Files
gcp_solution_architecture_a…/app/a2a_server.py
Jonathan Boniface ef37f59bad feat: opt-in published Terraform modules from approved sources (iteration 8)
Infrastructure code is plain resources by default. An Execution may opt in
with terraform_modules; the deployment default is TERRAFORM_MODULES, off.

- A module is accepted only from the registry namespaces the provider bundle
  lists (terraform-google-modules), at one exact version. Any other source is
  refused before Terraform runs; until now nothing stopped generated code
  naming any source for terraform init to download.
- terraform_module looks up a module's inputs, outputs and submodules from
  the registry and says when it cannot be used with the pinned provider.
- skills/google-cloud-iac-modules: when a module fits, and four modules with
  examples a test validates against hashicorp/google 8.6.0.
- Opted in, the IaC author and validator get the Skill and the lookup. Not
  opted in, they get neither and a module block is refused in code.
- A Change request follows its Baseline. execution.json records the choice.
- case-07 runs case 03's request with modules opted into.

Azure (aztfm) and AWS (terraform-aws-modules) sources are recorded in the
plan for when those Cloud providers are added. No case has run with this.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 22:09:29 +01:00

200 lines
8.1 KiB
Python

"""Maps A2A tasks onto Executions: a task is an Execution, its context is the Session.
A new task starts an Execution; a message on an input-required task answers the
Execution's prompt. Approvals and permissions count only as structured data parts,
never as free text.
Before a new task starts anything its message is read (app/nodes/intake.py): a
question is answered and starts nothing. A message that reads as a change to
delivered work also starts nothing: the caller is told which Execution it would
change and how to name it as the Baseline, because a Change request never starts
from earlier work on a guess (ADR-0006).
"""
from __future__ import annotations
import logging
from collections.abc import Mapping
from typing import Any
from a2a.helpers.proto_helpers import (
get_data_parts,
new_data_part,
new_task,
new_text_part,
)
from a2a.server.agent_execution import AgentExecutor, RequestContext
from a2a.server.events import EventQueue
from a2a.server.tasks import TaskUpdater
from a2a.types import TaskState
from google.protobuf import json_format
from app.agent import ExecutionEngine, Interaction
from app.nodes.execution_nodes import GUIDE_FILENAME
from app.nodes.intake import ABOUT, CHANGE_REQUEST, QUESTION, Intake, change_note
from app.presentation import describe_prompt
from app.workflows.deliverables import DeliverableKind
logger = logging.getLogger(__name__)
STRUCTURED_REPLY_KEYS = ("approval", "permission")
class SolutionArchitectExecutor(AgentExecutor):
def __init__(self, engine: ExecutionEngine, intake: Intake | None = None) -> None:
self._engine = engine
self._intake = intake
async def execute(self, context: RequestContext, event_queue: EventQueue) -> None:
task_id = context.task_id or ""
updater = TaskUpdater(event_queue, task_id, context.context_id or "")
new = context.current_task is None
try:
if new:
history = [context.message] if context.message is not None else []
await event_queue.enqueue_event(
new_task(
task_id,
context.context_id or "",
TaskState.TASK_STATE_SUBMITTED,
history=history,
)
)
await updater.start_work()
options = _options(context)
# A pre-approved caller is a batch job: it sends work, not questions.
if (
self._intake
and not options.get("pre_approved")
and not options.get("baseline")
and not options.get("new_request")
):
reading = await self._intake.read(
context.get_user_input(),
options.get("customer_case"),
scope=task_id,
)
if reading.kind == QUESTION:
await updater.complete(
updater.new_agent_message(
[
new_text_part(ABOUT),
new_data_part(
{"status": "answered", "intake": QUESTION}
),
]
)
)
return
if reading.kind == CHANGE_REQUEST and reading.latest:
await updater.complete(
updater.new_agent_message(
[
new_text_part(change_note(reading.latest)),
new_data_part(
{
"status": "answered",
"intake": CHANGE_REQUEST,
"baseline": reading.latest,
}
),
]
)
)
return
interaction = await self._engine.start(
context.get_user_input(),
session_id=context.context_id,
customer_case=options.get("customer_case"),
baseline=options.get("baseline"),
overlay_skills=options.get("overlay_skills") or (),
terraform_modules=options.get("terraform_modules"),
execution_id=task_id,
pre_approved=bool(options.get("pre_approved", False)),
allow_cloud_read=options.get("allow_cloud_read"),
)
else:
await updater.start_work()
interaction = await self._engine.resume(task_id, _reply(context))
except Exception as exc:
logger.exception("Execution %s failed", task_id)
await updater.failed(
updater.new_agent_message(
[
new_text_part(
f"The Execution failed: {type(exc).__name__}: {exc}"
)
]
)
)
return
await self._publish(interaction, updater)
async def cancel(self, context: RequestContext, event_queue: EventQueue) -> None:
# The checkpoint stays; cancelling only ends the A2A task.
updater = TaskUpdater(
event_queue, context.task_id or "", context.context_id or ""
)
await updater.cancel()
async def _publish(self, interaction: Interaction, updater: TaskUpdater) -> None:
if interaction.status == "awaiting_user" and interaction.prompt is not None:
prompt = dict(interaction.prompt)
await updater.requires_input(
updater.new_agent_message(
[
new_text_part(describe_prompt(prompt)),
new_data_part({"prompt": prompt}),
]
)
)
return
deliverables: Mapping[str, Mapping[str, Any]] = interaction.state.get(
"deliverables", {}
)
guide = deliverables.get(DeliverableKind.SOLUTION_GUIDE.value)
if guide:
await updater.add_artifact(
[new_text_part(guide["content"], media_type="text/markdown")],
name=GUIDE_FILENAME,
)
for kind, deliverable in deliverables.items():
for name, content in deliverable.get("attachments", {}).items():
await updater.add_artifact(
[new_text_part(content)], name=name, metadata={"deliverable": kind}
)
warnings = list(interaction.state.get("warnings", []))
summary = (
"The solution architecture is complete and approved."
if interaction.status == "completed"
else "The Execution ended with warnings: " + "; ".join(warnings)
)
await updater.complete(
updater.new_agent_message(
[
new_text_part(summary),
new_data_part({"status": interaction.status, "warnings": warnings}),
]
)
)
def _options(context: RequestContext) -> dict[str, Any]:
options: dict[str, Any] = {}
if context.message is not None and context.message.HasField("metadata"):
options.update(json_format.MessageToDict(context.message.metadata))
options.update(context.metadata)
return options
def _reply(context: RequestContext) -> dict[str, Any]:
if context.message is not None:
for data in get_data_parts(list(context.message.parts)):
if isinstance(data, Mapping) and any(
k in data for k in STRUCTURED_REPLY_KEYS
):
return {k: data[k] for k in STRUCTURED_REPLY_KEYS if k in data}
return {"reply": context.get_user_input()}