feat(ralph): smart gates + DAG + dashboard live (W3)

Restructurare Ralph QC loop pe smart gate dispatcher tag-driven (în loc de
5 faze fixe), DAG dependsOn cu propagare blocked, retry guard 3-strike, rate
limit detection, plus dashboard live cu polling 5s.

Changes:
- tools/ralph_prd_generator.py: parametru optional final_plan_path; când e
  furnizat, invocă Claude Opus pe final-plan.md pentru extragere user stories
  cu schema extinsă (tags, dependsOn, acceptanceCriteria 3-5). Backward compat
  păstrat — fără final_plan_path, fallback la heuristic-ul vechi.
- tools/ralph/prd-template.json: schema W3 (tags[], dependsOn[], retries,
  failed, blocked, failureReason, requiresDesignReview).
- tools/ralph/prompt.md: 4 faze (impl, base quality, smart gates, commit) +
  dispatcher pe story.tags. Tags vide → run-all-gates fallback (safe default).
- tools/ralph_dag.py (nou): tag validation heuristic anti-silent-regression
  (force ui dacă diff atinge .vue/.tsx/.html/.css/.scss; force db pentru
  migrations sau .sql; force vercel dacă există vercel.json) + topological
  sort cu blocked propagation + atomic prd.json updates.
- tools/ralph/ralph.sh: --max-turns 30, DAG-aware story selection, retry
  counter cu auto-fail la 3, rate limit detection (sleep 30min + 1 retry),
  CLI subcommands prin tools/ralph_dag.py helper.
- dashboard/handlers/ralph.py (nou): /api/ralph/status + /<slug>/log + /prd
  + /stop. Defensive vs corrupt prd.json. Sandbox-ed PID kill.
- dashboard/ralph.html (nou): live cards 3/2/1 col responsive, polling 5s,
  drawer pentru log/PRD viewer, status colors (--status-running/blocked/
  failed/complete declarate inline), Lucide icons cu aria-labels.
- dashboard/api.py: mount /api/ralph/* (GET status/log/prd, POST stop).
- tests/: 72 teste noi (smart gates, DAG, retry, dashboard endpoint).

Note arhitecturale:
- Polling 5s ales peste SSE/WebSocket (suficient pentru iter Ralph 8-15min)
- Tag validation rulează POST-iter pe diff git pentru anti-silent-regression
- Rate limit retry: 1 dată per rulare, apoi mark failed=rate_limited

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-04-26 18:36:35 +00:00
parent e06a79d98c
commit 655ed3ae09
11 changed files with 2282 additions and 189 deletions

250
tests/test_dag_execution.py Normal file
View File

@@ -0,0 +1,250 @@
"""Tests for W3 DAG execution + retry guard.
Acoperă:
- topological_eligible: alegere story DAG-aware (passes/failed/blocked propagation)
- cmd_incr_retry: 3-retry guard cu auto-fail la max_retries
- cmd_mark_failed: propagare blocked la dependenți
- _normalize_story: validează schema extinsă (tags, dependsOn, retries, blocked)
"""
from __future__ import annotations
import json
import sys
from pathlib import Path
import pytest
PROJECT_ROOT = Path(__file__).resolve().parents[1]
if str(PROJECT_ROOT) not in sys.path:
sys.path.insert(0, str(PROJECT_ROOT))
from tools import ralph_dag # noqa: E402
from tools.ralph_prd_generator import _normalize_story # noqa: E402
# ── topological_eligible ───────────────────────────────────────
def _stories(*specs):
"""Helper: build minimal stories list. Each spec = (id, priority, dependsOn, flags...)"""
