tags v0.2.1 to v0.3.5 were made on 2026-10-29. The version in `pyproject.toml` stayed at 0.2.0 through v0.2.4 and at 0.3.0 from v0.3.0 to v0.3.5; 0.3.6 is the first release whose package version equals its tag.
38
4
 
39
5
  ## 0.5.9 - 2026-10-29
40
6
 
@@ -16,15 +16,20 @@ The checks are in `src/smd_engine/validate/`, one module per subject, run in the
16
16
  - **Hints (G.P.'s «Программа действий», GP-Guide11, read against GP-T5):** an object-board element seen from a position with `stance: inside` (`object-from-inside`, ¶36); a new cycle whose object board is seen from an `external` position with no `reflected_from` (`external-without-reflection`, ¶35, ¶47); a goal with no breakdown: no `goal_source`, `arises_from`, gap linked by a relation, `difficulty` or `problem` thesis naming it, or `state.difficulty` (`goal-without-breakdown`, ¶43; also an `assigned` goal with no `assigned_by` and a `breakdown` goal with no `arises_from`); a position with `acts_as` (not `functionary`) and no `represents` relation (`position-without-object`, GP-Guide10 ¶7, ¶19); an object-board element with `modality: ought` and no fact linked by a relation, `in`, or a factual thesis (`ought-without-fact`, GP-T5 p. 252); `ought` on a `natural` element (`design-the-natural`, GP-STS); theses, other than `open`, with neither `feeds` nor `about` (`information-without-cell`, ¶5). Where the 1981 and 1988 texts disagree the 1988 lectures win: GP-Guide5 ¶7 required a forecast ("Управлять без прогноза в принципе невозможно"); GP-T5 p. 251 calls forecasting in оргуправление a substitution by scientistic principles, so `ought-without-fact` asks for the natural processes, not a forecast.
17
17
  - **Hints (G.P.'s management scheme, `management-four-boards`, GP-T5):** a `leads` relation to someone with goals `goal_source: own` (seen from the target or from a position that `belongs` to it) (`lead-vs-manage`, pp. 169-170); a `manages` relation with no element the source `produces` and the target `uses` (`manage-via-organizedness`, pp. 222-223); a position that `manages` with no retrospective or no projective knowledge seen from it or linked to it (`manage-without-knowledge`, p. 241); a `leads` relation whose ends hold no post (itself, or one it `belongs` to), or with no line of `subordinates` from the leader's post down to the other (`leads-outside-organisation`, pp. 242, 254); a `represents` target outside the kinds expected for the position's `acts_as`: organiser to `post`, `functional_place`, `mission`, `monosystem`; leader to `post` (not `group`: the informal group is outside the administrative structure, pp. 224-225, 254); manager to `work_process`, `trajectory`, `self_motion`; politician to `club`, `position`; functionary to none (`representation-fit`; the table is the engine's reading of GP-T5 pp. 170-175, 213, 223, 242-244, 253-254, not a table of G.P.'s); an upper board with content but no past of the object (a `factual` kind, `knowledge_retro`, or `modality: fact` on the managed board), no `ideal_state`, or no `captures` from the upper board to the managed one (`development-step`, «шаг развития», «минимальная единица», pp. 229, 232, 238); a managed board with fewer than two monosystems, or monosystems with no `parasitic`, `symbiotic` or `mechanism_of` relation (`polysystem`, pp. 206-213).
18
18
  - **Errors (Beer layer of `management-four-boards`, gate Бир3 В):** `vsm_system` on a kind the profile does not flag `vsm`; `of_whole` or `written` on a kind the profile does not flag `of_whole` (gates Г2 В, Г5б Б); `threshold` or `recovery_time` on an element that is not an `algedonic_channel`. The schema refuses a `vsm_system` that is not a non-empty list of distinct integers 1-5, an `of_whole` or `written` that is not a boolean, an empty `threshold` and a `recovery_time` that is not a number > 0. A profile that enables `single-node-metasystem`, `absorbed-into-system-1` or `no-model-of-whole` with no kind flagged `vsm`, or `no-model-of-whole` or `model-of-whole-one-filler` with no kind flagged `of_whole`, is refused.
