Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions src/agent_env/a2a_agent/a2a_agent.py
Original file line number Diff line number Diff line change
Expand Up @@ -215,6 +215,13 @@ def operation(extension: dict | None, name: str) -> tuple[dict | None, str | Non
return method, spec.path
return method, config.get("endpoint") or (spec.path if spec else None)

@staticmethod
def tool_settings(card: dict) -> dict:
"""The agent's own-tool selection and peer-tools setting as deploy_agent stamped them on its card
(``builtin_tools``, ``peer_tools``); only the ones set. Sent again with each conversation's config, so an
agent that restarted since its deploy doesn't get its tools back."""
return {k: card[k] for k in ("builtin_tools", "peer_tools") if card.get(k) is not None}

@staticmethod
def negotiate_agent_config(card: dict, desired: dict) -> Optional[NegotiatedAgentConfig]:
"""Filter ``desired`` to the fields the card's agent-config extension accepts (capability negotiation)."""
Expand Down
4 changes: 3 additions & 1 deletion src/agent_env/agents/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,9 @@ mostly the lxml and Pillow they depend on, and their pins match the Codex and Ge
task's documents read the same way in each. A `deploy_agent` `builtin_tools`
selection picks the CLI's own tools: `{"disable": "*"}` and `{"only": [...]}` become its `--tools` list, a list to
turn off becomes `--disallowedTools`, which keeps the CLI's defaults for the rest, and any selection adds
`--strict-mcp-config`, so its MCP servers are only the task env's. Tool search is off unless the agent's environment sets
`--strict-mcp-config`, so its MCP servers are only the task env's. A `deploy_agent` `peer_tools: false` also leaves out the peer_agents MCP
server, so the agent can't message the task's other agents; an agent whose agent-config takes no `peer_tools` fails
the deploy. Tool search is off unless the agent's environment sets
`ENABLE_TOOL_SEARCH`, so the CLI sends the env's tools and its own up front. The names are `ClaudeCodeTool`'s in
`claude_bridge.py`: every tool the pinned CLI has, though which it offers by default depends on the model. A test fails
when the Dockerfile's `CLAUDE_CODE_VERSION` moves, until `ClaudeCodeTool` and `TOOLS_CLI_VERSION` move with it.
Expand Down
7 changes: 5 additions & 2 deletions src/agent_env/task_step/task_steps/always_on.py
Original file line number Diff line number Diff line change
Expand Up @@ -122,6 +122,7 @@
PromptAgentTaskStep,
fetch_native_trajectory,
prescribe_formats,
resend_tool_settings,
)
from agent_env.task_step.task_steps.agent.user_sim_common import iso as _iso, person_of, task_zone, utc as _utc
from agent_env.task_step.task_steps.agent.user_sim_common import attached, file_name, for_agent, recorded_opening
Expand Down Expand Up @@ -592,7 +593,7 @@ async def _configure_agents(self) -> None:
"max_thinking_tokens": overrides.get("agent_max_thinking_tokens") or step.max_thinking_tokens,
"output_format": step.output_format, "timeout_seconds": step.timeout_seconds,
"task_id": ctx.metadata.get("task_id"), "agentenv_tools": step.agentenv_tools,
"builtin_tools": card.get("builtin_tools")}
**A2AAgent.tool_settings(card)}
desired = {k: v for k, v in desired.items() if v is not None}
model_params = get_config().get_model_params(step.model_params)
if model_params:
Expand All @@ -612,6 +613,7 @@ async def _configure_agents(self) -> None:
"on its first wake instead")

