codd-dev 2.17.0__tar.gz → 2.18.0__tar.gz

{codd_dev-2.17.0 → codd_dev-2.18.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codd-dev
-Version: 2.17.0
+Version: 2.18.0
 Summary: CoDD: Coherence-Driven Development — cross-artifact change impact analysis
 Project-URL: Homepage, https://github.com/yohey-w/codd-dev
 Project-URL: Repository, https://github.com/yohey-w/codd-dev
@@ -68,6 +68,14 @@ codd dag verify --auto-repair --max-attempts 10  # AI fixes coherence violations
 
 That's it. Three commands, three feedback loops, one coherent project.
 
+### 💡 Already shipping? Describe what you want fixed.
+
+```bash
+codd fix "login error messages are hard to understand"   # natural-language phenomenon mode
+```
+
+`codd fix [PHENOMENON]` (v2.16.0+) is the **second entry-point** of CoDD's North Star: tell CoDD a desired change in plain words, it identifies the affected design docs via lexicon + semantic scoring, updates them with an LLM, and runs the DAG verify gate before any code is touched. Add `--dry-run` to preview, `--non-interactive` for CI.
+
 > Real-world: dogfooded against a Next.js + Prisma + PostgreSQL LMS. See [Case study](#-case-study-real-world-lms).
 
 ---
@@ -79,6 +87,7 @@ That's it. Three commands, three feedback loops, one coherent project.
 | 🔍 **`codd elicit`** | LLM finds **specification holes** in your requirements, scoped against industry-standard lexicons (BABOK, OWASP, WCAG, PCI DSS, ISO 25010, …). |
 | 🔄 **`codd diff`** | Detects **drift** between requirements and the actual implementation (brownfield-friendly). |
 | 🛠️ **`codd dag verify --auto-repair`** | Validates the requirements → design → implementation → tests DAG; an LLM proposes patches when violations appear and the loop retries until SUCCESS or MAX_ATTEMPTS. |
+| 🎯 **`codd fix`** / **`codd fix [PHENOMENON]`** | Two modes. *Legacy*: auto-detects test/CI failures, maps them to the DAG, asks an LLM to patch implementation, re-runs the verify gate. *PHENOMENON* (v2.16.0+, North Star entry-point 2): take a natural-language phenomenon (e.g. `"dashboard layout breaks on mobile"`), identify candidate design docs via Tier-1 lexicon + Tier-2 semantic scoring, update them with the LLM, then propagate. Full interactive HITL + `--non-interactive` + `--dry-run`. |
 | 📦 **38 lexicon plug-ins** | Industry standards bundled as opt-in coverage axes — Web (WCAG / OWASP / Web Vitals / WebAuthn / forms / SEO / PWA / browser-compat / responsive), Mobile (HIG / Material 3 / a11y / MASVS), Backend (REST / GraphQL / gRPC / events), Data (SQL / JSON Schema / event sourcing / governance), Ops (CI/CD / Kubernetes / Terraform / observability / DORA), Compliance (ISO 27001 / HIPAA / PCI DSS / GDPR / EU AI Act), Process (ISO 25010 / 29119 / DDD / 12-factor / i18n / model cards / API rate-limit), and Methodology (BABOK). |
 | 🌐 **`codd brownfield`** | Extract → diff → elicit pipeline: point CoDD at an existing codebase and it reverse-engineers requirements, finds drift, and surfaces gaps in one shot. |
 | 🎯 **`codd init --suggest-lexicons --llm-enhanced`** | LLM reads your code/docs, identifies data types and function traits, and recommends which lexicons to install (with confidence + reasoning). |
@@ -153,14 +162,16 @@ This is what lets CoDD ship one core that works for Next.js, Django, FastAPI, Ra
 