out = []
for spec in specs:
sid, prio, deps, *rest = spec
s = {
"id": sid,
"priority": prio,
"dependsOn": list(deps),
"passes": False,
"failed": False,
"blocked": False,
"retries": 0,
}
for flag in rest:
s[flag] = True
out.append(s)
return out
class TestTopologicalEligible:
def test_no_deps_lowest_priority_picked(self):
stories = _stories(
("US-002", 20, []),
("US-001", 10, []),
)
chosen = ralph_dag.topological_eligible(stories)
assert chosen["id"] == "US-001"
def test_dependent_skipped_until_dep_passes(self):
stories = _stories(
("US-001", 10, []),
("US-002", 20, ["US-001"]),
)
# US-001 not done yet → US-001 picked
assert ralph_dag.topological_eligible(stories)["id"] == "US-001"
# Mark US-001 passes → US-002 eligible
stories[0]["passes"] = True
assert ralph_dag.topological_eligible(stories)["id"] == "US-002"
def test_failed_dep_propagates_blocked(self):
stories = _stories(
("US-001", 10, []),
("US-002", 20, ["US-001"]),
)
stories[0]["failed"] = True
chosen = ralph_dag.topological_eligible(stories)
assert chosen is None # US-002 marcat blocked, nimic eligibil
assert stories[1]["blocked"] is True
assert stories[1]["failureReason"] == "blocked_by:US-001"
def test_independent_runs_when_other_chain_failed(self):
# US-001 failed → US-002 blocked, dar US-003 e independent → eligibil
stories = _stories(
("US-001", 10, []),
("US-002", 20, ["US-001"]),
("US-003", 30, []),
)
stories[0]["failed"] = True
chosen = ralph_dag.topological_eligible(stories)
assert chosen["id"] == "US-003"
assert stories[1]["blocked"] is True
def test_chain_blocking_propagates_transitively(self):
# US-001 → US-002 → US-003. US-001 failed → US-002 blocked → US-003 blocked.
stories = _stories(
("US-001", 10, []),
("US-002", 20, ["US-001"]),
("US-003", 30, ["US-002"]),
)
stories[0]["failed"] = True
chosen = ralph_dag.topological_eligible(stories)
assert chosen is None
assert stories[1]["blocked"] is True
assert stories[2]["blocked"] is True
def test_all_complete_returns_none(self):
stories = _stories(("US-001", 10, []))
stories[0]["passes"] = True
assert ralph_dag.topological_eligible(stories) is None
def test_already_blocked_story_skipped(self):
stories = _stories(
("US-001", 10, []),
("US-002", 20, []),
)
stories[0]["blocked"] = True
chosen = ralph_dag.topological_eligible(stories)
assert chosen["id"] == "US-002"
# ── cmd_incr_retry / cmd_mark_failed (file-based) ──────────────
@pytest.fixture
def prd_path(tmp_path):
"""Construiește un prd.json minimal pentru test."""
data = {
"projectName": "test-proj",
"branchName": "ralph/test-proj",
"userStories": [
{
"id": "US-001", "title": "first", "priority": 10,
"dependsOn": [], "tags": [], "acceptanceCriteria": ["a"],
"passes": False, "failed": False, "blocked": False, "retries": 0,
},
{
"id": "US-002", "title": "second", "priority": 20,
"dependsOn": ["US-001"], "tags": [], "acceptanceCriteria": ["a"],
"passes": False, "failed": False, "blocked": False, "retries": 0,
},
],
}
p = tmp_path / "prd.json"
p.write_text(json.dumps(data), encoding="utf-8")
return p
class TestRetryGuard:
def test_incr_retry_increments_count(self, prd_path):
rc = ralph_dag.cmd_incr_retry(prd_path, "US-001")
assert rc == 0
data = json.loads(prd_path.read_text())
assert data["userStories"][0]["retries"] == 1
assert data["userStories"][0]["failed"] is False
def test_three_retries_marks_failed_max_retries(self, prd_path):
# incr 3 times
for _ in range(3):
ralph_dag.cmd_incr_retry(prd_path, "US-001")
data = json.loads(prd_path.read_text())
assert data["userStories"][0]["retries"] == 3
assert data["userStories"][0]["failed"] is True
assert data["userStories"][0]["failureReason"] == "max_retries"
def test_max_retries_propagates_blocked_to_dependent(self, prd_path):
for _ in range(3):
ralph_dag.cmd_incr_retry(prd_path, "US-001")
