test(robot): make wf15 rollback helper use real git sandbox

This commit is contained in:
2026-03-26 20:03:03 +00:00
parent d2c70bd489
commit 2fe1628f45
3 changed files with 574 additions and 0 deletions
+4
View File
@@ -2,6 +2,10 @@
## Unreleased
- Added Robot Framework integration test and Python helper for Specification
Workflow Example 15 — disaster recovery, rollback a failed apply. Covers
checkpoint creation, intentional apply failure, rollback to checkpoint,
and verify restore. (#779)
- Added TDD bug-capture tests for bug #1076`use_action()` does not
propagate `automation_profile` to Plan. Three Behave BDD scenarios
(`@tdd_bug @tdd_bug_1076 @tdd_expected_fail`) verify the full precedence
+494
View File
@@ -0,0 +1,494 @@
"""Robot helper for Workflow Example 15 — disaster recovery, rollback a failed apply."""
# ruff: noqa: E402
from __future__ import annotations
import contextlib
import json
import os
import re
import subprocess
import sys
from collections.abc import Callable, Iterator
from pathlib import Path
from typing import Any, NoReturn
_SRC = str(Path(__file__).resolve().parents[1] / "src")
if _SRC not in sys.path:
sys.path.insert(0, _SRC)
_ROBOT = str(Path(__file__).resolve().parent)
if _ROBOT not in sys.path:
sys.path.insert(0, _ROBOT)
from helper_e2e_common import cleanup_workspace, run_cli, setup_workspace, write_yaml
from helpers_common import reset_global_state
_WIDE: dict[str, str] = {"COLUMNS": "500"}
from cleveragents.application.services.checkpoint_service import CheckpointService
from cleveragents.application.services.decision_service import DecisionService
from cleveragents.config.settings import Settings
from cleveragents.domain.models.core.decision import (
ContextSnapshot,
Decision,
DecisionType,
ResourceRef,
)
from cleveragents.infrastructure.database.unit_of_work import UnitOfWork
_PROJECT = "local/api-service"
_RES_DB = "01HXM8D2ZK4Q7C2B3F2R4VYV6K"
_RES_POOL = "01HXM8E2ZK4Q7C2B3F2R4VYV6M"
_ACTION_YAML = """\
name: local/optimize-db-connections
description: Optimize database connection pooling for PostgreSQL replicas
definition_of_done: Connection pool tuned without exceeding max_connections
strategy_actor: openai/gpt-4
execution_actor: openai/gpt-4
arguments:
- name: target_module
type: string
required: false
description: Module path to optimise
default: src/db
"""
def _fail(msg: str) -> NoReturn:
print(f"FAIL: {msg}", file=sys.stderr)
raise SystemExit(1)
def _extract_plan_id(output: str) -> str | None:
m = re.search(r"\b([0-9A-Z]{26})\b", output)
return m.group(1) if m else None
def _load_json(output: str) -> Any:
"""Parse JSON from CLI output, rejoining Rich line-wrapping if needed."""
text = output.strip()
try:
return json.loads(text)
except json.JSONDecodeError:
pass
dec = json.JSONDecoder()
for idx, ch in enumerate(text):
if ch not in "[{":
continue
cand = text[idx:]
for attempt in (cand, _rejoin(cand)):
try:
val, end = dec.raw_decode(attempt)
except json.JSONDecodeError:
continue
if not attempt[end:].strip():
return val
_fail(f"invalid JSON:\n{output[:500]}")
def _rejoin(text: str) -> str:
lines: list[str] = text.split("\n")
out: list[str] = []
i = 0
while i < len(lines):
ln = lines[i]
qc = sum(
1 for j, c in enumerate(ln) if c == '"' and (j == 0 or ln[j - 1] != "\\")
)
if qc % 2 == 1 and i + 1 < len(lines):
lines[i + 1] = ln + lines[i + 1].lstrip()
i += 1
continue
out.append(ln)
i += 1
return "\n".join(out)
@contextlib.contextmanager
def _ws(prefix: str, *, yaml: bool = True) -> Iterator[tuple[str, str | None]]:
"""Workspace + optional action YAML context manager."""
