v0.1.36

Added

• Modular parser architecture — All 13 language parsers extracted from generator.py into okf/parsers/ (one file per language). Adding a new language is a single file + single registry entry. See docs/development.md.
• okf enrich command — Standalone enrichment of an existing bundle without re-scanning. Reads source_root from bundle index.md frontmatter (stored at generate time). Usage: okf enrich <bundle_dir> [--mode base|deep|security|full].
• Enrichment mode selection — --enrich now accepts mode qualifiers: --enrich base (descriptions/docstrings), --enrich deep (+usage examples, side effects, security, complexity), --enrich security (risk audit only), --enrich full (all passes). Backward compatible — plain --enrich defaults to base.
• --security standalone mode — okf generate --security <source_dir> <bundle_dir> audits security/complexity on an existing bundle. Uses _upsert_section() for surgical file patching — only touches Security/Complexity sections, leaves all others intact.
• Semantic related-links pass — enrich_related_semantic() uses LLM to find relevant cross-links beyond the call graph. Gated behind enrich.semantic_related.enabled. Includes deterministic candidate pre-filtering (_candidate_pool()) to keep prompt sizes bounded.
• source_root stored in bundle metadata — render_root_index() writes source_root to index.md frontmatter. _read_source_root() reads it back. _read_body() falls back to bundle metadata when no source dir is provided.
• _get() multi-level dot-notation — now supports arbitrary depth (e.g. enrich.deep.provider).
• Anthropic SDK support — _resolve_client() uses anthropic.Anthropic for "anthropic" provider, openai.OpenAI for all others.
• docs/development.md — New developer guide with full parser onboarding walkthrough, method reference table, and shared utilities index.
• 7 new unit tests — provider resolution cascade, per-mode overrides, built-in provider lookups, multi-level _get.
• 6 new CLI test scenarios in TEST.md — okf enrich modes and --enrich mode qualifier validation.

Changed

• generator.py reduced from 3509 to ~1630 lines — all parsers extracted to okf/parsers/. Remaining code (rendering, enrichment, scanning, writer, main) is focused and readable.
• Concept dataclass moved to okf/parsers/base.py — re-exported from generator.py for backward compat.
• config.py DEFAULTS now includes providers.* and enrich.* sections. Default LLM provider changed to "local".
• load() uses copy.deepcopy(DEFAULTS) instead of mutable reference merge — fixes test isolation.
• DEEP_ENRICH_PROMPT expanded from 2 fields to 4 (usage_example, side_effects, security, complexity). max_tokens bumped 350→500.
• _read_body() now accepts optional bundle_dir parameter for body-dependent passes without a source directory argument.
• CLI docstring updated to document --enrich modes and okf enrich command.
• README updated: new Enrichment Modes section, multi-provider routing config example, modular parsers table with file paths, comparison table with 2 new rows, FAQ links to docs/development.md.
• Version bumped to 0.1.36.

---