 ## 🧭 Roadmap
 
-- **v2.17.0 (current)** — `node_completeness` honours `kind: common` (cmd_470). Fixes a v2.15.0 oversight: `expects` edges pointing at common (shared infrastructure) nodes were misreported as missing impl files even when the file existed. 6 new tests, 2914 total PASS, SKIP=0. See [CHANGELOG](CHANGELOG.md).
+- **v2.18.0 (current)** — Greenfield triage from [@v-kato](https://github.com/v-kato) (cmd_473). Issue #20: `codd implement run --language` now overrides `project.language` per invocation (no more re-init when `codd init --language` was the wrong choice). Issue #21: `detailed_design` added to `DEFAULT_NODE_PREFIXES` so `codd plan --init` no longer emits node ids its own `codd validate` rejects. Issue #22: `_strip_code_fence` is non-greedy + non-anchored, dropping trailing markdown prose that LLMs occasionally appended after the closing fence. 12 new tests, 2937 total PASS, SKIP=0.
+- **v2.17.1** — emergency patch for `codd fix [PHENOMENON]` (cmd_471). Issue #23: `codd/fix/templates/*.txt` are now shipped inside the wheel (a hatch `include` regression caused `FileNotFoundError` after `pip install`). Issue #24: replaced `---` wrappers in `design_update.txt` / `risk_assessment.txt` with `<document>` / `<diff>` XML-style tags so the rendered prompt no longer collides with markdown frontmatter or unified-diff `--- a/path` lines. 11 new tests, 2925 total PASS, SKIP=0.
+- **v2.17.0** — `node_completeness` honours `kind: common` (cmd_470). Fixes a v2.15.0 oversight: `expects` edges pointing at common (shared infrastructure) nodes were misreported as missing impl files even when the file existed. 6 new tests, 2914 total PASS, SKIP=0.
 - **v2.16.0** — `codd fix [PHENOMENON]` — North Star entry-point 2 (cmd_468). Express a desired change in natural language; CoDD identifies affected design docs via Tier-1 lexicon + Tier-2 semantic scoring, updates them with LLM, runs the DAG verify gate. Full interactive HITL (candidate selection, ambiguity clarification, risk confirmation) with `--non-interactive` for CI. 66 new tests, 2908 total PASS, SKIP=0.
 - **v2.15.0** — `kind: common` for shared infrastructure (cmd_467). C5 amber −79.2% on dogfood project (125 → 26). `**` glob translator fix.
 - **v2.14.0** — 8 structural gaps closed (cmd_466 dogfood). Sidecar `<test>.codd.yaml` with `verified_by:` (C6) / `axis_matrix:` (C9); lexicon schema SSoT; completeness_audit batch; `scan.exclude` bug fix (−52%); `codd dag verify --auto-repair`; elicit mock-AI sentinel; AI timeout 3600 s SSoT. Red 22 → 0.
 - **v2.13.0** — Opt-out protection: `OptOutPolicy` requires `justification` + `expires_at`. Silent SKIP abolished; severity preserved.
 - **v2.12.0** — Test-completeness gates: C7 amber promotion + C8 `ci_health` static check.
 - **v2.11.0** — Sprint-less `codd implement` (`--design <path> --output <dir>` directly).
-- **v2.18.0 (next)** — impl/test auto-propagation from PHENOMENON (AC #8 completion); Codex wrapper for PHENOMENON mode.
+- **v2.19.0 (next)** — impl/test auto-propagation from PHENOMENON (AC #8 completion); Codex wrapper for PHENOMENON mode.
 
 ---
 

{codd_dev-2.17.0 → codd_dev-2.18.0}/README.md

@@ -28,6 +28,14 @@ codd dag verify --auto-repair --max-attempts 10  # AI fixes coherence violations
 
 That's it. Three commands, three feedback loops, one coherent project.
 
+### 💡 Already shipping? Describe what you want fixed.
+
+```bash
+codd fix "login error messages are hard to understand"   # natural-language phenomenon mode
+```
+
+`codd fix [PHENOMENON]` (v2.16.0+) is the **second entry-point** of CoDD's North Star: tell CoDD a desired change in plain words, it identifies the affected design docs via lexicon + semantic scoring, updates them with an LLM, and runs the DAG verify gate before any code is touched. Add `--dry-run` to preview, `--non-interactive` for CI.
+
 > Real-world: dogfooded against a Next.js + Prisma + PostgreSQL LMS. See [Case study](#-case-study-real-world-lms).
 