data = json.loads(prd_path.read_text())
# US-002 depinde de US-001 → blocked
assert data["userStories"][1]["blocked"] is True
def test_unknown_story_returns_error(self, prd_path):
rc = ralph_dag.cmd_incr_retry(prd_path, "US-999")
assert rc == 1
class TestMarkFailed:
def test_mark_failed_sets_flags(self, prd_path):
rc = ralph_dag.cmd_mark_failed(prd_path, "US-001", "rate_limited")
assert rc == 0
data = json.loads(prd_path.read_text())
assert data["userStories"][0]["failed"] is True
assert data["userStories"][0]["failureReason"] == "rate_limited"
def test_mark_failed_propagates_blocked(self, prd_path):
ralph_dag.cmd_mark_failed(prd_path, "US-001", "rate_limited")
data = json.loads(prd_path.read_text())
assert data["userStories"][1]["blocked"] is True
assert data["userStories"][1]["failureReason"] == "blocked_by:US-001"
class TestNextStory:
def test_next_story_prints_id(self, prd_path, capsys):
rc = ralph_dag.cmd_next_story(prd_path)
assert rc == 0
captured = capsys.readouterr()
assert captured.out.strip() == "US-001"
def test_next_story_returns_1_when_none_eligible(self, prd_path, capsys):
# Mark all complete
data = json.loads(prd_path.read_text())
for s in data["userStories"]:
s["passes"] = True
prd_path.write_text(json.dumps(data))
rc = ralph_dag.cmd_next_story(prd_path)
assert rc == 1
# ── _normalize_story (PRD generator schema) ────────────────────
class TestNormalizeStory:
def test_default_fields_populated(self):
s = _normalize_story({"title": "x"}, idx=0)
# Schema W3 fields trebuie să existe toate
for key in ("id", "title", "description", "priority", "acceptanceCriteria",
"tags", "dependsOn", "passes", "failed", "blocked", "retries",
"failureReason", "notes"):
assert key in s, f"Missing schema field: {key}"
assert s["passes"] is False
assert s["failed"] is False
assert s["blocked"] is False
assert s["retries"] == 0
def test_invalid_tags_filtered(self):
s = _normalize_story({"title": "x", "tags": ["frontend", "ui", "made-up"]}, idx=0)
assert s["tags"] == ["ui"] # frontend & made-up nu sunt în VALID_TAGS
def test_empty_acceptance_gets_default(self):
s = _normalize_story({"title": "x"}, idx=0)
assert len(s["acceptanceCriteria"]) >= 1
def test_ui_tag_implies_browser_check(self):
s = _normalize_story({"title": "x", "tags": ["ui"]}, idx=0)
assert s["requiresBrowserCheck"] is True
def test_explicit_browser_check_preserved(self):
s = _normalize_story({"title": "x", "tags": [], "requiresBrowserCheck": True}, idx=0)
assert s["requiresBrowserCheck"] is True
def test_id_auto_generated_from_idx(self):
s = _normalize_story({"title": "x"}, idx=4)
assert s["id"] == "US-005"
def test_id_preserved_when_provided(self):
s = _normalize_story({"id": "US-042", "title": "x"}, idx=0)
assert s["id"] == "US-042"
def test_depends_on_preserves_strings_filters_garbage(self):
s = _normalize_story({"title": "x", "dependsOn": ["US-001", "", None, "US-002"]}, idx=0)
assert s["dependsOn"] == ["US-001", "US-002"]

View File

@@ -0,0 +1,209 @@
"""Tests for /api/ralph/* endpoints (dashboard live).
Acoperă:
- /api/ralph/status: list cards cu state + count + fetchedAt
- /api/ralph/<slug>/log: tail progress.txt
- /api/ralph/<slug>/prd: full prd.json
- _ralph_validate_slug: path traversal protection
- corrupt prd.json: graceful error (status='error', not 500)
"""
from __future__ import annotations
import json
import sys
from pathlib import Path
import pytest
PROJECT_ROOT = Path(__file__).resolve().parents[1]
DASH = PROJECT_ROOT / "dashboard"
if str(DASH) not in sys.path:
sys.path.insert(0, str(DASH))
@pytest.fixture(scope="module")
def ralph_module():
from handlers import ralph as _r # type: ignore
return _r
@pytest.fixture
def handler(ralph_module, tmp_path, monkeypatch):
"""Build a stubbed handler with a temp WORKSPACE_DIR."""