19
- - **Hints (Beer layer of `management-four-boards`, BEER1972):** one filler fills systems 3, 4 and 5 across the posts and positions it fills (`single-node-metasystem`, Гл. 9, 13): the transfer will remove the metasystem with the person. The filler of a post or position is the one actor in its `fillers`; a post with no fillers is filled by the one position that `belongs` to it; a position with no fillers that belongs to posts with one and the same filler is that filler; an element with several fillers, or a post with several positions, is shared (a multinode, Гл. 14) and counts for nobody. An `algedonic_channel` with a `threshold` and no `recovery_time`, or with no `escalates` relation from it to a `post`, `position` or `place_for_others` (`line-without-recovery`, Гл. 17). A place filling system 5 that `captures` places of the business filling systems 4 and 5 between them (`absorbed-into-system-1`, Гл. 15): the business becomes part of the buyer's system 1, its name and identity at risk. The engine has no scenario table: a transfer scenario is laid out on the boards as the buyer's post and its `captures` relations (G.P.'s «захват», GP-T5 pp. 226-227). A model of the whole is a `model` with `of_whole: true`; a post or position holds it when it `uses` it (gate Г2 В). A post or position filling system 4 that holds no model of the whole, itself, through a position that `belongs` to it, through the post it belongs to or through another position of that post (`no-model-of-whole`, Гл. 13); the message says whether no model of the whole is on the boards or none is held there. A written model (`written: true`) does not count unless a place `uses` it: a written model still needs a keeper (gate Г7г А). A model of the whole whose holders all resolve to one named filler, by the rule above (`model-of-whole-one-filler`, Гл. 13): it leaves with the person at transfer. A model with `written: true` is written down and kept apart from the people and gets no such hint (gate Г5б Б): the person leaves, the record stays. A reflexive position counts by its fillers as a post does; a holder of another kind (a group, a place for others), a shared holder, or a holder with no named filler gives no hint, nor does a model nobody `uses`.
19
+ - **Hints (Beer layer of `management-four-boards`, BEER1972):** one filler fills systems 3, 4 and 5 across the posts and positions it fills (`single-node-metasystem`, Гл. 9, 13): the transfer will remove the metasystem with the person. The filler of a post or position is the one actor in its `fillers`; a post with no fillers is filled by the one position that `belongs` to it; a position with no fillers that belongs to posts with one and the same filler is that filler; an element with several fillers, or a post with several positions, is shared (a multinode, Гл. 14) and counts for nobody. An `algedonic_channel` with a value (a `threshold` or a `line`, gate КЛ1 В) and no recovery time (neither `recovery_time` nor `line.recovery_days`), or with no `escalates` relation from it to a `post`, `position` or `place_for_others` (`line-without-recovery`, Гл. 17). A place filling system 5 that `captures` places of the business filling systems 4 and 5 between them (`absorbed-into-system-1`, Гл. 15): the business becomes part of the buyer's system 1, its name and identity at risk. The engine has no scenario table: a transfer scenario is laid out on the boards as the buyer's post and its `captures` relations (G.P.'s «захват», GP-T5 pp. 226-227). A model of the whole is a `model` with `of_whole: true`; a post or position holds it when it `uses` it (gate Г2 В). A post or position filling system 4 that holds no model of the whole, itself, through a position that `belongs` to it, through the post it belongs to or through another position of that post (`no-model-of-whole`, Гл. 13); the message says whether no model of the whole is on the boards or none is held there. A written model (`written: true`) does not count unless a place `uses` it: a written model still needs a keeper (gate Г7г А). A model of the whole whose holders all resolve to one named filler, by the rule above (`model-of-whole-one-filler`, Гл. 13): it leaves with the person at transfer. A model with `written: true` is written down and kept apart from the people and gets no such hint (gate Г5б Б): the person leaves, the record stays. A reflexive position counts by its fillers as a post does; a holder of another kind (a group, a place for others), a shared holder, or a holder with no named filler gives no hint, nor does a model nobody `uses`.
20
20
  - **Errors (method review 31.10.2026, gates Ф В and У В; `validate/graph.py`):** a cycle of `belongs` relations, of `subordinates` relations, or of move sources (`move.from` of elements and theses, longer than the move from its own result reported above); each cycle is one error naming its ids. A cycle of `flow` is not an error: a position pays into a chain and receives its result.