 ---
@@ -39,6 +47,7 @@ That's it. Three commands, three feedback loops, one coherent project.
 | 🔍 **`codd elicit`** | LLM finds **specification holes** in your requirements, scoped against industry-standard lexicons (BABOK, OWASP, WCAG, PCI DSS, ISO 25010, …). |
 | 🔄 **`codd diff`** | Detects **drift** between requirements and the actual implementation (brownfield-friendly). |
 | 🛠️ **`codd dag verify --auto-repair`** | Validates the requirements → design → implementation → tests DAG; an LLM proposes patches when violations appear and the loop retries until SUCCESS or MAX_ATTEMPTS. |
+| 🎯 **`codd fix`** / **`codd fix [PHENOMENON]`** | Two modes. *Legacy*: auto-detects test/CI failures, maps them to the DAG, asks an LLM to patch implementation, re-runs the verify gate. *PHENOMENON* (v2.16.0+, North Star entry-point 2): take a natural-language phenomenon (e.g. `"dashboard layout breaks on mobile"`), identify candidate design docs via Tier-1 lexicon + Tier-2 semantic scoring, update them with the LLM, then propagate. Full interactive HITL + `--non-interactive` + `--dry-run`. |
 | 📦 **38 lexicon plug-ins** | Industry standards bundled as opt-in coverage axes — Web (WCAG / OWASP / Web Vitals / WebAuthn / forms / SEO / PWA / browser-compat / responsive), Mobile (HIG / Material 3 / a11y / MASVS), Backend (REST / GraphQL / gRPC / events), Data (SQL / JSON Schema / event sourcing / governance), Ops (CI/CD / Kubernetes / Terraform / observability / DORA), Compliance (ISO 27001 / HIPAA / PCI DSS / GDPR / EU AI Act), Process (ISO 25010 / 29119 / DDD / 12-factor / i18n / model cards / API rate-limit), and Methodology (BABOK). |
 | 🌐 **`codd brownfield`** | Extract → diff → elicit pipeline: point CoDD at an existing codebase and it reverse-engineers requirements, finds drift, and surfaces gaps in one shot. |
 | 🎯 **`codd init --suggest-lexicons --llm-enhanced`** | LLM reads your code/docs, identifies data types and function traits, and recommends which lexicons to install (with confidence + reasoning). |
@@ -113,14 +122,16 @@ This is what lets CoDD ship one core that works for Next.js, Django, FastAPI, Ra
 
 ## 🧭 Roadmap
 
-- **v2.17.0 (current)** — `node_completeness` honours `kind: common` (cmd_470). Fixes a v2.15.0 oversight: `expects` edges pointing at common (shared infrastructure) nodes were misreported as missing impl files even when the file existed. 6 new tests, 2914 total PASS, SKIP=0. See [CHANGELOG](CHANGELOG.md).
+- **v2.18.0 (current)** — Greenfield triage from [@v-kato](https://github.com/v-kato) (cmd_473). Issue #20: `codd implement run --language` now overrides `project.language` per invocation (no more re-init when `codd init --language` was the wrong choice). Issue #21: `detailed_design` added to `DEFAULT_NODE_PREFIXES` so `codd plan --init` no longer emits node ids its own `codd validate` rejects. Issue #22: `_strip_code_fence` is non-greedy + non-anchored, dropping trailing markdown prose that LLMs occasionally appended after the closing fence. 12 new tests, 2937 total PASS, SKIP=0.
+- **v2.17.1** — emergency patch for `codd fix [PHENOMENON]` (cmd_471). Issue #23: `codd/fix/templates/*.txt` are now shipped inside the wheel (a hatch `include` regression caused `FileNotFoundError` after `pip install`). Issue #24: replaced `---` wrappers in `design_update.txt` / `risk_assessment.txt` with `<document>` / `<diff>` XML-style tags so the rendered prompt no longer collides with markdown frontmatter or unified-diff `--- a/path` lines. 11 new tests, 2925 total PASS, SKIP=0.
+- **v2.17.0** — `node_completeness` honours `kind: common` (cmd_470). Fixes a v2.15.0 oversight: `expects` edges pointing at common (shared infrastructure) nodes were misreported as missing impl files even when the file existed. 6 new tests, 2914 total PASS, SKIP=0.
 - **v2.16.0** — `codd fix [PHENOMENON]` — North Star entry-point 2 (cmd_468). Express a desired change in natural language; CoDD identifies affected design docs via Tier-1 lexicon + Tier-2 semantic scoring, updates them with LLM, runs the DAG verify gate. Full interactive HITL (candidate selection, ambiguity clarification, risk confirmation) with `--non-interactive` for CI. 66 new tests, 2908 total PASS, SKIP=0.
 - **v2.15.0** — `kind: common` for shared infrastructure (cmd_467). C5 amber −79.2% on dogfood project (125 → 26). `**` glob translator fix.
 - **v2.14.0** — 8 structural gaps closed (cmd_466 dogfood). Sidecar `<test>.codd.yaml` with `verified_by:` (C6) / `axis_matrix:` (C9); lexicon schema SSoT; completeness_audit batch; `scan.exclude` bug fix (−52%); `codd dag verify --auto-repair`; elicit mock-AI sentinel; AI timeout 3600 s SSoT. Red 22 → 0.
 - **v2.13.0** — Opt-out protection: `OptOutPolicy` requires `justification` + `expires_at`. Silent SKIP abolished; severity preserved.
 - **v2.12.0** — Test-completeness gates: C7 amber promotion + C8 `ci_health` static check.
 - **v2.11.0** — Sprint-less `codd implement` (`--design <path> --output <dir>` directly).
-- **v2.18.0 (next)** — impl/test auto-propagation from PHENOMENON (AC #8 completion); Codex wrapper for PHENOMENON mode.
+- **v2.19.0 (next)** — impl/test auto-propagation from PHENOMENON (AC #8 completion); Codex wrapper for PHENOMENON mode.
 