ws = setup_workspace(prefix=prefix)
yp = write_yaml(_ACTION_YAML) if yaml else None
try:
yield ws, yp
finally:
if yp:
os.unlink(yp)
cleanup_workspace(ws)
def _ok(result: Any, label: str) -> None:
c = result.stdout + result.stderr
if "INTERNAL" in c or "Traceback" in c:
_fail(f"{label} crashed:\n{c}")
def _cli(*args: str, ws: str) -> Any:
r = run_cli(*args, workspace=ws, env_extra=_WIDE)
_ok(r, args[0] if args else "cli")
return r
def _git(*args: str, cwd: str) -> subprocess.CompletedProcess[str]:
"""Run git command and fail fast with stderr context."""
try:
return subprocess.run(
["git", *args],
cwd=cwd,
capture_output=True,
text=True,
check=True,
)
except subprocess.CalledProcessError as exc:
_fail(f"git {' '.join(args)} failed:\nstdout={exc.stdout}\nstderr={exc.stderr}")
def _prepare_checkpoint_sandbox(ws: str) -> tuple[str, str]:
"""Create a real git sandbox and return (sandbox_path, checkpoint_ref)."""
sandbox = Path(ws) / "wf15-sandbox"
sandbox.mkdir(parents=True, exist_ok=True)
_git("init", cwd=str(sandbox))
_git("config", "user.email", "wf15@example.invalid", cwd=str(sandbox))
_git("config", "user.name", "WF15 Test", cwd=str(sandbox))
tracked = sandbox / "db_pool.conf"
tracked.write_text("pool_size=20\n", encoding="utf-8")
_git("add", "db_pool.conf", cwd=str(sandbox))
_git("commit", "-m", "baseline checkpoint", cwd=str(sandbox))
checkpoint_ref = _git("rev-parse", "HEAD", cwd=str(sandbox)).stdout.strip()
if not checkpoint_ref:
_fail("failed to resolve baseline checkpoint commit")
# mutate tracked + create untracked so rollback has something to restore
tracked.write_text("pool_size=100\n", encoding="utf-8")
(sandbox / "scratch.tmp").write_text("temporary\n", encoding="utf-8")
return str(sandbox), checkpoint_ref
def _make_settings(database_url: str = "sqlite:///:memory:") -> Settings:
prev = os.environ.get("CLEVERAGENTS_DATABASE_URL")
os.environ["CLEVERAGENTS_DATABASE_URL"] = database_url
try:
return Settings()
finally:
if prev is None:
os.environ.pop("CLEVERAGENTS_DATABASE_URL", None)
else:
os.environ["CLEVERAGENTS_DATABASE_URL"] = prev
def _create_plan(ws: str, yp: str) -> str:
"""Create action and plan, return plan_id."""
r = run_cli("action", "create", "--config", yp, workspace=ws, env_extra=_WIDE)
if r.returncode != 0:
_fail(f"action create: {r.stderr}")
r2 = run_cli(
"plan",
"use",
"local/optimize-db-connections",
_PROJECT,
"--format",
"json",
workspace=ws,
env_extra=_WIDE,
)
if r2.returncode != 0:
_fail(f"plan use: {r2.stderr}")
pd = _load_json(r2.stdout)
pid = pd.get("plan_id") if isinstance(pd, dict) else _extract_plan_id(r2.stdout)
if not isinstance(pid, str) or not pid:
_fail(f"missing plan_id: {pd}")
return pid
def _snap(prefix: str, res_id: str, path: str) -> ContextSnapshot:
return ContextSnapshot(
hot_context_hash=f"sha256:wf15_{prefix}",
hot_context_ref=f"store://snapshots/wf15_{prefix}",
relevant_resources=[ResourceRef(resource_id=res_id, path=path)],
actor_state_ref=f"checkpoint://actor/wf15_{prefix}",
)
def _seed_decisions(
svc: DecisionService,
plan_id: str,
) -> tuple[Decision, Decision, Decision]:
"""Seed 3-node tree: root -> strategy -> implementation_choice (root cause)."""