21
21
  - **Warnings (method review 31.10.2026):** an ontology position of `valid_for` with no line in `subjects` (positional-construction.md §5; the reverse of the warning of step 3).
22
- - **Hints (functional plan, functional-plan.md §4; `validate/graph.py`):** over the functions `in` a new frame: a function with no payer (no `flow` into it from a position or place for others), no input (no `flow` into it) or no receiver (no `flow` from it to a position or place for others), each slot met as well by a gap linked to the function by a relation, with `of: position` for the payer, `material` or `means` for the input, `position`, `product` or `function` for the receiver (`function-flows`); functions of a frame with two or more linked by no `flow` (only `flow`) to another of them (`unlinked-functions`); positions, other than the cycle's reflexive one (`reflected_from`, `stance: reflexive`), that perform, pay for and receive nothing in the plan (`idle-positions`; places for others are left out). Enabled on `program-three-boards` and `classic-two-boards`; the management profile draws its functions on process boards, not in frames.
22
+ - **Hints (functional plan, functional-plan.md §4; `validate/graph.py`):** over the functions `in` a new frame: a function with no payer (no `flow` into it from a position or place for others), no input (no `flow` into it) or no receiver (no `flow` from it to a position or place for others), each slot met as well by a gap linked to the function by a relation, with `of: position` for the payer, `material` or `means` for the input, `position`, `product` or `function` for the receiver (`function-flows`); functions of a frame with two or more linked by no `flow` (only `flow`) to another of them (`unlinked-functions`); positions, other than the cycle's reflexive one (`reflected_from`, `stance: reflexive`), that perform, pay for and receive nothing in the plan (`idle-positions`; places for others are left out). Enabled on `program-three-boards` and `classic-two-boards`; the management profile draws its functions on process boards, not in frames. With `carries` on a `flow` (`money`, `material`, `product`, `information`), the payer is an incoming flow of `money` from a position or place, an input an incoming flow of anything else, and a receiver an outgoing flow of anything but money to a position or place, or to another function of the same frame that has a receiver (followed along the chain, a cycle of functions walked once), so a middle function needs no line of its own to the receiver and only the end of the chain must reach a position or place; a supplier's `material` is an input and not a payer. A flow with no `carries` is read as before (any flow from a position or place pays, any flow in is an input, any flow out to a position or place receives), so records that never set it report what they did.
23
+ - **Errors (AI on the boards, gate ИИ1 Б):** `ai` on an element that is neither a function nor a role (a kind the profile flags `role`), or on a gap whose `of` is not `position`; `red_line` on a kind the profile flags `light_off` ("'red_line' is not for notes"; facts and signals on the default profile). `threshold` is allowed on any element with `red_line: true`, as the red line's free-text value. The schema refuses an `ai` other than `assists` or `performs` and a `red_line` that is not a boolean.
24
+ - **Hint `ai-decision-unanswered`** (functional-plan.md §3; `validate/ai.py`; all three profiles): a function whose effective `ai` (its own, else that of the role in `performed_by`) is `performs`, that is linked to a red line, and for which no role without `ai: performs` answers. A red line is an element with `red_line: true`, or an `algedonic_channel` with a `threshold` or a `line` (the management profile's carrier of a stakeholder's red line). Linked (`validate/redlines.py`, `linked`): the function is a red line itself, it is `in` one or one is `in` it, the role in its `performed_by` is one, or a relation of any kind, either way, joins it to one. A role answers when it is of a kind the profile flags `role`, not a gap (a missing role answers for nothing), has no `ai: performs`, and is the function's `performed_by`, or `subordinates` the function (role to function), or receives its `escalates` or a `flow` with `carries: information` (function to role). `subordinates` and `escalates` exist on the management profile only; elsewhere `performed_by` and the information flow back remain. A function with its own `ai: performs` whose `performed_by` is a role without `ai: performs` is answered by that role and gets no hint: the role answers for what AI does in it. The case behind it is the MIPT case's H64: some decisions cannot be passed on, to a model or to a new manager, without people of a certain role, and the owner records each as a yes/no red line. With no `ai` on the record the hint is silent.
25
+ - **Hint `red-line-without-value`** (functional-plan.md §4; all three profiles): an element with `red_line: true` and neither a `line` nor a `threshold`. A red line is an absolute value, a threshold or yes/no that data check (the case owner 26.10); the hint asks to name it. It is a hint, not an error: the group may mark a line before it has the value.
