Release notes | Hanalyx Docs

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

[0.3.1] - 2026-04-04

Route-path ID now takes priority over generic-filename logic for Next.js App Router (onboarding-route → onboarding)
Zod .min(N, "message") / .max(N, "message") / .email("msg") with custom error messages now extracted correctly
CLI discovers prisma/schema.prisma from parent directories when scanning a subdirectory
Inline Python comment false positives eliminated (# isort:skip, # noqa on import lines)

--overwrite flag for specter reverse — existing spec files are skipped by default, preserving manual edits
--exclude flag for specter reverse — scope scans with path exclusions
3-tier pre-release CI workflow validating against 12 open-source repos (chi, fiber, validator, FastAPI, pydantic, django, cal.com, create-t3-app, payload, tRPC, TanStack, refine)

specter.yaml project manifest — defines system metadata, domain grouping, coverage thresholds, and spec registry
specter init command — scaffolds specter.yaml from existing specs with --name and --force flags
Domain grouping — group specs by business area (payments, auth, content) with tier inheritance
Tier cascade — spec tier → domain tier → system tier → default (2)
Configurable coverage thresholds per tier via manifest settings.coverage
Auto-maintained spec registry rebuilt on every sync run
Domain-level coverage aggregation
spec-manifest.spec.yaml — the manifest's own spec (10 constraints, 14 ACs)
104 tests (up from 85), 7 specs (up from 6)

Route-path ID now takes priority over generic-filename logic for Next.js App Router files (onboarding-route → onboarding, slug-route → blog-slug)
Zod .min(N, "message") and .max(N, "message") with custom error messages now extracted correctly (previously the closing paren regex failed when extra args were present)
Zod .email("message") and .url("message") with custom messages now extracted correctly
CLI discovers prisma/schema.prisma from parent directories when scanning a subdirectory like src/

--exclude flag for specter reverse — exclude paths from scanning (e.g., --exclude src/components --exclude "*.test.*")

Strip inline Python comments before constraint extraction — eliminates false positives from # isort:skip, # noqa, etc. on import lines (Django: 67 → 31 constraints, 36 false positives removed)

P0: Map unknown validation.rule values to "custom" — Specter no longer rejects its own output for Go struct tags (gte, lte, oneof), Python Field kwargs (min_length, max_length), or Prisma attrs (unique)
P1: Add .spec.tsx, .spec.jsx, .spec.js to TypeScript test file detection — previously 713 test assertions silently lost in refine
P1: Fix test description truncation on embedded quotes — it("'visible' value") no longer stops at the first '
P1: Filter Python comment directives (# isort, # noqa, # type:, # pragma) from constraint extraction
P2: Incorporate parent directory into spec ID for generic filenames (index.ts → auth-index, main.go → rest-main, route.ts → users-route)

normalizeValidationRule() in core engine with alias mapping (gte→min, lte→max, oneof→enum, etc.)
Generic filename detection list for spec ID collision prevention
Improvement roadmap document (docs/IMPROVEMENT_ROADMAP.md)
Updated CLAUDE.md with mission statement, design principle, and bug priority framework

Fix null validation.value crash when Zod patterns have no extractable literal (.email(), .url(), .optional(), .refine(), z.boolean(), z.array())
Fix all Next.js App Router files generating the same spec ID route — now derives ID from route path (e.g., /api/webhooks/stripe → webhooks-stripe)
Include source file path in validation_failed diagnostics for easier debugging
Find package.json/go.mod/pyproject.toml in parent directories for system name inference

TypeScript adapter: extract constraints from TypeScript enums, union types, and as const arrays
TypeScript adapter: extract constraints from Prisma schema files (.prisma) — field types, @unique, @db.VarChar(N), required/optional
TypeScript adapter: extract role and status constraints from code patterns (e.g., session.user.role === "ADMIN")
TypeScript adapter: recognize .url(), z.boolean(), z.array() Zod patterns
Zod enum now extracts values (e.g., z.enum(["a", "b"]) → Value: "a", "b")
85 tests (up from 77)

specter reverse command — reverse compiler extracts draft .spec.yaml from existing code
Plugin adapter architecture with 3 built-in adapters: TypeScript, Python, Go
TypeScript adapter: Zod schemas, Jest/Vitest tests, Next.js/Express routes
Python adapter: Pydantic models, pytest tests, FastAPI/Django routes
Go adapter: struct validate tags, table-driven tests, net/http/gin/chi routes
Gap detection: flags constraints without test coverage
Auto-detection of language adapter from file extensions
spec-reverse.spec.yaml (10 constraints, 14 ACs)
77 total tests (up from 37)

Migrated from TypeScript/Node.js to Go (single static binary, zero runtime dependencies)
All 5 MVP tools rewritten: parse, resolve, check, coverage, sync
Cross-platform binary distribution via goreleaser (linux, darwin, windows)
DEB package available for Debian/Ubuntu