Files
gcp_solution_architecture_a…/app/adk/fake_runtime.py
Jonathan Boniface f54af24169
Some checks failed
validation / verify (push) Failing after 10s
feat: a work breakdown structure with a computed timeline (iteration 10)
A new Deliverable in the packaging Phase, produced by a delivery planner
after validation.

- The planner writes the plan as data: work packages that name the Elements
  they deliver, and activities with days, what each follows, and whether the
  team or the customer owns it.
- Code checks the plan against the design: every Element belongs to a work
  package, and the activities can be scheduled.
- Code computes the schedule, with no more of the team's activities at once
  than the team has people, and the critical path.
- Code writes the WBS dictionary (a work package is done when the Checks on
  its Elements pass), the activity sequence and a timeline on a relative axis
  from Day 0.
- The plan is checked against the delivery window from the agreed scope; over
  it, the Deliverable says by how much and names the critical path.
- With a start date it also gives calendar dates and a Mermaid Gantt chart.
- The timeline is added to the solution guide, and the plan's length is in
  execution.json and the packaging Approval.

Not run with real models.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-11 19:54:17 +01:00

256 lines
8.4 KiB
Python

"""A scripted AgentRuntime for deterministic tests and evaluation dry runs.
Each role has a queue of turns; a turn is a list of tool calls, or a function of the
task that returns one. Calls go through the real ToolSpecs, so every rule the tools
enforce is exercised exactly as a model would meet it.
"""
from __future__ import annotations
import json
from collections import deque
from collections.abc import Callable, Mapping, Sequence
from dataclasses import dataclass, field
from typing import Any
from app.adk.runtime import (
AgentTask,
Answer,
BudgetExhausted,
Outcome,
Question,
RoleSpec,
Stopped,
ToolSpec,
)
from app.workflows.deliverables import DEPENDENCY_GRAPH, DeliverableKind
@dataclass(frozen=True)
class Call:
tool: str
args: Mapping[str, Any] = field(default_factory=dict)
Turn = Sequence[Call] | Callable[[AgentTask], Sequence[Call]]
# One small, coherent Requirements chain, so sample submissions pass every link.
SAMPLE_DATA: dict[str, str] = {
"scope.yaml": "engagement:\n kind: workload\nintent:\n statement: Accept events.\n",
"scope.json": json.dumps(
{
"kind": "workload",
"stated": {"intent": {"statement": "Accept events."}},
"to_decide": [],
"to_ask": [],
"not_applicable": [],
}
),
"requirements.json": json.dumps(
[
{"id": "FR-1", "kind": "functional", "text": "Accept events."},
{
"id": "NFR-1",
"kind": "non_functional",
"text": "99.9% availability.",
"conditions": ["The service stays available when one instance fails."],
},
]
),
"components.json": json.dumps(
[{"id": "C-1", "name": "Ingress", "requirements": ["FR-1", "NFR-1"]}]
),
"products.json": json.dumps([{"component": "C-1", "product": "Example Run"}]),
"recommendations.json": json.dumps(
[
{
"id": "DR-1",
"pillar": "security",
"text": "Use a dedicated identity.",
"components": ["C-1"],
"conditions": ["The service runs as an identity nothing else uses."],
}
]
),
"target-architecture.json": json.dumps(
{
"elements": [
{
"id": "E-1",
"name": "Ingress service",
"component": "C-1",
"product": "Example Run",
"resource_type": "example_run_service",
},
{
"id": "E-2",
"name": "Ingress identity",
"product": "Example IAM",
"resource_type": "example_service_account",
},
],
"relationships": [
{"id": "R-1", "from": "E-1", "to": "E-2", "kind": "runs as"}
],
"checks": [
{
"id": "K-1",
"applies_to": "E-1",
"statement": "The service runs at least two instances.",
"source": "NFR-1",
},
{
"id": "K-2",
"applies_to": "E-2",
"statement": "The identity is used by this service only.",