26
+ - **Errors (red line values, gate КЛ1 В; `validate/redlines.py`, `docs/red-lines.md`):** `line` on an element that is neither `red_line: true` nor an `algedonic_channel`; a `line` with both a corridor (`min`, `max`) and `expect`, or with neither; `min` not below `max`; `until` not after `since` (or after the record's `date` when there is no `since`); `line.recovery_days` and the channel's `recovery_time` both set and different; `observed` on an element with no `line`; an observed value of the wrong type (a boolean for a yes/no line, a number for a corridor); with a manifest, an unknown `observed` source. The schema refuses a `line` without `until`, and an observation without `value`, `date` or `source`. The schema also refuses `expect` without `metric`. **Warnings:** with a manifest, an observation whose source is not of kind `field_artefact`; observations dated before the line takes force (`since`, else the record's `date`), which are not read; two observations with the same date ("one observation per date").
27
+ - **Hints (red line values):** `red-line-breach` (field-result.md §5): the breaches of a line by its observations, each with its dates, by the rule of `src/smd_engine/redline.py` (observations from the line's start to `until`, sorted by date; a run of outside values spans from its first date to the next inside value, or, still open, to the later of the last observation and the record's `date`, never past `until`; it is a breach when longer than `line.recovery_days`, else `recovery_time`; any run when neither is set). `red-line-expired` (field-result.md §5): observations dated after `until`, or a record dated after it ("the line ended on <until>: renew or drop"). `red-line-outlives-pace` (functional-plan.md §4): a line in force (`until` minus `since`, or minus the record's `date`) more years than the `pace` of an element it is linked to (as above, itself included), the analogue of `ontology-outlives-pace`: a line is in force for a term matched to the pace of its tier (the case owner 26.10; the MIPT case's H24).
23
28
  - **Hints (typical errors of a group screen, qualification-pass.md §3.2; `program-three-boards`):** positions, other than the reflexive one, with no `shift` or `flow` to another position, place or cooperation (`positions-unlinked`, E2); object-board content, gaps and notes aside, linked by no relation, transition, `in`, `performed_by` or move, however indirectly, to an element of another board (`object-isolated`, E3; an ontology's `cooperation` counts as linked to its `valid_for` positions; `seen_from` does not count, since E3 asks for a link besides it); facts with no relation, transition, `in`, `feeds` or `about` (`isolated-facts`, E6, E11); the starting position of a `shift`, other than the reflexive one, with nothing seen from it on a board with halves while a new frame is drawn (`jump-to-frame`, E12; the engine's reading: the order of the transitions is not read); a transition with no end that is a position, a place for others or seen from a position (`transition-unanchored`, E14). E1, E7 and the rest of E14 are the hints on halves, entry and `seen_from` above; the other rows need reading.
24
29
  - **Hints (self-checks of the procedures):** an AI-drafted `cooperation` that no K targets (`ai-cooperation`, positional-construction.md §5); an ontology whose `valid_for` positions have no `flow` to or from its `cooperation` (`cooperation-flows`); a `detail` table with a column «что теряется» or "what is lost" whose cell is empty or says none (`нет`, `none`, `-`, `—`, `n/a`) in a row (`translation-loss`, stakeholder-ontology.md §5); on `program-three-boards`, a `garland` with fewer than two `belongs` relations into it, or whose events all cite one source ("one witness") (`garland`, visionary-tract.md §3); a bridge, a `situation` on the big half with a `respecify` move, whose move is not from a `model` and a `garland` (`bridge-from`, §5); a candidate move of the tract (`glue` on a model, `respecify` or `rescale` on a situation, `adjoin` on a goal) with no `via`; a move the group made in its talk needs none (`move-via`, §1). Not checked: that a stakeholder ontology names exactly one position (no field marks a record as one; several positions with no cooperation get the `ontology` hint), "one witness" by actor (signals carry no actor field), and three cycles in a row without a new scheme (the validator compares a record with its previous one only).
25
30
  - **Errors (hypothesis standard, gates М1 В and Ф В; `validate/hypotheses.py`):** `observed` with k greater than n; the schema refuses an unknown `type`, a `rule` without `metric`, `op` or `threshold` or with another `op`, an `n_min` below 1, and an `observed` without `k`, `n` or `artefact`. `observed.artefact` is checked against the manifest like `artefacts`. The schema refuses `==` as an `op`: equality is no refutation condition.