import constants # type: ignore
# Re-route WORKSPACE_DIR la tmp pentru izolare
monkeypatch.setattr(constants, "WORKSPACE_DIR", tmp_path)
class _Stub(ralph_module.RalphHandlers):
def __init__(self):
self.captured = None
self.captured_code = None
self.path = "/api/ralph/status"
def send_json(self, data, code=200):
self.captured = data
self.captured_code = code
def send_error(self, code):
self.captured = {"error_code": code}
self.captured_code = code
return _Stub()
def _make_ralph_project(workspace: Path, slug: str, stories: list, progress: str = "init"):
"""Create a fake ralph project under workspace/<slug>/scripts/ralph/."""
ralph_dir = workspace / slug / "scripts" / "ralph"
ralph_dir.mkdir(parents=True, exist_ok=True)
(ralph_dir / "prd.json").write_text(json.dumps({
"projectName": slug,
"branchName": f"ralph/{slug}",
"userStories": stories,
}), encoding="utf-8")
(ralph_dir / "progress.txt").write_text(progress, encoding="utf-8")
return ralph_dir
# ── /api/ralph/status ──────────────────────────────────────────
class TestStatus:
def test_empty_workspace_returns_empty(self, handler):
handler.handle_ralph_status()
assert handler.captured_code == 200
assert handler.captured["projects"] == []
assert "fetchedAt" in handler.captured
def test_status_skips_non_ralph_projects(self, handler, tmp_path):
# Create a project WITHOUT scripts/ralph
(tmp_path / "regular-proj").mkdir()
handler.handle_ralph_status()
assert handler.captured["projects"] == []
def test_status_lists_ralph_projects(self, handler, tmp_path):
_make_ralph_project(tmp_path, "proj-a", [
{"id": "US-001", "title": "a", "priority": 10, "passes": True,
"failed": False, "blocked": False, "retries": 0, "tags": []},
{"id": "US-002", "title": "b", "priority": 20, "passes": False,
"failed": False, "blocked": False, "retries": 0, "tags": ["ui"]},
])
handler.handle_ralph_status()
projects = handler.captured["projects"]
assert len(projects) == 1
p = projects[0]
assert p["slug"] == "proj-a"
assert p["storiesTotal"] == 2
assert p["storiesComplete"] == 1
assert p["storiesFailed"] == 0
assert p["storiesBlocked"] == 0
assert p["status"] == "idle" # not running (no .ralph.pid)
def test_status_corrupt_prd_returns_error_not_500(self, handler, tmp_path):
ralph_dir = tmp_path / "broken" / "scripts" / "ralph"
ralph_dir.mkdir(parents=True)
(ralph_dir / "prd.json").write_text("{not valid json", encoding="utf-8")
handler.handle_ralph_status()
assert handler.captured_code == 200
assert any(p.get("status") == "error" for p in handler.captured["projects"])
def test_status_count_matches_projects(self, handler, tmp_path):
_make_ralph_project(tmp_path, "p1", [])
_make_ralph_project(tmp_path, "p2", [])
handler.handle_ralph_status()
assert handler.captured["count"] == 2
def test_complete_status_when_all_pass(self, handler, tmp_path):
_make_ralph_project(tmp_path, "donezo", [
{"id": "US-001", "passes": True, "failed": False, "blocked": False,
"retries": 0, "tags": [], "title": "x", "priority": 10},
])
handler.handle_ralph_status()
p = handler.captured["projects"][0]
assert p["status"] == "complete"
def test_failed_status_propagation(self, handler, tmp_path):
_make_ralph_project(tmp_path, "broken-proj", [
{"id": "US-001", "passes": False, "failed": True, "blocked": False,
"retries": 3, "tags": [], "title": "x", "priority": 10,
"failureReason": "max_retries"},
])
handler.handle_ralph_status()
p = handler.captured["projects"][0]
assert p["status"] == "failed"
assert p["storiesFailed"] == 1
# ── /api/ralph/<slug>/log ──────────────────────────────────────
class TestLog:
def test_log_returns_progress_lines(self, handler, tmp_path):
_make_ralph_project(tmp_path, "p1", [], progress="line1\nline2\nline3")
handler.path = "/api/ralph/p1/log"
handler.handle_ralph_log("p1")
assert handler.captured_code == 200
assert handler.captured["lines"] == ["line1", "line2", "line3"]
assert handler.captured["total"] == 3
def test_log_invalid_slug_400(self, handler):
handler.handle_ralph_log("../etc/passwd")
assert handler.captured_code == 400
def test_log_path_traversal_blocked(self, handler):
handler.handle_ralph_log("..")
