Custom HTTP tools an administrator defines

A row in custom_tools becomes a ToolDef like any built-in, offered beside
the thirteen. The registry had to stop being an import-time constant for
that: `resolve_tools` now returns the schemas *and* the runners together,
carried to the loop on the ToolContext.

That closes a hole on the way. `run_tool` looked names up in the global
REGISTRY with no reference to what had been offered, so a model naming a
tool its chat was gated out of -- a family switched off, a permission the
reader lacks -- had it run anyway. The resolved set is now authoritative.

Arguments come from a model, so an argument may fill a hole but never move
the target: the scheme and host of a URL template are literal, values are
escaped for where they land, and the origin is pinned afterwards. Every
redirect hop is checked the way services/fetch.py checks one, and the
secret is dropped if a hop leaves the origin it was issued for.

Also fixes the tool-activity block claiming every library tool had
"searched the web", which it has done since the second family landed.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Jaroslav Beneš
2026-08-01 16:26:47 +02:00
parent d9f274ec1a
commit bc84fec21d
26 changed files with 2771 additions and 60 deletions
+303
View File
@@ -0,0 +1,303 @@
"""The custom-tools admin screen."""
from __future__ import annotations
import json
import httpx
import pytest
from fastapi.testclient import TestClient
from sqlalchemy import select
from lembas.db.models import CustomTool, Group, User
from lembas.security.passwords import hash_password
from lembas.services.crypto import UNCHANGED_SENTINEL, decrypt, encrypt
@pytest.fixture
def plain_user(client: TestClient, db, registered):
"""A second account, which is never an administrator."""
client.post("/auth/logout")
client.post(
"/auth/register",
data={"name": "Sam", "email": "sam@shire.test", "password": "correct horse battery"},
follow_redirects=False,
)
user = db.scalar(select(User).where(User.email == "sam@shire.test"))
user.role = "user"
user.active = True
db.commit()
client.post(
"/auth/login",
data={"email": "sam@shire.test", "password": "correct horse battery"},
follow_redirects=False,
)
return user
def _form(**overrides) -> dict:
base = {
"name": "Weather",
"slug": "weather",
"description": "Look up the weather for a city.",
"parameters": json.dumps(
{"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"]}
),
"method": "GET",
"url_template": "https://api.test/v1/{{city}}",
"headers": "Accept: application/json",
"secret_placement": "bearer",
"secret_name": "Authorization",
"response_mode": "json",
"max_chars": "8000",
"timeout": "20",
"position": "0",
"enabled": "true",
"public": "true",
}
base.update(overrides)
return {key: value for key, value in base.items() if value is not None}
def _create(client: TestClient, **overrides):
return client.post("/admin/tools", data=_form(**overrides), follow_redirects=False)
# --- Guards ------------------------------------------------------------------
def test_the_pages_are_refused_to_a_plain_user(client: TestClient, plain_user):
assert client.get("/admin/tools").status_code == 403
assert client.get("/admin/tools/new").status_code == 403
assert client.post("/admin/tools", data=_form()).status_code == 403
def test_a_tool_that_does_not_exist_is_a_404(client: TestClient, registered):
assert client.get("/admin/tools/nope/edit").status_code == 404
def test_new_is_not_parsed_as_a_tool_id(client: TestClient, registered):
"""FastAPI matches in registration order; /admin/models has been bitten by
exactly this."""
response = client.get("/admin/tools/new")
assert response.status_code == 200
assert "New tool" in response.text
# --- Creating and editing ----------------------------------------------------
def test_creating_a_tool_then_editing_it(client: TestClient, db, registered):
_create(client, secret="s3cret")
tool = db.scalar(select(CustomTool))
assert tool.slug == "weather"
assert tool.headers_json == {"Accept": "application/json"}
assert tool.parameters_json["properties"]["city"] == {"type": "string"}
assert decrypt(tool.secret_encrypted) == "s3cret"
assert tool.enabled is True
client.post(
f"/admin/tools/{tool.id}",
data=_form(name="Forecast", enabled=None, secret=UNCHANGED_SENTINEL),
follow_redirects=False,
)
db.refresh(tool)
assert tool.name == "Forecast"
# An unticked checkbox is simply absent from the post, which is the signal.
assert tool.enabled is False
assert decrypt(tool.secret_encrypted) == "s3cret", "the dots must keep the secret"
def test_clearing_the_field_removes_the_secret(client: TestClient, db, registered):
_create(client, secret="s3cret")
tool = db.scalar(select(CustomTool))
client.post(f"/admin/tools/{tool.id}", data=_form(secret=""), follow_redirects=False)
db.refresh(tool)
assert tool.secret_encrypted == ""
def test_a_secret_is_never_rendered_in_full(client: TestClient, db, registered):
_create(client, secret="super-secret-value")
tool = db.scalar(select(CustomTool))
page = client.get(f"/admin/tools/{tool.id}/edit").text
assert "super-secret-value" not in page
assert UNCHANGED_SENTINEL in page
# --- Validation reports back into the form -----------------------------------
def test_bad_parameter_json_is_reported_not_a_422(client: TestClient, db, registered):
response = _create(client, parameters="{not json")
assert response.status_code == 200
assert "not valid JSON" in response.text
assert db.scalar(select(CustomTool)) is None
def test_parameters_that_are_not_an_object_are_refused(client: TestClient, db, registered):