26
31
  - **Warnings (hypothesis standard):** a recorded `confirmed` while the whole Wilson interval of `observed` lies on the refuting side of the rule's threshold, or `refuted` while it lies on the holding side; with a manifest, an `observed.artefact` whose source is not of kind `field_artefact`.
27
- - **Hints (hypothesis standard, `procedures/field-result.md` §0):** an untyped hypothesis whose `raised_in` cycle has typed hypotheses (`hypothesis-type`; records with none typed, like the MIPT case's, get none); a `claim` or `estimate` with neither `rule` nor `revise_if` (`hypothesis-rule`; such a hypothesis gets no `revise-if` hint, nor does a `core`, `design` or `probe`, and a `rule` stands in for `revise_if`); a `rule` with no `basis` (`rule-basis`); an `estimate` whose `rule` has no `n_min` (`rule-n-min`); a `probe` confirmed or refuted, or a `core` or `design` not `untested` (`type-status`); for each hypothesis with `observed` still `untested` or `testing`, the verdict its rule proposes, by the 95% Wilson score interval of k/n when the threshold is a share strictly between 0 and 1 (`verdict`, field-result.md §3: too few, refuted, confirmed, or inconclusive and stays testing; the group signs `accept_result`).
32
+ - **Hints (hypothesis standard, `procedures/field-result.md` §0):** an untyped hypothesis whose `raised_in` cycle has typed hypotheses (`hypothesis-type`; records with none typed, like the MIPT case's, get none); a `claim` or `estimate` with neither `rule` nor `revise_if` (`hypothesis-rule`; such a hypothesis gets no `revise-if` hint, nor does a `core`, `design` or `probe`, and a `rule` stands in for `revise_if`); a `rule` with no `basis` (`rule-basis`); an `estimate` whose `rule` has no `n_min` (`rule-n-min`); a `probe` confirmed or refuted, or a `core` or `design` not `untested` (`type-status`); for each `testing` hypothesis with `observed`, the verdict its rule proposes (an `untested` one with `observed` gets the hint "observed data recorded but status is untested: set testing" instead), by the 95% Wilson score interval of k/n when the threshold is a share strictly between 0 and 1 (`verdict`, field-result.md §3: too few, refuted, confirmed, or inconclusive and stays testing; the group signs `accept_result`).
28
33
  - **Hint `board-dense`:** a board, record boards included, with more elements than the profile's `board_dense` (default 15; `docs/profiles.md`), worded as advice to fold what is settled or split a part off, never an error; it points here, since no procedure folds a board. The default is low on purpose and tunable: working memory holds about four chunks when rehearsal and chunking are blocked (Cowan, N. (2001). The magical number 4 in short-term memory: a reconsideration of mental storage capacity. *Behavioral and Brain Sciences*, 24(1), 87-114). That figure is about what one keeps in mind at once, not about a screen, and no study gives a number of elements per diagram, so 15 is a heuristic to be tuned on the case (the MIPT case's second cycle carried 33 elements on three boards, about 11 per board).
29
34
  - **Errors (record boards):** reported with the schema errors: a record board of an unknown template, with an id of a profile board or template or `records`/`sources`, or twice; an element of a known kind its record board's template does not take. Later: a record board id that is also an element id, a `seen_from` that is neither a position nor a reflexive position, an unknown `unfolds` or one of a kind outside the template's `unfolds`. **Warnings:** a record board with no `seen_from` when its template sets `needs_position`; a kind whose `home` is a template, on another board ("usually lies on a board of template 'process'").
30
35
  - **Errors (v0.3):** `scheme` on an element that is not a gap or on a thesis that is not a problem; a `detail.keys` column that is not among `columns`; with a taxonomy file, an unknown taxonomy or value in `tags` or in a bound column.
