274 lines
12 KiB
Python
274 lines
12 KiB
Python
"""Hermes chat config contracts."""
|
|
|
|
from __future__ import annotations
|
|
|
|
from test_hermes_chat_support import (
|
|
HERMES,
|
|
ROOT,
|
|
_documents,
|
|
yaml,
|
|
)
|
|
|
|
|
|
def test_chat_config_enables_real_research_compute_and_delegation():
|
|
configmap = _documents(HERMES / "chat-configmap.yaml")[0]
|
|
config = yaml.safe_load(configmap["data"]["config.yaml"])
|
|
|
|
# AUTO leaves effort unset so the target chosen by Switchyard owns it.
|
|
assert "reasoning_effort" not in config["agent"]
|
|
assert config["model"]["model"] == "atlas/auto/fast"
|
|
assert config["web"] == {
|
|
"backend": "ddgs",
|
|
"search_backend": "ddgs",
|
|
"extract_backend": "public-extract",
|
|
}
|
|
assert config["delegation"]["max_concurrent_children"] == 2
|
|
assert config["delegation"]["max_iterations"] == 80
|
|
assert config["agent"]["max_turns"] == 120
|
|
assert config["tool_loop_guardrails"]["hard_stop_enabled"] is True
|
|
for platform in ("cli", "api_server"):
|
|
toolsets = config["platform_toolsets"][platform]
|
|
assert "delegation" in toolsets
|
|
assert "browser" in toolsets
|
|
assert "python_sandbox" in toolsets
|
|
assert "vision" in toolsets
|
|
assert "web" in toolsets
|
|
assert "image_gen" in toolsets
|
|
assert "terminal" not in toolsets
|
|
assert "code_execution" not in toolsets
|
|
|
|
def test_agent_config_keeps_delegated_reviewers_from_owning_task_lifecycle():
|
|
configmap = _documents(HERMES / "agent-configmap.yaml")[0]
|
|
instructions = configmap["data"]["AGENTS.md"]
|
|
soul = configmap["data"]["SOUL.md"]
|
|
config = yaml.safe_load(configmap["data"]["config.yaml"])
|
|
|
|
assert (
|
|
"Only the foreground durable worker owns its Kanban task lifecycle"
|
|
in instructions
|
|
)
|
|
assert "Never unblock and then complete" in instructions
|
|
assert "gateway dispatcher can claim the transient" in instructions
|
|
for platform in ("cli", "api_server"):
|
|
assert "kanban" in config["platform_toolsets"][platform]
|
|
assert "must never complete, block, unblock, reclaim" in instructions
|
|
assert "task's final structured result itself" in instructions
|
|
assert "Atlas organization has private visibility" in instructions
|
|
assert "may be public or private" in instructions
|
|
assert "do not infer a\nrepository's visibility" in instructions
|
|
assert "already supplied through `GIT_ASKPASS`" in instructions
|
|
assert "Never call `kanban_show` without a known, non-empty task ID" in instructions
|
|
assert "bounded ad-hoc inspection and acceptance checks may" in instructions
|
|
assert "load implementation or TDD skills" in instructions
|
|
assert "Never call `kanban_show` without\na known, non-empty task ID" in soul
|
|
assert "must load a skill only when its workflow\nmaterially applies" in soul
|
|
assert "runtime-only `GIT_ASKPASS`" in soul
|
|
assert "`scm.bstein.dev` is Gitea, not GitHub" in instructions
|
|
assert "Never load or follow a GitHub/`gh`" in instructions
|
|
assert "/opt/coordinator/gitea_api.py METHOD /api/v1/..." in instructions
|
|
assert "JENKINS_BASE_URL" in instructions
|
|
assert "Do not\nuse `git reset --hard`" in instructions
|
|
rendered = (HERMES / "agent-deployment.yaml").read_text()
|
|
assert 'git config --global user.name "Hermes Agent"' in rendered
|
|
assert 'git config --global user.email "hermes@bstein.dev"' in rendered
|
|
|
|
def test_agent_image_completes_parked_kanban_tasks_atomically():
|
|
dockerfile = (ROOT / "dockerfiles" / "Dockerfile.hermes-agent").read_text()
|
|
|
|
assert (
|
|
"AND status IN ('running', 'ready', 'blocked', 'scheduled')"
|
|
in dockerfile
|
|
)
|
|
assert (
|
|
"Atomically mark running, ready, blocked, or scheduled tasks done"
|
|
in dockerfile
|
|
)
|
|
assert "kind IN ('created', 'blocked', 'unblocked')" in dockerfile
|
|
assert "payload.get(\"status\") == \"blocked\"" in dockerfile
|
|
assert "hermes-kanban-blocked-regression.py" in dockerfile
|
|
|
|
def test_sandbox_shares_only_the_tenant_workspace_without_credentials():
|
|
sandbox_docs = _documents(HERMES / "chat-sandbox.yaml")