assert handler.captured_code == 400
def test_log_missing_progress_returns_empty(self, handler, tmp_path):
ralph_dir = tmp_path / "noprogress" / "scripts" / "ralph"
ralph_dir.mkdir(parents=True)
(ralph_dir / "prd.json").write_text("{}") # no progress.txt
handler.path = "/api/ralph/noprogress/log"
handler.handle_ralph_log("noprogress")
assert handler.captured_code == 200
assert handler.captured["lines"] == []
# ── /api/ralph/<slug>/prd ──────────────────────────────────────
class TestPrd:
def test_prd_returns_full_json(self, handler, tmp_path):
stories = [{"id": "US-001", "passes": False, "title": "t", "priority": 10}]
_make_ralph_project(tmp_path, "p1", stories)
handler.handle_ralph_prd("p1")
assert handler.captured_code == 200
assert handler.captured["projectName"] == "p1"
assert len(handler.captured["userStories"]) == 1
def test_prd_404_when_missing(self, handler, tmp_path):
(tmp_path / "ghost").mkdir() # exists, but no prd.json
handler.handle_ralph_prd("ghost")
assert handler.captured_code == 404
def test_prd_invalid_slug_400(self, handler):
handler.handle_ralph_prd("/etc/passwd")
assert handler.captured_code == 400
# ── _ralph_validate_slug ───────────────────────────────────────
class TestValidateSlug:
def test_valid_slug_returns_path(self, handler, tmp_path):
(tmp_path / "good-slug").mkdir()
result = handler._ralph_validate_slug("good-slug")
assert result is not None
assert result.name == "good-slug"
def test_slash_rejected(self, handler):
assert handler._ralph_validate_slug("a/b") is None
def test_dotdot_rejected(self, handler):
assert handler._ralph_validate_slug("..") is None
def test_empty_rejected(self, handler):
assert handler._ralph_validate_slug("") is None
def test_nonexistent_returns_none(self, handler):
assert handler._ralph_validate_slug("does-not-exist") is None

174
tests/test_smart_gates.py Normal file
View File

@@ -0,0 +1,174 @@
"""Tests for W3 smart gates tag validation heuristic.
Acoperă:
- infer_tags_from_paths: detect ui/db/vercel pe baza file extensions / paths
- force_include_tags: combinare tags Opus + tags inferate din diff (anti-silent-regression)
- Toate combinatii de tag types (ui, db, vercel, refactor, docs, backend, infra)
- Edge cases: tags vide, tags invalide, empty diff
"""
from __future__ import annotations
import sys
from pathlib import Path
import pytest
PROJECT_ROOT = Path(__file__).resolve().parents[1]
if str(PROJECT_ROOT) not in sys.path:
sys.path.insert(0, str(PROJECT_ROOT))
from tools.ralph_dag import ( # noqa: E402
VALID_TAGS,
force_include_tags,
infer_tags_from_paths,
)
# ── infer_tags_from_paths ──────────────────────────────────────
class TestInferTags:
def test_empty_diff_no_tags(self):
assert infer_tags_from_paths([]) == []
def test_only_readme_no_tags(self):
assert infer_tags_from_paths(["README.md", "CHANGELOG.md"]) == []
def test_vue_triggers_ui(self):
assert infer_tags_from_paths(["src/App.vue"]) == ["ui"]
def test_tsx_triggers_ui(self):
assert infer_tags_from_paths(["app/page.tsx"]) == ["ui"]
def test_jsx_triggers_ui(self):
assert infer_tags_from_paths(["src/Button.jsx"]) == ["ui"]
def test_html_triggers_ui(self):
assert infer_tags_from_paths(["dashboard/index.html"]) == ["ui"]
def test_css_scss_trigger_ui(self):
assert infer_tags_from_paths(["src/main.css"]) == ["ui"]
assert infer_tags_from_paths(["src/main.scss"]) == ["ui"]
def test_svelte_triggers_ui(self):
assert infer_tags_from_paths(["src/App.svelte"]) == ["ui"]
def test_migrations_triggers_db(self):
assert infer_tags_from_paths(["db/migrations/0001_init.sql"]) == ["db"]
def test_top_level_migrations_triggers_db(self):
assert infer_tags_from_paths(["migrations/2026/04/add_users.sql"]) == ["db"]
def test_sql_outside_migrations_still_triggers_db(self):