@@ -7,7 +7,7 @@ Exit codes: 0 success; 1 the check found errors (or, with `--strict`, warnings;
7
7
  ```bash
8
8
  uv run smd new path/to/case --profile program-three-boards --lang en --title "Family bakery" # a case folder that validates at once: the first record, a source manifest, a README with the next steps; --case sets the case id (default: from the folder name); nothing is overwritten
9
9
  uv run smd status cases/example/bakeries-c2.board.yaml --lang en -o out/status.md # the core view of a board set on one page of Markdown: state line, open K, hypotheses, sign-offs waiting, transitions, counts
10
- uv run smd next cases/example/bakeries-c2.board.yaml # the procedure to run next, with one line why and at most two alternatives; a rule table, no model
10
+ uv run smd next cases/example/bakeries-c2.board.yaml # the procedure to run next, with one line why and at most two alternatives, each with its section; hint pointers and a few rules, no model
11
11
  uv run smd validate cases/example --sources cases/example/sources.yaml --hints # a directory: each cycle is also checked against the one it revises
12
12
  uv run smd validate cases/example_ru --sources cases/example_ru/sources.yaml # the same case with Russian labels
13
13
  uv run smd validate path/to/case --taxonomies path/to/taxonomies.yaml # check tags and detail columns bound by detail.keys against the case's taxonomies
@@ -30,7 +30,7 @@ uv run smd vault cases/example/bakeries-c2.board.yaml out/vault --sources cases/
30
30
  uv run smd pack solo-session cases/example/bakeries-c2.board.yaml --sources cases/example/sources.yaml -o out/pack.md # attach to any model; --sources carries the manifest, --taxonomies the taxonomy file; with board sets, the skeleton of the latest cycle as "Cells to fill"; prints `pack sha256: <hex>` to stderr, the hash of what it wrote, for `run.pack_sha256`
31
31
  uv run smd sources cases/example/sources.yaml --clean-room --deny-lineage earlier-concept
32
32
  uv run smd sources path/to/sources.yaml --folder /local/folder
33
- uv run smd sources path/to/sources.yaml --folder /local/folder --refresh [ids] # recompute sha256 of the named sources (default: every decision_log) and, where a source carries `lines`, its line count; edit only those lines of the manifest
33
+ uv run smd sources path/to/sources.yaml --folder /local/folder --refresh [ids] # recompute sha256 of the named sources (default: every decision_log); a source that already carries `lines` gets its count rewritten, one that does not is not given one; edit only those lines of the manifest
34
34
  uv run smd compare a.board.yaml b.board.yaml -o out/compare.md --merge out/merge.board.yaml
35
35
  uv run smd eval run.board.yaml cases/example/golden/bakeries-c2-expected.board.yaml --min-agreement 10 --max-only-expected 3 # did a procedure run get worse: the five compare counts, the `run` stamps of both records (procedure, version, model, date, `pack_sha256`, `inputs`, `temperature`, `sample` when present), theses matched by cites, K and H matched by wording; exit 1 on a missed threshold
36
36
  uv run smd agree out/run1.board.yaml out/run2.board.yaml out/run3.board.yaml [--fields belt,kind] [--json] # how stable are the labels across N runs of one procedure on one input: Krippendorff's alpha per label field, and the share of cite addresses every run found; see below
@@ -45,7 +45,7 @@ uv run smd apply cases/example/bakeries-c2.board.yaml cases/example/patches/bake
45
45
 
46
46
  `publish` validates the record and the earlier cycles it `revises` (found next to it). An error anywhere in that chain stops the bundle. With `--strict`, a warning on the published record stops it too; warnings on earlier cycles are printed, marked "earlier cycle: not failing", and the bundle is written. The bundle's index links files relative to the `--out` folder, so the folder can be moved or sent: the record file and `--archive` files are linked by relative paths, a file that shares no folder with `--out` except the root is named without a link, and `YAML:N` is a line number, not a VS Code link. `smd index` links the same way, relative to the folder of `-o`, or to the current folder when it prints the page; neither page holds an absolute path of the building machine.
47
47
 
48
- `sources --folder` checks that every listed file is present, has a sha256, and matches it. `--refresh` rewrites the sha256 instead, for logs that grow.
48
+ `sources --folder` checks that every listed file is present, has a sha256, and matches it. `--refresh` rewrites the sha256 instead, for logs that grow. It prints `unchanged` or `sha256 updated, N lines` for each source with a file; N is the line count of the file as it is now, printed whether or not the entry carries `lines`, and only an entry that has a `lines:` line has it written to the manifest.
