v1.9.3

DeepDoc 1.9.3 fixes backend navigation quality for real-world repos: path-slug sections produced by the classify LLM (e.g. new-src-api-services-order-index-ts) are now detected and replaced with proper domain sections, database and runtime buckets use flat canonical section names, and OpenAPI specs are rewritten so Fumadocs can resolve paths directly.

Added

• Added _looks_like_path_slug_section() in bucket_injection.py and nav_shaping.py — detects all-lowercase-hyphenated section values that look like file path cluster IDs (e.g. new-src-api-services-sync-index-ts) so they are always recategorized rather than preserved.
• Added _is_backend_placeholder_section() in bucket_injection.py — rejects generic placeholder section names (Architecture, Core, Features, Services, Subsystems, Runtime & Frameworks) for backend repos so they are replaced with a specific domain section.
• Added _BACKEND_INTEGRATION_TOKENS, _BACKEND_OPERATION_TOKENS, and _BACKEND_RUNTIME_TOKENS token sets in bucket_injection.py for finer-grained backend section routing (Integrations, Operations, Background Jobs).
• Added a dedicated _BACKEND_ORDER section ranking list in nav_shaping.py — backend repos now get a fixed reader-first ordering: Start Here → Overview → Core Workflows → API Reference → Data Model → Background Jobs → Integrations → Operations → Supporting Infrastructure → tail sections.
• Added _spec_base_path() and _write_spec() helpers in pipeline_v2.py — _spec_base_path extracts the server base path from OpenAPI 3 servers[] or Swagger 2 basePath; _write_spec writes the rewritten spec as YAML or JSON depending on file suffix.
• Added "Last indexed" provenance badge to the generated Fumadocs docs page (scaffold_files.py) — reads deepdoc_generated_at and deepdoc_generated_commit from page frontmatter and renders a muted timestamp + short commit ID above the page body.

Changed

• _canonical_section_for_bucket() now rejects path-slug sections and backend placeholder sections before preserving the existing section, so classify-step noise no longer leaks into final nav.
• Backend bucket section routing reordered: Data Layer check moved before runtime/integration checks; Subsystems fallback replaced with Core Workflows; added explicit Operations bucket for config/logging/monitoring tokens; Supporting Infrastructure replaces the old Runtime & Frameworks catch-all for middleware/auth/route/controller buckets.
• Normalized all database bucket sections from Database > Database & Schema to flat Data Model (specializations.py).
• Normalized all runtime bucket sections from Background Jobs > Background Jobs & Runtime to flat Background Jobs (specializations.py).
• OpenAPI staging in pipeline_v2.py now rewrites the spec before saving: bakes the server base path into every path key and resets servers to [{"url": "/"}] so Fumadocs can do a direct dict lookup without prepending the server origin.
• OpenAPI nav entries are now placed under a top-level API Playground folder instead of being nested inside the existing API Reference folder (engine.py).

Tests

• Added test_shape_plan_nav_recategorizes_path_slug_backend_sections — verifies that new-src-* path-slug sections are all replaced and buckets land in Core Workflows, Integrations, Operations, and Supporting Infrastructure as appropriate.
• Updated test_shape_plan_nav_backend_uses_reader_flow_and_dedupes_setup section assertion from Architecture to Core Workflows to match new routing.
• Updated test_specialized_bucket_injection_splits_large_database_docs_and_adds_runtime_pages assertion from Database > Database & Schema to Data Model.
• Updated test_build_fumadocs_from_plan_creates_site_scaffold to assert provenance badge fields (deepdoc_generated_at, deepdoc_generated_commit, Last indexed:, Intl.DateTimeFormat) are present in the generated docs page.