if self.us.get("prompt_set"): # deploy_user_sim_agent set the user sim's prompt and model at deploy
await resend_tool_settings(self.user_agent, self.user_url) # its tools, though, again, as the agent's
self.user_model, self.user_model_from = self.us["model"], self.us["model_source"]
self.user_setup = dict(self.us["setup"])
self.usage.start(ModelRole.USER_SIMULATOR, self.user_model, self.user_agent)
Expand All @@ -620,7 +622,8 @@ async def _configure_agents(self) -> None:
self.user_model, self.user_model_from = self._user_model()
user_config = {"model": self.user_model, "system_prompt": setup.get("system_prompt"),
"output_format": setup.get("output_format")}
user_config = {k: v for k, v in user_config.items() if v is not None}
user_config = {k: v for k, v in user_config.items() if v is not None} | A2AAgent.tool_settings(
self.user_agent.a2a_card or {})
negotiated = A2AAgent.negotiate_agent_config(self.user_agent.a2a_card or {}, user_config)
if negotiated:
await protocol.post_agent_config(self.user_url + negotiated.endpoint, negotiated.fields,
Expand Down
27 changes: 26 additions & 1 deletion src/agent_env/task_step/task_steps/deploy_agent.py
Original file line number Diff line number Diff line change
Expand Up @@ -151,10 +151,15 @@ def _skill_fields(skill: dict) -> dict:



def _config_request(agent_config: dict | None) -> dict:
"""The ``set`` request description of a card's agent-config extension."""
return ((agent_config or {}).get("params") or {}).get("methods", {}).get("set", {}).get("request", {})


def _check_builtin_tools(agent_config: dict | None, agent_name: str, selection: dict) -> None:
"""Refuse a selection of the agent's own tools that its card's agent-config extension can't take: it has no
``builtin_tools``, or its schema doesn't name a tool the selection does."""
request = ((agent_config or {}).get("params") or {}).get("methods", {}).get("set", {}).get("request", {})
request = _config_request(agent_config)
if "builtin_tools" not in request.get("supported", []):
raise RuntimeError(f"Agent '{agent_name}' cannot have its own tools selected: its card's "
f"{AGENT_CONFIG_V1.uri} takes no 'builtin_tools'")
Expand All @@ -165,6 +170,14 @@ def _check_builtin_tools(agent_config: dict | None, agent_name: str, selection:
if error is not None:
raise RuntimeError(f"Agent '{agent_name}' can't take builtin_tools {selection}: {error.message}")


def _check_peer_tools(agent_config: dict | None, agent_name: str) -> None:
"""Refuse to turn off the peer_agents tools of an agent whose card's agent-config extension can't take
``peer_tools``: it would keep them."""
if "peer_tools" not in _config_request(agent_config).get("supported", []):
raise RuntimeError(f"Agent '{agent_name}' cannot have its peer_agents tools turned off: its card's "
f"{AGENT_CONFIG_V1.uri} takes no 'peer_tools'")