49
49
 
50
50
  `agree` takes two or more board sets, runs of one procedure on one input (the golden README shows how to make them), and measures how stable the labels the model writes are. It warns, and goes on, when `run.procedure`, `run.pack_sha256` or `run.inputs` differ between the records or are missing from some. The records must follow one profile.
51
51
  - **Units.** A thesis is matched across runs by the address it cites, as `eval` pairs theses (same source, a `#t=` time within 30 s); a run with no thesis at an address gives a missing value. A thesis with no `cites` has no address and is not counted. An element of the first record is matched in each other run as `compare` matches it; elements only a later run has are not units.
@@ -64,29 +64,30 @@ uv run smd apply cases/example/bakeries-c2.board.yaml cases/example/patches/bake
64
64
  `status <record>` prints the core view, in this order and nothing else:
65
65
  - the state line (cycle, version, status, the cycle it revises, how many items are accepted and open) with `state.difficulty` and `state.next_move`;
66
66
  - the K not closed, open first and then partly answered;
67
- - the hypotheses in `testing`; then, for each `untested` or `testing` hypothesis with `rule` and `observed`, the verdict the rule proposes by the Wilson interval (`docs/data-model.md`, Hypotheses; a proposal, the group signs `accept_result`); then the `untested` ones in leverage order, below;
67
+ - the hypotheses in `testing`; then, for each `testing` hypothesis with `rule` and `observed`, the verdict the rule proposes by the Wilson interval (`docs/data-model.md`, Hypotheses; a proposal, the group signs `accept_result`); then the `untested` ones in leverage order, below;
68
+ - the red lines their observations breach, each with the dates of its breaches (`docs/red-lines.md`);
68
69
  - AI-voiced items waiting for a human sign-off: thesis variants (`ai_variant`), candidate moves, K and H drafted by an AI that are not closed or resolved;
69
70
  - the transitions;
70
- - the counts of elements per board (halves of the situation board apart) and per kind.
71
+ - the counts of elements per board (halves of the situation board apart) and per kind; when any element has `ai` (gate ИИ1 Б), one more line, «функции: выполняет ИИ n, ИИ помогает k, из m» ("functions: AI performs n, AI assists k, of m"): the functions by their effective `ai` (`docs/data-model.md`, Elements) and all functions. `smd index` prints the same line at the top of the page; neither prints it for a record with no `ai`.
71
72
 
72
73
  Lists are cut to a fixed length and end in "and N more" (8 K, 5 hypotheses in testing, 8 verdicts, 5 untested, 5 transitions, 6 ids per waiting item, 8 kinds), one line per item. The page for the MIPT case's cycle 5 (132 elements, 19 K, 40 hypotheses) is 46 lines. The labels come from the same `LABELS` tables as the other pages; `--lang ru` is the default.
73
74
 
74
75
  **Hypothesis leverage (a proposal).** For each untested hypothesis the engine collects what a refutation would reach: its `targets`; its `qualifications` (K); the ontology records whose `test` is this hypothesis; the theses whose `feeds` or `about` hold one of its targets; and the elements one relation away from a target (either end of a relation, a target not counted twice). The score is the size of that set, the order is by score and then by id with numbers read as numbers (H2 before H10). It is the same for the same record. It is not a probability and carries no weight: a hypothesis can be worth testing for a reason the board does not show. The group chooses which to test; the order only saves it from counting by hand. `smd status` lists it, and `smd index` shows the score in the hypotheses table (column "reach (proposal)", the parts of the count as the tooltip).
75
76
 
76
- **`next <record>`.** The rules are read top to bottom; the first whose condition holds names the procedure to run (`smd pack <procedure> <record>` is printed), the next two are shown as alternatives, each procedure once. A procedure written for another profile than the record's is skipped, as `smd pack` would refuse it. If no rule fires, the output says so. The record must validate; its hint findings are read through `validate_record`. The rules sit in one list, `RULES` in `src/smd_engine/nextstep.py`, in the order of `procedures/solo-session.md` (section 3) and the `next` keys of `procedures/README.md`:
77
+ **`next <record>`.** Candidates come from two places, and are then put in one order. The record must validate; its hint findings are read through `validate_record`.
78
+ - **Hints.** Every hint finding carries the procedure and section that act on it (`HINT_PROCEDURES`, `src/smd_engine/validate/context.py`; the same pointer `validate --hints` prints). Each is a candidate with the hint's message as its reason (the reference and the message, cut at 200 characters). A hint that points to a document and not to a procedure (`board-dense`, `docs/checks.md`) is printed as a note under the answer, never as a procedure.