 ---
 

{codd_dev-2.17.0 → codd_dev-2.18.0}/codd/cli.py

@@ -2109,6 +2109,15 @@ def implement_augment_cmd(
     help="Seconds before one chunk is interrupted",
 )
 @click.option("--enable-typecheck-loop", is_flag=True, default=False, help="Run configured typecheck repair loop after implementation")
+@click.option(
+    "--language",
+    default=None,
+    help=(
+        "Override project.language for this invocation only. Useful when "
+        "`codd init --language js` was chosen but a downstream spec/design "
+        "doc requires TypeScript (Issue #20). codd.yaml is not modified."
+    ),
+)
 def implement_run_cmd(
     task_id: str | None,
     project_path: str,
@@ -2117,6 +2126,7 @@ def implement_run_cmd(
     chunk_size: int | None,
     timeout_per_chunk: int,
     enable_typecheck_loop: bool,
+    language: str | None,
 ):
     """Run implementation with optional derived step injection."""
     from codd.implementer import auto_detect_task, implement_tasks
@@ -2140,6 +2150,7 @@ def implement_run_cmd(
                 chunk_size=chunk_size,
                 timeout_per_chunk=timeout_per_chunk,
                 history=None,
+                language=language,
             )
         except (FileNotFoundError, ValueError, CoddCLIError) as exc:
             click.echo(f"Error: {exc}")
@@ -2165,6 +2176,7 @@ def implement_run_cmd(
             task=task_id,
             ai_command=ai_cmd,
             use_derived_steps=_optional_bool(use_derived_steps),
+            language=language,
         )
     except (FileNotFoundError, ValueError, CoddCLIError) as exc:
         click.echo(f"Error: {exc}")
@@ -2237,6 +2249,7 @@ def _run_chunked_implementation(
     chunk_size: int,
     timeout_per_chunk: int,
     history: str | None,
+    language: str | None = None,
 ):
     if not task_id:
         raise ValueError("--task is required when chunked execution is enabled")
@@ -2245,6 +2258,13 @@ def _run_chunked_implementation(
     from codd.implementer.chunked_runner import ChunkedRunner
 
     config = _load_optional_project_config(project_root)
+    if language:
+        # Per-invocation language override (Issue #20, v-kato): align with
+        # implement_tasks() so both code paths honour the CLI flag without
+        # mutating codd.yaml on disk.
+        project_cfg = dict(config.get("project") or {})
+        project_cfg["language"] = language
+        config = {**config, "project": project_cfg}
     task_item, steps = _chunked_task_and_steps(project_root, config, task_id)
     resolved_ai_command = generator_module._resolve_ai_command(config, ai_cmd, command_name="implement")
 