response = _create(client, parameters='{"type": "string"}')
assert "must be a JSON object" in response.text
assert db.scalar(select(CustomTool)) is None
def test_a_slug_colliding_with_a_builtin_is_refused(client: TestClient, db, registered):
response = _create(client, slug="web_search")
assert "built-in tool" in response.text
assert db.scalar(select(CustomTool)) is None
def test_a_duplicate_slug_is_refused(client: TestClient, db, registered):
_create(client)
response = _create(client, name="Other")
assert "already a tool" in response.text
assert len(list(db.scalars(select(CustomTool)))) == 1
def test_a_url_whose_host_is_a_variable_is_refused(client: TestClient, db, registered):
response = _create(client, url_template="https://{{city}}.test/x")
assert "literal" in response.text
assert db.scalar(select(CustomTool)) is None
def test_a_rejected_save_keeps_what_was_typed(client: TestClient, db, registered):
_create(client)
tool = db.scalar(select(CustomTool))
response = client.post(
f"/admin/tools/{tool.id}",
data=_form(name="Renamed", parameters="{oops"),
follow_redirects=False,
)
assert "Renamed" in response.text, "the form still holds the submitted name"
db.refresh(tool)
assert tool.name == "Weather", "and the stored row is untouched"
# --- Access ------------------------------------------------------------------
def test_making_a_tool_public_clears_its_groups(client: TestClient, db, registered):
group = Group(name="Council")
db.add(group)
db.commit()
_create(client, public=None, group_ids=group.id)
tool = db.scalar(select(CustomTool))
assert [g.name for g in tool.groups] == ["Council"]
client.post(
f"/admin/tools/{tool.id}",
data={**_form(public="true"), "group_ids": group.id},
follow_redirects=False,
)
db.refresh(tool)
assert tool.groups == []
def test_a_restricted_tool_is_hidden_from_a_user_outside_its_groups(db, registered):
from lembas.services import tool_access
group = Group(name="Council")
outsider = User(name="Sam", email="s@shire.test", password_hash=hash_password("x"))
db.add_all([group, outsider])
db.add(
CustomTool(
slug="weather", name="Weather", url_template="https://api.test/", public=False
)
)
db.commit()
tool = db.scalar(select(CustomTool))
tool.groups = [group]
db.commit()
assert tool_access.visible_custom_tools(db, outsider) == []
outsider.groups = [group]
db.commit()
assert len(tool_access.visible_custom_tools(db, outsider)) == 1
# --- Running one by hand -----------------------------------------------------
def test_the_test_button_shows_what_the_model_would_read(
client: TestClient, db, registered, mock_http, monkeypatch
):
monkeypatch.setattr(
"socket.getaddrinfo", lambda *a, **k: [(2, 1, 6, "", ("93.184.216.34", 80))]
)
mock_http(lambda _r: httpx.Response(200, json={"summary": "Sunny"}))
_create(client, response_mode="json", response_path="summary")
tool = db.scalar(select(CustomTool))
response = client.post(
f"/admin/tools/{tool.id}/test", data={"arguments": '{"city": "Minas Tirith"}'}
)
assert response.status_code == 200
assert "Sunny" in response.text
def test_a_failing_test_is_recorded_on_the_row(
client: TestClient, db, registered, mock_http, monkeypatch
):
monkeypatch.setattr(
"socket.getaddrinfo", lambda *a, **k: [(2, 1, 6, "", ("93.184.216.34", 80))]
)
mock_http(lambda _r: httpx.Response(503, text="down"))
_create(client)
tool = db.scalar(select(CustomTool))
client.post(f"/admin/tools/{tool.id}/test", data={"arguments": "{}"})
db.refresh(tool)
assert "503" in tool.last_error
assert tool.last_checked_at is not None
def test_bad_test_arguments_are_reported(client: TestClient, db, registered):
_create(client)
tool = db.scalar(select(CustomTool))
response = client.post(f"/admin/tools/{tool.id}/test", data={"arguments": "[1, 2]"})
assert "JSON object" in response.text
# --- The list ----------------------------------------------------------------
def test_the_list_searches_and_filters(client: TestClient, db, registered):
_create(client)
_create(client, name="Tickets", slug="tickets", url_template="https://jira.test/{{city}}")
tool = db.scalar(select(CustomTool).where(CustomTool.slug == "tickets"))
tool.enabled = False
db.commit()
page = client.get("/admin/tools").text
assert "Weather" in page and "Tickets" in page
assert "Tickets" not in client.get("/admin/tools?filter=enabled").text
assert "Weather" not in client.get("/admin/tools?q=tick").text
def test_deleting_a_tool_leaves_its_prompt_override_alone(client: TestClient, db, registered):
"""The override outlives the row, which is what lets a tool be recreated
under the same slug without losing the wording somebody chose."""
from lembas.services import prompts as prompts_service
_create(client, guidance="- Default wording.")
tool = db.scalar(select(CustomTool))
prompts_service.save(db, {"tool.custom_weather": "- Edited wording."})
client.post(f"/admin/tools/{tool.id}/delete", follow_redirects=False)
assert db.scalar(select(CustomTool)) is None
assert prompts_service.stored(db)["tool.custom_weather"] == "- Edited wording."
def test_the_secret_survives_a_round_trip_through_the_form(client: TestClient, db, registered):
"""A regression guard on the sentinel: the field is rendered with dots, and
submitting the page unchanged must not overwrite the key with them."""
db.add(
CustomTool(
slug="weather",
name="Weather",
url_template="https://api.test/",
secret_encrypted=encrypt("s3cret"),
)
)
db.commit()
tool = db.scalar(select(CustomTool))
client.post(
f"/admin/tools/{tool.id}", data=_form(secret=UNCHANGED_SENTINEL), follow_redirects=False
)
db.refresh(tool)
assert decrypt(tool.secret_encrypted) == "s3cret"