coulomb/markitect-main
2
0
Fork 0
Code Issues 61 Pull Requests Actions Projects 12 Releases 1 Wiki Activity
639 Commits 1 Branch 12 Tags
b7e9edbb4b2fc61b87cbf8a26fc7d03d3023d573
Commit Graph

5 Commits

Author SHA1 Message Date
tegwick
81a4c8796a feat(infospace): add L2 entity classification with type × VSM matrix (S2.9) 
Implements the L2 typed-entities layer — each entity is assigned an
Entity Type (Element, Process, Relation, Principle, Institution) and a
VSM System (S1–S5) by an LLM, with one-sentence rationales for each.

New modules:
- markitect/infospace/classification.py — EntityClassification dataclass
  + ENTITY_TYPES / VSM_SYSTEMS controlled vocabularies
- markitect/infospace/classification_io.py — write/read classification
  files (YAML frontmatter + markdown body, mirrors evaluation_io)
- markitect/infospace/classifier.py — build_classification_prompt(),
  parse_classification_response(), run_entity_classification(); batch
  runner writes files incrementally (same resumable pattern as evaluate)

CLI: markitect infospace classify [--entity SLUG] [--provider P] [--model M]
  - Incremental skip: checks output/classifications/ for existing files
  - Defaults to openrouter provider; 2000 max_tokens (Gemini 2.5 Flash
    uses ~787 thinking tokens, so 800 was too low)

CLI: markitect infospace classify-summary [--update-metrics]
  - Entity type counts + VSM system counts with percentages
  - 5 × 6 type × VSM matrix (spots structural blind spots at a glance)
  - --update-metrics writes type_distribution, type_entropy,
    vsm_type_matrix_cells to metrics.yaml

Config: InfospaceConfig gains classifications_dir (default output/classifications)
Schema: schemas/typed-entity-schema-v1.0.md — type/VSM vocabulary tables,
  rationale format rules, validation rules, metrics enabled at L2
infospace.yaml: schemas.typed_entity references typed-entity-schema-v1.0.md

Seed classifications (3): division_of_labour (Process/S1),
  natural_price_as_central_price (Principle/S2),
  invisible_hand_mechanism (Principle/S4)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-02-23 09:35:58 +01:00
tegwick
2d45425b25 feat(infospace): add L3 relation graph with VSM-aware triplets (S2.8) 
Implements the L3 relation graph layer — a directed graph of (Subject,
Predicate, Object) triplets annotated with VSM channel codes and feedback
roles. Triplets are authored as markdown files under output/relations/,
parsed into RelationMeta dataclasses, and analysed with networkx.

New modules:
- markitect/infospace/relation_models.py — RelationMeta dataclass +
  RELATION_TYPES controlled vocabulary (15 relation classes → VSM codes)
- markitect/infospace/relation_parser.py — parse_relation_file() and
  parse_relations_directory()

New schema: examples/infospace-with-history/schemas/relation-schema-v1.0.md
  — file naming convention, required sections, controlled vocabulary table

15 seed relation files covering the three core WoN feedback loops:
  - Capital Accumulation loop (positive reinforcement, S1/S3)
  - Market Price Balancing loop (negative feedback, S2/S3)
  - Market Extent mutual dependency (S1/S2)
  Plus structural relations: wages regulation, rent residual, price
  decomposition, invisible hand coordination

CLI: markitect infospace relations [--entity SLUG] [--vsm FILTER]
     [--loops] [--stats]
  - Builds directed graph from parsed files
  - Detects feedback loops via nx.simple_cycles()
  - 6 loops found from 15 seed relations (3 intended + 3 emergent)
  - --stats aggregates by VSM system code (strips parentheticals)

Config: InfospaceConfig gains relations_dir (default output/relations)
infospace.yaml: schemas.relation references relation-schema-v1.0.md

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-02-23 06:04:28 +01:00
tegwick
fa27572f43 fix(example): skip prompt writes when output exists, add quality rubrics 
INFRA-TASKS #5 — process_chapters.py now skips writing *-prompt.md files
when the corresponding output file already exists on disk. DB-only rebuilds
no longer dirty the working tree with unchanged prompt content.

INFRA-TASKS #8 — Added '## Quality Metrics' section to the entity and VSM
mapping schemas, defining the five evaluation dimensions (Definition Precision,
Source Grounding, Domain Placement, VSM Relevance, Explanatory Value) with
1–5 rubrics used by the evaluate-entity template.

Also updated INFRA-TASKS.md to reflect current resolution status for tasks
4–19 across S2 and S3.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-02-23 06:04:09 +01:00
tegwick
8095a1da4c fix(example): standardise domain enum and source chapter format in schema/rules 
Two root causes of metric fragmentation observed in collection checks:

1. Schema's Economic Domain used free-form examples ("labour economics,
   trade theory") which overrode the enum in extraction-rules.md, causing
   the LLM to produce multi-domain strings and non-canonical values.
   Fix: schema now specifies the exact 7-value enum with descriptions.

2. Source Chapter had no format constraint, producing 9 different formats
   for 7 chapters (full titles, mixed Roman/Arabic numerals, asterisks).
   Fix: extraction-rules now mandate "Book [Roman], Chapter [n]" exactly.

These fixes are prerequisites for clean reprocessing (S3.2 continuation).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-02-19 13:01:09 +01:00
tegwick
fecc2fd4fa feat(llm): add LLM integration module with OpenRouter and Claude Code adapters 
Implements markitect/llm/ package with concrete LLMAdapter implementations:
- OpenRouterAdapter: HTTP via urllib with retry/backoff on 429/5xx
- ClaudeCodeAdapter: subprocess-based Claude CLI with stdin piping
- Factory pattern: create_adapter("openrouter") or create_adapter("claude-code")
- API key resolution chain: constructor > env var > project-root key file
- 42 unit tests, 2 integration tests (gated on API key / CLI availability)

Also adds the infospace-with-history example with Wealth of Nations VSM
analysis pipeline, templates, schemas, source chapters, and processed
output for chapters 1-2. process_chapters.py now supports --provider
and --model flags for automatic LLM-driven processing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-02-11 01:17:58 +01:00