class DeployAgentTaskStep(TaskStep):
type: ClassVar[str] = "deploy_agent"
entity_refs = (
Expand Down Expand Up @@ -208,6 +221,7 @@ def __init__(
network_policy: Optional[dict] = None,
role: Optional[str] = None,
builtin_tools: Optional[dict] = None,
peer_tools: Optional[bool] = None,
enable_agent_changelog: bool = False,
agent_changelog_toolcall_position_exclusive: Optional[int] = None,
sandbox_name: Optional[str] = None,
Expand Down Expand Up @@ -241,6 +255,10 @@ def __init__(
self.role = role or None
self.builtin_tools = (None if builtin_tools is None
else BuiltinTools[str].model_validate(builtin_tools).model_dump(exclude_none=True))
# False: the agent gets no peer_agents MCP tools, so it can't message the task's other agents.
if peer_tools is not None and not isinstance(peer_tools, bool):
raise ValueError(f"peer_tools must be true or false, not {peer_tools!r}")
self.peer_tools = peer_tools
self.enable_agent_changelog = enable_agent_changelog
self.agent_changelog_object_url = agent_changelog_object_url
self.agent_changelog_toolcall_position_exclusive = agent_changelog_toolcall_position_exclusive
Expand Down Expand Up @@ -279,6 +297,7 @@ def to_dict(self) -> dict:
base["network_policy"] = self.network_policy
base["role"] = self.role
base["builtin_tools"] = self.builtin_tools
base["peer_tools"] = self.peer_tools
base["enable_agent_changelog"] = self.enable_agent_changelog
base["agent_changelog_object_url"] = self.agent_changelog_object_url
base["agent_changelog_toolcall_position_exclusive"] = self.agent_changelog_toolcall_position_exclusive
Expand Down Expand Up @@ -312,6 +331,7 @@ def from_dict(cls, data: dict) -> DeployAgentTaskStep:
network_policy=data.get("network_policy"),
role=data.get("role"),
builtin_tools=data.get("builtin_tools"),
peer_tools=data.get("peer_tools"),
enable_agent_changelog=data.get("enable_agent_changelog", False),
agent_changelog_object_url=data.get("agent_changelog_object_url"),
agent_changelog_toolcall_position_exclusive=data.get("agent_changelog_toolcall_position_exclusive"),
Expand Down Expand Up @@ -464,6 +484,11 @@ async def _execute(self, context: TaskStepContext, deployed_agents: list) -> Tas
_check_builtin_tools(A2AAgent.find_extension(card, A2AAgent.EXT_AGENT_CONFIG), self.agent_name,
self.builtin_tools)
identity_payload["builtin_tools"] = self.builtin_tools
if self.peer_tools is not None:
if not self.peer_tools:
# Before the first request to the agent, as for builtin_tools.
_check_peer_tools(A2AAgent.find_extension(card, A2AAgent.EXT_AGENT_CONFIG), self.agent_name)
identity_payload["peer_tools"] = self.peer_tools
Comment thread
kaihackney-scale marked this conversation as resolved.
Resolved
identity = A2AAgent.negotiate_agent_config(card, identity_payload)
if self.role is not None:
carried_by = [via for via, carried in (
Expand Down
21 changes: 17 additions & 4 deletions src/agent_env/task_step/task_steps/prompt_agent.py
Original file line number Diff line number Diff line change
Expand Up @@ -556,8 +556,8 @@ async def _converse(
"output_format": self.output_format, "timeout_seconds": self.timeout_seconds,
"task_id": context.metadata.get("task_id"),
"agentenv_tools": self.agentenv_tools,
# Again each turn, so an agent that restarted since deploy_agent sent it doesn't get its own tools back.
"builtin_tools": card.get("builtin_tools"),
# Again, so an agent that restarted since deploy_agent sent them doesn't get its tools back.
**A2AAgent.tool_settings(card),
}
desired_config = {k: v for k, v in desired_config.items() if v is not None}
model_params = get_config().get_model_params(self.model_params)
Expand All @@ -576,13 +576,17 @@ async def _converse(

is_user_sim = user_sim is not None and self.user_a2a_url is None
if built_in is not None: # deploy_user_sim_agent set its model and answer format
await resend_tool_settings(user_sim, user_url)
context.record_model(ModelRole.USER_SIMULATOR.record_key, self.id, built_in[0].get("model"))
usage.start(ModelRole.USER_SIMULATOR, built_in[0].get("model"), user_sim)
elif is_user_sim:
user_output_format = self.user_output_format or _DEFAULT_USER_SIM_OUTPUT_FORMAT
user_model = self._resolve_user_model(context)
user_config = {k: v for k, v in {"output_format": user_output_format, "model": user_model}.items() if v is not None}
negotiated = A2AAgent.negotiate_agent_config(user_sim.a2a_card or {}, user_config)
user_card = user_sim.a2a_card or {}
# Its own tools and peer tools as deploy_agent set them, again, as for the agent above.
user_config = {k: v for k, v in {"output_format": user_output_format, "model": user_model}.items()
if v is not None} | A2AAgent.tool_settings(user_card)
negotiated = A2AAgent.negotiate_agent_config(user_card, user_config)
if negotiated:
await protocol.post_agent_config(user_url + negotiated.endpoint, negotiated.fields,
headers=user_sim.control_headers)
Expand Down Expand Up @@ -1421,6 +1425,15 @@ async def prescribe_formats(context: TaskStepContext, step_id: str, agent: Deplo
return