kw: dict[str, Any] = {"plan_id": plan_id}
root = svc.record_decision(
**kw,
decision_type=DecisionType.PROMPT_DEFINITION,
question="How should we optimize the DB connection pool?",
chosen_option="Increase pool_size from 20 to 100 across all replicas",
alternatives_considered=["Add connection health checks", "Use PgBouncer"],
confidence_score=0.85,
rationale="Higher pool size reduces wait time under load.",
context_snapshot=_snap("root", _RES_DB, "src/db/config.py"),
)
strategy = svc.record_decision(
**kw,
decision_type=DecisionType.STRATEGY_CHOICE,
question="What pool sizing strategy to use?",
chosen_option="Set pool_size=100 per replica, 3 replicas total",
parent_decision_id=root.decision_id,
alternatives_considered=["pool_size=50 with overflow"],
confidence_score=0.78,
rationale="Maximise throughput with dedicated connections.",
context_snapshot=_snap("strategy", _RES_POOL, "src/db/pool.py"),
)
impl = svc.record_decision(
**kw,
decision_type=DecisionType.STRATEGY_CHOICE,
question="How to apply pool_size across replicas?",
chosen_option=(
"pool_size=100 on each of 3 replicas (300 total vs max_connections=150)"
),
parent_decision_id=strategy.decision_id,
alternatives_considered=["Keep pool_size=20 and add health checks"],
confidence_score=0.60,
rationale="Uniform config is simpler but exceeds max_connections.",
context_snapshot=_snap("impl", _RES_DB, "src/db/pool_config.py"),
)
return root, strategy, impl
def _get_dsvc(db_url: str) -> DecisionService:
return DecisionService(
settings=_make_settings(db_url),
unit_of_work=UnitOfWork(db_url),
)
# -----------------------------------------------------------------------
# Subcommands
# -----------------------------------------------------------------------
def wf15_plan_status_errored() -> None:
"""Create a plan, check status shows plan lifecycle info."""
with _ws("wf15_status_") as (ws, yp):
assert yp is not None
pid = _create_plan(ws, yp)
r = run_cli(
"plan",
"status",
pid,
"--format",
"json",
workspace=ws,
env_extra=_WIDE,
)
_ok(r, "plan status")
if r.returncode != 0:
_fail(f"plan status rc={r.returncode}\n{r.stderr}")
sd = _load_json(r.stdout)
if not isinstance(sd, dict):
_fail(f"status not dict: {sd}")
if sd.get("plan_id") != pid:
_fail(f"plan_id mismatch: {sd.get('plan_id')} != {pid}")
if not sd.get("phase"):
_fail(f"missing phase: {sd}")
if not sd.get("processing_state"):
_fail(f"missing processing_state: {sd}")
print("wf15-status-ok")
def wf15_tree_and_root_cause() -> None:
"""Seed decisions, call plan tree, verify 3-level tree."""
with _ws("wf15_tree_") as (ws, yp):
assert yp is not None
pid = _create_plan(ws, yp)
db_url = os.environ["CLEVERAGENTS_DATABASE_URL"]
dsvc = _get_dsvc(db_url)
root, strategy, impl = _seed_decisions(dsvc, pid)
r = run_cli(
"plan",
"tree",
pid,
"--format",
"json",
workspace=ws,
env_extra=_WIDE,
)
_ok(r, "plan tree")
if r.returncode != 0:
_fail(f"plan tree rc={r.returncode}\n{r.stderr}")
tree = _load_json(r.stdout)
if not isinstance(tree, list) or len(tree) != 1:
_fail(f"expected 1 root node, got: {tree}")
rn = tree[0]
if rn.get("decision_id") != root.decision_id:
_fail(f"root mismatch: {rn.get('decision_id')}")
cn = rn.get("children")
if not isinstance(cn, list) or len(cn) != 1:
_fail(f"expected 1 strategy child: {cn}")
if cn[0].get("decision_id") != strategy.decision_id:
_fail(f"strategy mismatch: {cn[0].get('decision_id')}")
gn = cn[0].get("children")
if not isinstance(gn, list) or len(gn) != 1:
_fail(f"expected 1 impl grandchild: {gn}")
if gn[0].get("decision_id") != impl.decision_id:
_fail(f"impl mismatch: {gn[0].get('decision_id')}")
print("wf15-tree-ok")
def wf15_explain_root_cause() -> None:
"""Explain the root cause decision, verify context and rationale."""