79
+ - **Rules.** The conditions that no hint reports stay explicit rules, `RULES` in `src/smd_engine/nextstep.py`, each with the procedure and section it points to:
77
80
 
78
81
  | # | Condition | Procedure | Why |
79
82
  |---|---|---|---|
80
- | 1 | the record revises a cycle and names no `reflected_from`, or `state.difficulty` is empty | `reflexive-protocol` | the reflexive questions of the closed cycle come first (gate 0) |
81
- | 2 | a thesis has neither `feeds` nor `about` | `theses-to-board` | information not laid into a cell of a scheme |
82
- | 3 | the small half of the situation board is empty (hint) | `small-situation` | the four questions come first |
83
- | 4 | the big half is empty (hint) | `visionary-tract` | the big situation is not laid out |
84
- | 5 | fewer than two positions and no place for others (hint) | `positional-construction` | the positional set looks incomplete |
85
- | 6 | a new frame has no functions, or functions with no performer (hint) | `functional-plan` | the plan names who performs and who pays |
86
- | 7 | a gap or a problem thesis names no block of a named scheme (hint) | `problematization` | plus and minus on a named scheme |
87
- | 8 | the record has no K | `qualification-pass` | not yet checked against the criteria |
88
- | 9 | cycle 1 and no theses | `ingest-transcript` | with a recorded group talk, start from it (`next` cannot see the sources) |
89
- | 10 | a hypothesis is in `testing` | `field-result` | if the field gave an artefact, record the result |
90
- | 11 | none of the hints of rules 3 to 7 is open and the record has K | `reflexive-protocol` | closes the cycle and opens the next |
83
+ | 1 | the record revises a cycle and names no `reflected_from`, or `state.difficulty` is empty | `reflexive-protocol` §2, ranked before everything (gate 0) | the reflexive questions of the closed cycle come first |
84
+ | 2 | a thesis has neither `feeds` nor `about`, and no hint points to `theses-to-board` §2 (a hint at §3, E3, does not count) | `theses-to-board` §2 | information not laid into a cell of a scheme (the hint `information-without-cell` says the same when the profile enables it) |
85
+ | 3 | the record has no K | `qualification-pass` §3 | not yet checked against the criteria |
86
+ | 4 | cycle 1 and no theses | `ingest-transcript` §0 | with a recorded group talk, start from it (`next` cannot see the sources) |
87
+ | 5 | a hypothesis is in `testing` | `field-result` §1 | if the field gave an artefact, record the result |
88
+ | 6 | the record has K, and no hint points to a procedure before `reflexive-protocol` in `ORDER` (below) | `reflexive-protocol` §7 | closes the cycle and opens the next |
89
+
90
+ - **Order.** The candidates are sorted by `ORDER` in `nextstep.py`. Gate 0 (rule 1) goes before everything, as `procedures/solo-session.md` section 2 puts the reflexive answers before step 1. Then the step-2 procedures, which lay information on the board before the thinking steps (GP-Guide11 paragraph 5; the hint `information-without-cell`): `ingest-transcript`, `theses-to-board`, `field-result`. Then the order of `solo-session.md` section 3: `small-situation`, `visionary-tract`, `stakeholder-ontology` (before the positional construction, as the note under that table says), `positional-construction`, `functional-plan`, `problematization`, `qualification-pass`. Then `reflexive-protocol` (step 7, which closes the cycle) and `solo-session` (which only orders the others). A procedure is listed once, with the sections of all its candidates (`functional-plan §3, §4`) and the reason of the first of them, rules before hints. A procedure written for another profile than the record's is skipped, as `smd pack` would refuse it.
91
+ - **Answer.** The first is the proposal (`smd pack <procedure> <record>` is printed), the next two are alternatives, each with its section. If nothing is left, the output says so. The same record always gives the same answer, and the answer is a proposal: the owner or the group may run another step (`solo-session`, section 3).
91
92
 
92
93
  What people have to decide (K not closed, AI variants, `state.open`) is not a procedure; `next` reports the counts above its answer. The answer is a proposal: the owner or the group may run another step (`solo-session`, section 3).