async def resend_tool_settings(agent: DeployedAgent, url: str) -> None:
"""Send ``agent`` its own-tool and peer-tools settings again, as deploy_agent stamped them on its card, for a user
sim whose model and prompt deploy_user_sim_agent already set; nothing when it has none, or its card takes neither."""
card = agent.a2a_card or {}
negotiated = A2AAgent.negotiate_agent_config(card, A2AAgent.tool_settings(card))
if negotiated:
await protocol.post_agent_config(url + negotiated.endpoint, negotiated.fields, headers=agent.control_headers)


class StoredTrajectory(NamedTuple):
"""Where a turn's native trajectory was stored, None when the agent returned none, and the format its answer
named, None when it named none."""
Expand Down
41 changes: 41 additions & 0 deletions tst/unit/task_step/test_always_on_user_sim_tools.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
"""always_on sends its user sim's own-tool and peer-tools settings again, as deploy_agent stamped them on its card,
for a user sim deploy_user_sim_agent set up (its prompt and model already set) as for one it sets up itself; so a user
sim that restarted since its deploy doesn't get its tools back."""

from __future__ import annotations

import pytest

from agent_env.a2a_agent import A2AAgent
from agent_env.task_step.task_steps import always_on as ao
from tst.unit.task_step.always_on_harness import USER_SIM, Rig

pytestmark = pytest.mark.usefixtures("local_stores")

SAVED = {"builtin_tools": {"disable": "*"}, "peer_tools": False}
CARD = {"capabilities": {"extensions": [{"uri": A2AAgent.EXT_AGENT_CONFIG, "params": {
"endpoint": "/ext/agent-config", "methods": {"set": {"request": {"supported": [
"model", "system_prompt", "output_format", "builtin_tools", "peer_tools"]}}}}}]}, **SAVED}
PROMPT_SET = {"prompt_set": True, "multi_day": False, "model_source": "user_sim_step",
"setup": {"model": "sim-model", "model_from": "user_sim_step", "accepted_fields": ["model"]}}


@pytest.mark.parametrize("user", [PROMPT_SET, {}], ids=["deploy_user_sim_agent set it up", "always_on sets it up"])
@pytest.mark.asyncio
async def test_the_user_sim_is_sent_its_saved_tool_settings(monkeypatch, tmp_path, user):
posted: list[tuple[str, dict]] = []

async def post(url, fields, headers=None):
posted.append((url, fields))

monkeypatch.setattr(ao.protocol, "post_agent_config", post)
rig = Rig(monkeypatch, tmp_path)
await rig.arm()
ctx = rig.context(user=user)
next(a for a in ctx.deployed_agents if a.agent_name == "user-sim").a2a_card = CARD
await rig.run(rig.step(max_wakes=1), ctx)

sent = [fields for url, fields in posted if url.startswith(USER_SIM)]
assert sent and all(fields.items() >= SAVED.items() for fields in sent)
if user.get("prompt_set"): # its prompt and model stay deploy_user_sim_agent's: only the tool settings go
assert sent == [SAVED]
47 changes: 47 additions & 0 deletions tst/unit/task_step/test_deploy_agent_role_delivery.py
Original file line number Diff line number Diff line change
Expand Up @@ -229,6 +229,53 @@ async def test_a_selection_the_agent_cannot_take_fails_before_any_request(wiring
assert wiring.agent.close.await_count == 1


PEER_CONFIG = _agent_config(["name", "description", "peer_tools"])


@pytest.mark.asyncio
async def test_peer_tools_off_is_sent_once_with_its_identity_and_kept_on_the_card(wiring):
context = await wiring.deploy(_card(HEADERS_MCP, PEER_CONFIG), role=None, peer_tools=False)

assert wiring.bodies("/ext/agent-config") == [{"name": "solver", "peer_tools": False}]
assert context.deployed_agents[0].a2a_card["peer_tools"] is False