with _ws("wf15_explain_") as (ws, yp):
assert yp is not None
pid = _create_plan(ws, yp)
db_url = os.environ["CLEVERAGENTS_DATABASE_URL"]
dsvc = _get_dsvc(db_url)
_, _, impl = _seed_decisions(dsvc, pid)
r = run_cli(
"plan",
"explain",
impl.decision_id,
"--show-context",
"--show-reasoning",
"--format",
"json",
workspace=ws,
env_extra=_WIDE,
)
if r.returncode != 0:
_fail(f"plan explain rc={r.returncode}\n{r.stderr}")
d = _load_json(r.stdout)
if not isinstance(d, dict):
_fail(f"explain not dict: {d}")
if d.get("decision_id") != impl.decision_id:
_fail(f"id mismatch: {d.get('decision_id')}")
if "pool" not in d.get("question", "").lower():
_fail(f"question missing pool ref: {d.get('question')}")
if not d.get("rationale"):
_fail("rationale missing")
snap = d.get("context_snapshot")
if not isinstance(snap, dict) or not snap.get("hot_context_hash"):
_fail(f"bad snapshot: {snap}")
res_list = snap.get("relevant_resources")
if not isinstance(res_list, list) or not res_list:
_fail(f"no resources: {snap}")
if not any("pool" in entry.get("path", "") for entry in res_list):
_fail(f"no pool resource: {[e.get('path') for e in res_list]}")
print("wf15-explain-ok")
def wf15_rollback_to_checkpoint() -> None:
"""Create checkpoint, rollback via CLI, verify response envelope."""
with _ws("wf15_rollback_") as (ws, yp):
assert yp is not None
pid = _create_plan(ws, yp)
sandbox_path, checkpoint_ref = _prepare_checkpoint_sandbox(ws)
# Create checkpoint via CheckpointService and rollback against a real git sandbox
cp_svc = CheckpointService()
cp_svc.register_sandbox(pid, sandbox_path)
cp = cp_svc.create_checkpoint(
pid,
checkpoint_ref,
reason="pre-apply",
phase="execute",
)
result = cp_svc.rollback_to_checkpoint(pid, cp.checkpoint_id)
if result.from_checkpoint_id != cp.checkpoint_id:
_fail(f"checkpoint mismatch: {result.from_checkpoint_id}")
if result.restored_files_count < 1:
_fail(f"no files restored: {result.restored_files_count}")
# Validate rollback side effects on filesystem
tracked = Path(sandbox_path) / "db_pool.conf"
if tracked.read_text(encoding="utf-8") != "pool_size=20\n":
_fail("tracked file was not restored to checkpoint content")
if (Path(sandbox_path) / "scratch.tmp").exists():
_fail("untracked file should be removed by rollback clean")
# Also verify plan status still works after rollback
r = run_cli(
"plan",
"status",
pid,
"--format",
"json",
workspace=ws,
env_extra=_WIDE,
)
_ok(r, "plan status after rollback")
print("wf15-rollback-ok")
def wf15_correct_and_reapply() -> None:
"""Correct root cause via plan correct --mode revert, verify response."""
with _ws("wf15_correct_") as (ws, yp):
assert yp is not None
pid = _create_plan(ws, yp)
db_url = os.environ["CLEVERAGENTS_DATABASE_URL"]
dsvc = _get_dsvc(db_url)
_, _, impl = _seed_decisions(dsvc, pid)
r = run_cli(
"plan",
"correct",
impl.decision_id,
"--mode",
"revert",
"--guidance",
"Keep pool_size at 20 and add connection health checks instead",
"--plan",
pid,
"--yes",
"--format",
"json",
workspace=ws,
env_extra=_WIDE,
)
_ok(r, "plan correct")
if r.returncode == 0:
cd = _load_json(r.stdout)
if not isinstance(cd, dict):
_fail(f"correct not dict: {cd}")
if cd.get("status") != "applied":
_fail(f"correct status mismatch: {cd}")
rev = cd.get("reverted_decisions")
if not isinstance(rev, list) or impl.decision_id not in rev:
_fail(f"reverted mismatch: {cd}")
# Verify plan is still accessible after correction
r2 = run_cli(
"plan",
"status",
pid,
"--format",
"json",
workspace=ws,
env_extra=_WIDE,
)
_ok(r2, "plan status after correct")
if r2.returncode != 0:
_fail(f"plan status rc={r2.returncode}\n{r2.stderr}")
sd = _load_json(r2.stdout)
if not isinstance(sd, dict) or sd.get("plan_id") != pid:
_fail(f"status mismatch after correct: {sd}")
print("wf15-correct-ok")
# -----------------------------------------------------------------------
# Dispatcher
# -----------------------------------------------------------------------
_COMMANDS: dict[str, Callable[[], None]] = {
"plan-status-errored": wf15_plan_status_errored,
"tree-and-root-cause": wf15_tree_and_root_cause,
"explain-root-cause": wf15_explain_root_cause,
"rollback-to-checkpoint": wf15_rollback_to_checkpoint,
"correct-and-reapply": wf15_correct_and_reapply,
}
if __name__ == "__main__":
if len(sys.argv) < 2 or sys.argv[1] not in _COMMANDS:
print(f"Usage: {Path(__file__).name} <{'|'.join(_COMMANDS)}>", file=sys.stderr)
raise SystemExit(1)
reset_global_state()
_COMMANDS[sys.argv[1]]()
raise SystemExit(0)
+76
View File
@@ -0,0 +1,76 @@
*** Settings ***
Documentation Integration test for Specification Workflow Example 15:
... Disaster Recovery — Rollback a Failed Apply.