|
|
deployments = [doc for doc in sandbox_docs if doc["kind"] == "Deployment"]
|
|
assert len(deployments) == 8
|
|
for ordinal, deployment in enumerate(deployments):
|
|
pod_spec = deployment["spec"]["template"]["spec"]
|
|
container = pod_spec["containers"][0]
|
|
assert pod_spec["automountServiceAccountToken"] is False
|
|
assert container["securityContext"]["readOnlyRootFilesystem"] is True
|
|
assert container["securityContext"]["runAsNonRoot"] is True
|
|
assert container["securityContext"]["runAsGroup"] == 10000
|
|
assert not container.get("env")
|
|
assert {mount["mountPath"] for mount in container["volumeMounts"]} == {
|
|
"/tmp",
|
|
"/workspace",
|
|
"/opt/data/workspace",
|
|
}
|
|
workspace_volume = next(
|
|
item for item in pod_spec["volumes"] if item["name"] == "workspace"
|
|
)
|
|
assert workspace_volume["persistentVolumeClaim"]["claimName"] == (
|
|
f"workspace-hermes-chat-tenant-{ordinal}"
|
|
)
|
|
|
|
statefulset = _documents(HERMES / "chat-statefulset.yaml")[0]
|
|
templates = statefulset["spec"]["volumeClaimTemplates"]
|
|
workspace = next(item for item in templates if item["metadata"]["name"] == "workspace")
|
|
assert workspace["spec"]["resources"]["requests"]["storage"] == "10Gi"
|
|
assert workspace["spec"]["accessModes"] == ["ReadWriteMany"]
|
|
|
|
pod_spec = statefulset["spec"]["template"]["spec"]
|
|
hermes = next(item for item in pod_spec["containers"] if item["name"] == "hermes")
|
|
startup = hermes["args"][0]
|
|
assert "hermes-chat-sandbox-${ordinal}.hermes-chat-sandbox" in startup
|
|
assert any(
|
|
mount["name"] == "workspace" and mount["mountPath"] == "/opt/data/workspace"
|
|
for mount in hermes["volumeMounts"]
|
|
)
|
|
|
|
policies = _documents(HERMES / "networkpolicy.yaml")
|
|
deny = next(
|
|
item
|
|
for item in policies
|
|
if item["metadata"]["name"] == "hermes-chat-sandbox-deny"
|
|
)
|
|
assert deny["spec"]["ingress"] == []
|
|
assert deny["spec"]["egress"] == []
|
|
for ordinal in range(4):
|
|
policy = next(
|
|
item
|
|
for item in policies
|
|
if item["metadata"]["name"] == f"hermes-chat-sandbox-tenant-{ordinal}"
|
|
)
|
|
assert policy["spec"]["podSelector"]["matchLabels"][
|
|
"ai.bstein.dev/tenant-ordinal"
|
|
] == str(ordinal)
|
|
source = policy["spec"]["ingress"][0]["from"][0]["podSelector"][
|
|
"matchLabels"
|
|
]
|
|
assert source["statefulset.kubernetes.io/pod-name"] == (
|
|
f"hermes-chat-tenant-{ordinal}"
|
|
)
|
|
|
|
def test_gateway_image_honors_ui_model_and_caps_reasoning():
|
|
dockerfile = (ROOT / "dockerfiles" / "Dockerfile.hermes-agent").read_text()
|
|
assert (
|
|
"FROM nousresearch/hermes-agent@sha256:"
|
|
"47d4bd4cc420b70e40ed75efdade373e45b86b7382d4013a054208982bb6ba08"
|
|
) in dockerfile
|
|
assert "_resolve_request_route" in dockerfile
|
|
assert 'allowed_providers = {"atlas-switchyard"}' in dockerfile
|
|
assert 'reasoning_effort=body.get("reasoning_effort")' in dockerfile
|
|
assert 'reasoning_config = {"enabled": True, "effort": "xhigh"}' in dockerfile
|
|
assert "ddgs==9.14.4" in dockerfile
|
|
assert "specific not in _LEGACY_WEB_BACKENDS" in dockerfile
|
|
assert '"pre_internal_route"' in dockerfile
|
|
assert "pre_internal_route hook failed" in dockerfile
|
|
assert '"pre_subagent_route"' in dockerfile
|
|
assert "pre_subagent_route hook failed" in dockerfile
|
|
assert "agent._hermes_explicit_model_pick = bool" in dockerfile
|
|
assert "agent._hermes_oneshot = True" in dockerfile
|
|
assert 'getattr(parent_agent, "_hermes_oneshot", False)' in dockerfile
|
|
assert 'var != "CLAUDE_CODE_OAUTH_TOKEN"' in dockerfile
|
|
assert '"source": "managed_claude_code_subscription"' in dockerfile
|
|
assert '"name": "Claude Code subscription (Switchyard)"' in dockerfile
|
|
assert "Anthropic API / direct OAuth (not used by Switchyard)" in dockerfile
|
|
assert "scheduleTerminalPaint" in dockerfile
|
|
assert "sessionTree.childrenByParent" in dockerfile
|
|
assert "SessionActivityPanel" in dockerfile
|
|
assert "include_children=${includeChildren}" in dockerfile
|
|
assert "_dashboard_session_is_active" in dockerfile