@pytest.mark.asyncio
async def test_peer_tools_left_unset_sends_none(wiring):
await wiring.deploy(_card(HEADERS_MCP, PEER_CONFIG), role=None)

assert wiring.bodies("/ext/agent-config") == [{"name": "solver"}]


@pytest.mark.asyncio
@pytest.mark.parametrize("card", [_card(HEADERS_MCP, ROLE_CONFIG), _card(HEADERS_MCP)],
ids=["config-without-it", "no-agent-config"])
async def test_peer_tools_off_on_an_agent_that_cannot_take_it_fails_before_any_request(wiring, card):
with pytest.raises(RuntimeError, match="takes no 'peer_tools'"):
await wiring.deploy(card, role=None, peer_tools=False)

assert wiring.requests == []
assert wiring.agent.close.await_count == 1


@pytest.mark.asyncio
async def test_peer_tools_on_for_an_agent_that_cant_take_it_is_dropped_and_the_deploy_goes_on(wiring):
# true is the default an agent already has, so it isn't checked against the card: negotiation leaves it out.
context = await wiring.deploy(_card(HEADERS_MCP, ROLE_CONFIG), role=None, peer_tools=True)

assert wiring.bodies("/ext/agent-config") == [{"name": "solver"}]
assert "peer_tools" not in context.deployed_agents[0].a2a_card


def test_peer_tools_is_a_bool_and_round_trips():
with pytest.raises(ValueError):
DeployAgentTaskStep(id="ag", version=None, peer_tools="no")
step = DeployAgentTaskStep(id="ag", version=None, peer_tools=False)
assert DeployAgentTaskStep.from_dict({**step.to_dict(), "version": 1}).peer_tools is False
assert DeployAgentTaskStep.from_dict({**DeployAgentTaskStep(id="ag", version=None).to_dict(),
"version": 1}).peer_tools is None


def test_a_selection_takes_exactly_one_of_disable_and_only():
for selection in ({}, {"disable": "*", "only": ["Read"]}, {"enable": ["Read"]}):
with pytest.raises(ValueError):
Expand Down
26 changes: 26 additions & 0 deletions tst/unit/task_step/test_prompt_agent_built_in_user_sim.py
Original file line number Diff line number Diff line change
Expand Up @@ -412,3 +412,29 @@ async def test_agent_triggers_registered_for_the_built_in_user_are_refused_befor
"user sim aren't supported yet"):
await _step().execute(ctx)
assert "user_sim_reads" not in ctx.metadata


@pytest.mark.asyncio
async def test_a_built_in_user_sim_is_sent_its_saved_tool_settings_again_and_nothing_else(monkeypatch):
# deploy_user_sim_agent set her model and format; her own tools and peer tools, stamped on her card by deploy, go
# again, so a user sim that restarted since doesn't get its tools back.
Peers(monkeypatch, {AGENT: [text("draft one"), text("glad to help")], USER: [said("done", "perfect")]})
ctx = await _context(monkeypatch)
hannah = next(a for a in ctx.deployed_agents if a.agent_name == "hannah")
card = json.loads(json.dumps(_CONFIG_CARD))
card["capabilities"]["extensions"][0]["params"]["methods"]["set"]["request"]["supported"] += [
"builtin_tools", "peer_tools"]
hannah.a2a_card = {**card, "builtin_tools": {"disable": "*"}, "peer_tools": False}
posted = []

async def post_agent_config(url, payload, timeout_seconds=60, *, headers=None):
posted.append((url, payload))

# _context's deploy step left its own recorder in place of post_agent_config; record this run's posts instead.
monkeypatch.setattr("agent_env.a2a_agent.protocol.post_agent_config", post_agent_config)

await _step().execute(ctx)

assert [p for url, p in posted if url == f"{USER}/ext/agent-config"] == [
{"builtin_tools": {"disable": "*"}, "peer_tools": False}]
assert all(not url.startswith(USER) or set(p) == {"builtin_tools", "peer_tools"} for url, p in posted)
Loading
Loading