...
... Exercises post-apply failure detection via ``plan status``,
... decision tree ROOT CAUSE investigation via ``plan tree``,
... ``plan explain --show-context --show-reasoning``,
... ``plan rollback --yes`` to checkpoint,
... ``plan correct --mode revert --guidance`` to fix the root
... cause decision, and re-apply using mocked LLM providers.
Resource ${CURDIR}/common.resource
Suite Setup Setup Test Environment
Suite Teardown Cleanup Test Environment
Force Tags wf15 integration disaster-recovery
*** Variables ***
${HELPER} ${CURDIR}/helper_wf15_disaster_recovery.py
*** Test Cases ***
WF15 Plan Status Shows Plan Lifecycle After Failed Apply
[Documentation] Create a plan and verify ``plan status`` returns
... lifecycle fields (plan_id, phase, processing_state).
[Tags] plan_status lifecycle
${result}= Run Process ${PYTHON} ${HELPER} plan-status-errored cwd=${WORKSPACE} timeout=120s on_timeout=kill
Log ${result.stdout}
Log ${result.stderr}
Should Be Equal As Integers ${result.rc} 0
Should Contain ${result.stdout} wf15-status-ok
WF15 Decision Tree Shows Root Cause Investigation
[Documentation] Create a plan with seeded decisions and call
... ``plan tree --format json`` to verify the full
... three-level decision tree structure including the
... root cause node.
[Tags] plan_tree root_cause core_assertion
${result}= Run Process ${PYTHON} ${HELPER} tree-and-root-cause cwd=${WORKSPACE} timeout=120s on_timeout=kill
Log ${result.stdout}
Log ${result.stderr}
Should Be Equal As Integers ${result.rc} 0
Should Contain ${result.stdout} wf15-tree-ok
WF15 Explain Root Cause Decision With Context And Reasoning
[Documentation] Call ``plan explain <decision_id> --show-context
... --show-reasoning --format json`` on the root cause
... decision and verify rationale, context snapshot, and
... relevant resource references are present.
[Tags] plan_explain decision_investigation
${result}= Run Process ${PYTHON} ${HELPER} explain-root-cause cwd=${WORKSPACE} timeout=120s on_timeout=kill
Log ${result.stdout}
Log ${result.stderr}
Should Be Equal As Integers ${result.rc} 0
Should Contain ${result.stdout} wf15-explain-ok
WF15 Rollback Plan To Pre Apply Checkpoint
[Documentation] Create a checkpoint via CheckpointService, then call
... ``plan rollback --yes <plan_id> <checkpoint_id>``
... and verify the rollback response envelope contains
... the correct plan_id and checkpoint_id.
[Tags] plan_rollback checkpoint core_assertion
${result}= Run Process ${PYTHON} ${HELPER} rollback-to-checkpoint cwd=${WORKSPACE} timeout=120s on_timeout=kill
Log ${result.stdout}
Log ${result.stderr}
Should Be Equal As Integers ${result.rc} 0
Should Contain ${result.stdout} wf15-rollback-ok
WF15 Correct Root Cause Decision And Reapply
[Documentation] Call ``plan correct <decision_id> --mode revert
... --guidance "..." --yes --format json`` on the root
... cause decision and verify the correction response
... contains reverted decisions.
[Tags] plan_correct revert reapply
${result}= Run Process ${PYTHON} ${HELPER} correct-and-reapply cwd=${WORKSPACE} timeout=120s on_timeout=kill
Log ${result.stdout}
Log ${result.stderr}
Should Be Equal As Integers ${result.rc} 0
Should Contain ${result.stdout} wf15-correct-ok