assert infer_tags_from_paths(["scripts/seed.sql"]) == ["db"]
def test_vercel_json_only(self):
assert infer_tags_from_paths([], has_vercel_json=True) == ["vercel"]
def test_combined_ui_db_vercel(self):
result = infer_tags_from_paths(
["app/page.tsx", "db/migrations/0001.sql"], has_vercel_json=True
)
assert result == ["ui", "db", "vercel"]
def test_dedup_when_multiple_files_same_category(self):
result = infer_tags_from_paths(["a.tsx", "b.vue", "c.css"])
assert result == ["ui"]
def test_case_insensitive_extensions(self):
assert infer_tags_from_paths(["src/App.TSX"]) == ["ui"]
assert infer_tags_from_paths(["db/Init.SQL"]) == ["db"]
# ── force_include_tags ─────────────────────────────────────────
class TestForceIncludeTags:
def test_existing_only_no_diff(self):
assert force_include_tags(["backend"], [], False) == ["backend"]
def test_diff_inferred_added_to_existing(self):
# Opus marcat docs, dar diff atinge .tsx → ui forțat
result = force_include_tags(["docs"], ["src/Page.tsx"], False)
assert "docs" in result
assert "ui" in result
def test_filters_invalid_tags_from_existing(self):
# Tag-ul "frontend" nu e în VALID_TAGS — trebuie eliminat
result = force_include_tags(["frontend", "ui"], [], False)
assert "frontend" not in result
assert "ui" in result
def test_empty_when_no_existing_no_diff(self):
assert force_include_tags([], [], False) == []
def test_dedup_existing_and_inferred(self):
# Existing are ui, diff are .tsx → un singur ui în output
result = force_include_tags(["ui"], ["src/A.tsx"], False)
assert result.count("ui") == 1
def test_vercel_added_when_vercel_json_present(self):
result = force_include_tags(["backend"], [], has_vercel_json=True)
assert "vercel" in result
assert "backend" in result
def test_all_valid_tags_preserved(self):
# Verifică că force_include nu strică tags valide existente
all_valid = list(VALID_TAGS)
result = force_include_tags(all_valid, [], False)
for t in all_valid:
assert t in result
def test_order_existing_first_then_inferred(self):
# Existing tags trebuie să apară primele (stabilitate API)
result = force_include_tags(["backend"], ["src/Page.tsx", "db/migrations/0001.sql"], False)
assert result[0] == "backend"
assert "ui" in result and "db" in result
# ── Smart gates dispatcher contract (combinatii tag → expected gates) ─────────
# Acesta e un table-test pentru contractul dispatcher-ului din prompt.md.
# Verifică doar mapping-ul tag → gate name (specifice prompt.md), nu execuția.
GATE_MAPPING = {
"refactor": "/workflow:simplify",
"ui": "/qa",
"vercel": "gh pr checks",
"db": "schema diff",
"docs": None, # docs => doar typecheck base
"backend": "/review",
"infra": "/review",
}
class TestGateMapping:
"""Validează că prompt.md menționează gate-urile așteptate per tag."""
@pytest.fixture(scope="class")
def prompt_md(self):
path = PROJECT_ROOT / "tools" / "ralph" / "prompt.md"
return path.read_text(encoding="utf-8")
def test_refactor_gate_documented(self, prompt_md):
assert "/workflow:simplify" in prompt_md
def test_ui_gate_documented(self, prompt_md):
assert "/qa" in prompt_md
assert "agent-browser" in prompt_md.lower()
def test_vercel_gate_documented(self, prompt_md):
assert "gh pr checks" in prompt_md
def test_db_gate_documented(self, prompt_md):
assert "schema diff" in prompt_md.lower() or "alembic" in prompt_md.lower()
def test_backend_gate_documented(self, prompt_md):
assert "/review" in prompt_md
def test_run_all_fallback_documented(self, prompt_md):
# Tags vide → run-all-gates fallback (safe default)
assert "tags vide" in prompt_md.lower() or "run-all-gates" in prompt_md.lower()
def test_dag_dependson_documented(self, prompt_md):
assert "dependsOn" in prompt_md or "DAG" in prompt_md