codd_dev-2.18.0/codd/fix/templates/clarification_question.txt

@@ -0,0 +1,34 @@
+You are CoDD's clarification interviewer. The user described a
+phenomenon, but the description is too ambiguous to act on safely.
+Generate one short clarification question and 2-4 candidate options,
+then a final free-text option.
+
+STRICT RULES:
+- Question and options MUST be derived from the lexicon and the
+  phenomenon analysis. Do NOT invent terms not present in either.
+- Each option is a single short sentence (<=80 chars).
+- The final option must always be "Other (free text)".
+- Do not propose options that contradict the project lexicon.
+- Output JSON only.
+
+Output shape:
+{
+  "question": "string",
+  "options": [
+    {"id": "a", "label": "string"},
+    {"id": "b", "label": "string"},
+    ...
+    {"id": "other", "label": "Other (free text)"}
+  ]
+}
+
+PHENOMENON_TEXT:
+{phenomenon_text}
+
+PHENOMENON_ANALYSIS (JSON):
+{analysis_json}
+
+PROJECT_LEXICON (may be empty):
+{lexicon_context}
+
+JSON OUTPUT:

codd_dev-2.18.0/codd/fix/templates/design_update.txt

@@ -0,0 +1,33 @@
+You are CoDD's design document updater. The user reported a phenomenon
+they want fixed. Update the target design document to reflect the
+required change.
+
+STRICT RULES:
+1. Preserve every existing user_journey, acceptance_criterion, persona,
+   and lexicon_term unless explicit removal was requested.
+2. Never touch the frontmatter `codd:` block (codd.* metadata is owned
+   by the tooling).
+3. Make the smallest cohesive edit that captures the new requirement.
+   Add a bullet, refine a sentence, append an acceptance criterion.
+4. Do NOT rewrite the whole document. No reorganization. No tone change.
+5. If the phenomenon does not actually require any design change,
+   return the document UNCHANGED (exact byte-for-byte).
+6. Do not delete content unless ALLOW_DELETE is "true".
+
+PHENOMENON_TEXT:
+{phenomenon_text}
+
+PHENOMENON_ANALYSIS (JSON):
+{analysis_json}
+
+ALLOW_DELETE: {allow_delete}
+
+CURRENT_DOCUMENT_PATH: {target_path}
+
+CURRENT_DOCUMENT_CONTENT:
+<document>
+{current_content}
+</document>
+
+Return ONLY the full updated document body (including frontmatter).
+No commentary, no code fences, no <document> tags, no diff markers.

codd_dev-2.18.0/codd/fix/templates/phenomenon_parse.txt

@@ -0,0 +1,34 @@
+You are CoDD's PHENOMENON parser. The user described a phenomenon they
+want fixed in their software. Extract a structured analysis.
+
+Return ONLY a JSON object with this exact shape:
+
+{
+  "intent": "improvement | bugfix | new_feature | clarification | unknown",
+  "subject_terms": ["natural-language keyword", "..."],
+  "lexicon_hits": ["normalized lexicon id", "..."],
+  "ambiguity_score": 0.0,
+  "acceptance_signal": "short natural-language description of done"
+}
+
+Rules:
+- intent: choose exactly one. "unknown" if not derivable.
+- subject_terms: 1-6 short noun-phrases that describe what the phenomenon
+  is about, in the user's own words. No metaphors, no boilerplate.
+- lexicon_hits: only include ids that appear in the supplied lexicon.
+  If no lexicon hit, return [].
+- ambiguity_score: 0.0 = totally clear and actionable.
+  1.0 = the phenomenon is so vague that any fix could be wrong.
+  Be honest. 0.4+ means a clarification question is warranted.
+- acceptance_signal: at most 140 chars. "" if cannot be inferred.
+
+PHENOMENON_TEXT:
+{phenomenon_text}
+
+PROJECT_LEXICON (may be empty):
+{lexicon_context}
+
+DESIGN_DOC_SUMMARIES (id → frontmatter description, may be empty):
+{design_summaries}
+
+JSON OUTPUT:

codd_dev-2.18.0/codd/fix/templates/risk_assessment.txt

@@ -0,0 +1,32 @@
+You are CoDD's risk classifier. Given a unified diff produced by an
+automated fix, decide whether it includes a destructive or otherwise
+risky change that warrants explicit user confirmation.
+
+Return ONLY a JSON object:
+
+{
+  "risky": true | false,
+  "categories": ["schema_migration", "dependency_add", "mass_deletion",
+                 "test_removal", "config_change"],
+  "summary": "one sentence explaining the risk, or '' if not risky"
+}
+
+Detection guidance:
+- schema_migration: prisma/migrations, alembic versions, *.sql files,
+  DDL keywords in the diff (ALTER, DROP, CREATE TABLE).
+- dependency_add: changes to pyproject.toml [project.dependencies],
+  package.json dependencies, requirements*.txt, Gemfile, Cargo.toml.
+- mass_deletion: removed lines outnumber added lines by 3x or more,
+  OR a whole file is removed.
+- test_removal: tests or spec files deleted, OR test cases removed
+  without replacement.
+- config_change: CI workflow, deploy config, security policy, env vars.
+
+DIFF:
+<diff>
+{diff_text}
+</diff>
+
+HEURISTIC_FLAGS (already detected): {heuristic_flags}
+
+JSON OUTPUT:

{codd_dev-2.17.0 → codd_dev-2.18.0}/codd/implementer.py

@@ -382,10 +382,19 @@ def implement_tasks(
     clean: bool = False,
     use_derived_steps: bool | None = None,
     task: str | None = None,
+    language: str | None = None,
     **_ignored: Any,
 ) -> list[ImplementationResult]:
     project_root = Path(project_root).resolve()
     config = _load_project_config(project_root)
+    if language:
+        # Per-invocation language override (Issue #20, v-kato): mismatched
+        # codd init --language doesn't force a full re-init; spec authors can
+        # ship an implement run with --language typescript and the project's
+        # codd.yaml stays untouched.
+        project_cfg = dict(config.get("project") or {})
+        project_cfg["language"] = language
+        config = {**config, "project": project_cfg}
     design_node = design or task
     if not design_node:
         raise ValueError("--design is required")
@@ -1382,7 +1391,11 @@ def _prepend_traceability_comment(relative_path: str, comment_block: str, conten
 
 def _strip_code_fence(block: str) -> str:
     stripped = block.strip()
-    fenced = re.match(r"^```(?:[a-zA-Z0-9_+-]+)?\s*\n(?P<body>.*)\n```$", stripped, re.DOTALL)
+    # Non-greedy `.*?` captures up to the FIRST closing fence; any trailing
+    # prose/markdown after the fence is discarded (Issue #22, v-kato).
+    # Drop the `$` end-of-string anchor so the match still wins when the
+    # LLM ignored the "no commentary" instruction and appended explanations.
+    fenced = re.match(r"^```(?:[a-zA-Z0-9_+-]+)?\s*\n(?P<body>.*?)\n```", stripped, re.DOTALL)
     if fenced:
         return fenced.group("body")
     return stripped

{codd_dev-2.17.0 → codd_dev-2.18.0}/codd/validator.py

@@ -24,6 +24,7 @@ DEFAULT_NODE_PREFIXES = {
     "design",
     "detail",
     "detailed",
+    "detailed_design",
     "doc",
     "endpoint",
     "file",

{codd_dev-2.17.0 → codd_dev-2.18.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "codd-dev"
-version = "2.17.0"
+version = "2.18.0"
 description = "CoDD: Coherence-Driven Development — cross-artifact change impact analysis"
 readme = "README.md"
 license = "MIT"
@@ -63,6 +63,7 @@ include = [
     "codd/**/*.py",
     "codd/**/*.md",
     "codd/**/*.yaml",
+    "codd/**/*.txt",
     "codd/hooks/recipes/*.json",
     "codd/hooks/recipes/*.sh",
     "codd/templates/**/*.j2",