|
|
assert 'searchParams.get("lineage_root") || resumeParam' in dockerfile
|
|
assert "followActiveLineage" in dockerfile
|
|
assert 'next.delete("lineage_root")' in dockerfile
|
|
activity = (
|
|
ROOT / "dockerfiles" / "hermes-session-activity-panel.tsx"
|
|
).read_text()
|
|
assert "Live worker activity" in activity
|
|
assert "api.getSessionMessages" in activity
|
|
assert "aria-live=\"polite\"" in activity
|
|
assert "role=\"log\"" in activity
|
|
assert "aria-busy={running && messages.length === 0}" in activity
|
|
assert 'detail.source !== "api_server"' not in activity
|
|
assert "Poll-backed transcript; terminal rendering is not required." in activity
|
|
assert "DEFAULT_VISIBLE_MESSAGES = 250" in activity
|
|
assert "current + DEFAULT_VISIBLE_MESSAGES" in activity
|
|
assert 'Show up to{" "}' in activity
|
|
assert "Math.min(" in activity
|
|
assert "visibleMessages.map" in activity
|
|
assert "response.total_messages" in activity
|
|
assert "visibleLimit)," in activity
|
|
assert "function displayText(value: unknown)" in activity
|
|
assert "open && (" in activity
|
|
assert 'document.getElementById("hermes-resume-bootstrap")?.remove()' in activity
|
|
assert 'id="hermes-resume-bootstrap"' in dockerfile
|
|
assert 'role="status"' in dockerfile
|
|
assert 'aria-busy="true"' in dockerfile
|
|
assert "Opening Hermes worker activity…" in dockerfile
|
|
assert 'params.has("resume")' in dockerfile
|
|
assert "Open Telegram setup" in dockerfile
|
|
|
|
def test_telegram_router_local_build_defaults_to_cluster_architecture():
|
|
dockerfile = (
|
|
ROOT / "dockerfiles" / "Dockerfile.hermes-chat-router"
|
|
).read_text(encoding="utf-8")
|
|
|
|
assert "ARG TARGETOS=linux" in dockerfile
|
|
assert "ARG TARGETARCH=arm64" in dockerfile
|
|
assert "GOOS=${TARGETOS} GOARCH=${TARGETARCH}" in dockerfile
|
|
|
|
def test_chat_oauth_allows_stale_service_worker_retirement():
|
|
documents = _documents(HERMES / "oauth2-proxy.yaml")
|
|
deployment = next(
|
|
document
|
|
for document in documents
|
|
if document["kind"] == "Deployment"
|
|
and document["metadata"]["name"] == "oauth2-proxy-hermes-chat"
|
|
)
|
|
args = deployment["spec"]["template"]["spec"]["containers"][0]["args"]
|
|
|
|
assert "--skip-auth-route=GET=^/sw[.]js([?].*)?$" in args
|
|
assert "--custom-templates-dir=/etc/oauth2-proxy/templates" in args
|
|
|
|
template = (HERMES / "oauth2-proxy-templates" / "error.html").read_text()
|
|
assert 'http-equiv="refresh"' not in template
|
|
assert "Unable to find a valid CSRF token" in template
|
|
assert "expired or was already used" in template
|
|
assert "sessionStorage" in template
|
|
assert "Automatic recovery paused" in template
|
|
assert "/start?rd=/" in template
|
|
container = deployment["spec"]["template"]["spec"]["containers"][0]
|
|
assert "v7.15.3@sha256:10a1165743a192e" in container["image"]
|
|
assert "--cookie-csrf-per-request=true" in args
|
|
assert "--cookie-csrf-per-request-limit=8" in args
|
|
assert "--trusted-proxy-ip=10.42.0.0/16" in args
|
|
assert "--api-route=^/api/" in args
|
|
assert "--api-route=^/health$" in args
|
|
assert "--cookie-expire=168h" in args
|
|
assert "--cookie-refresh=19m" in args
|
|
assert "--session-store-type=redis" in args
|
|
assert any(
|
|
arg.startswith("--redis-connection-url=redis://hermes-oauth-sessions.")
|
|
for arg in args
|
|
)
|
|
|
|
agent = _documents(HERMES / "agent-deployment.yaml")[0]
|
|
agent_pod = agent["spec"]["template"]["spec"]
|
|
agent_oauth = next(
|
|
container for container in agent_pod["containers"]
|
|
if container["name"] == "oauth2-proxy"
|
|
)
|
|
assert "--custom-templates-dir=/etc/oauth2-proxy/templates" in agent_oauth["args"]
|
|
assert {
|
|
"name": "oauth-templates",
|
|
"mountPath": "/etc/oauth2-proxy/templates",
|
|
"readOnly": True,
|
|
} in agent_oauth["volumeMounts"]
|
|
agent_template_volume = next(
|
|
volume for volume in agent_pod["volumes"]
|
|
if volume["name"] == "oauth-templates"
|
|
)
|
|
assert agent_template_volume["configMap"]["name"] == (
|
|
"hermes-chat-oauth-templates"
|
|
)
|