"source": "DR-1",
},
],
}
),
"implementation.json": json.dumps(
[
{"check": "K-1", "file": "terraform/main.tf", "address": "example.run"},
{"check": "K-2", "file": "terraform/main.tf", "address": "example.sa"},
]
),
"work-breakdown.json": json.dumps(
{
"work_packages": [
{
"id": "WP-1",
"name": "Ingress",
"parent": None,
"elements": ["E-1", "E-2"],
"scope": "The service and its identity.",
"produces": ["a running service"],
"skills": ["platform engineering"],
}
],
"activities": [
{
"id": "A-1",
"work_package": "WP-1",
"name": "Create the identity",
"days": 1,
"after": [],
},
{
"id": "A-2",
"work_package": "WP-1",
"name": "Deploy the service",
"days": 2,
"after": ["A-1"],
},
],
}
),
"verdicts.json": json.dumps(
{
"dry_run": {"verdict": "pass", "evidence": "terraform validate passed."},
"checks": [
{"check": "K-1", "verdict": "pass", "evidence": "min instances is 2."},
{"check": "K-2", "verdict": "pass", "evidence": "One binding."},
],
"proposed_additions": [],
}
),
}
def sample_submission(kind: DeliverableKind, note: str = "") -> Call:
"""A `submit_deliverable` call that meets the Deliverable's contract."""
spec = DEPENDENCY_GRAPH[kind]
parts = [f"# {kind}", note or "Summary."]
parts += [f"## {section}\n{note or 'Details.'}" for section in spec.sections]
attachments = {
name: SAMPLE_DATA.get(name, f"# {name}\n") for name in spec.attachments
}
return Call(
"submit_deliverable",
{"content": "\n".join(parts) + "\n", "attachments": attachments},
)
@dataclass(frozen=True)
class RunRecord:
role: str
task: AgentTask
tools: tuple[str, ...]
results: tuple[tuple[str, str], ...] # (tool name, message)
outcome: Outcome
class ScriptExhausted(AssertionError):
"""A role was run more times than the test scripted."""
class FakeAgentRuntime:
def __init__(self, turns: Mapping[str, Sequence[Turn]]) -> None:
self._turns = {role: deque(role_turns) for role, role_turns in turns.items()}
self.runs: list[RunRecord] = []
def remaining(self, role: str) -> int:
return len(self._turns.get(role, ()))
async def run(
self, role: RoleSpec, task: AgentTask, tools: Sequence[ToolSpec]
) -> Outcome:
queue = self._turns.get(role.name)
if not queue:
raise ScriptExhausted(f"No scripted turn left for role {role.name!r}")
turn = queue.popleft()
calls = turn(task) if callable(turn) else turn
by_name = {tool.name: tool for tool in tools}
results: list[tuple[str, str]] = []
outcome: Outcome = Stopped(
"The script ended without a tool that ends the turn."
)
for index, call in enumerate(calls):
if index >= role.budget:
outcome = BudgetExhausted(
f"{role.name} used all {role.budget} tool calls before finishing."
)
break
if call.tool not in by_name:
raise ValueError(
f"{role.name} called {call.tool!r}, which it was not offered: "
f"{sorted(by_name)}"
)
result = await by_name[call.tool].handler(**call.args)
results.append((call.tool, result.message))
if result.outcome is not None:
outcome = result.outcome
break
self.runs.append(
RunRecord(role.name, task, tuple(by_name), tuple(results), outcome)
)
return outcome
class FakeDecider:
"""A scripted Decider: answers by question id, or declines so callers fall back.
`answers` maps a question id to an Answer, or to a function of (state, question)
returning one. An id with no entry makes the whole decision unavailable.
"""
def __init__(self, answers: Mapping[str, Any] | None = None) -> None:
self.answers = dict(answers or {})
self.asked: list[tuple[str, dict[str, Question], str]] = []
async def decide(
self, state: str, questions: Mapping[str, Question], *, scope: str = ""
) -> Mapping[str, Answer] | None:
self.asked.append((state, dict(questions), scope))
decided = {}
for name, question in questions.items():
if name not in self.answers:
return None
answer = self.answers[name]
decided[name] = answer(state, question) if callable(answer) else answer
return decided