mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-06-27 11:22:03 +00:00
2f1a47b90e
Make verification closure the default coding behavior after landed file edits while keeping bounded retries and config/env switches for users who need to disable it.
139 lines
4.2 KiB
Python
139 lines
4.2 KiB
Python
import json
|
|
from pathlib import Path
|
|
|
|
from agent.verification_evidence import (
|
|
mark_workspace_edited,
|
|
record_terminal_result,
|
|
)
|
|
from agent.verification_stop import (
|
|
build_verify_on_stop_nudge,
|
|
verify_on_stop_enabled,
|
|
)
|
|
|
|
|
|
def _node_project(root: Path) -> None:
|
|
(root / "package.json").write_text(
|
|
json.dumps({"scripts": {"test": "vitest", "lint": "eslint ."}}),
|
|
encoding="utf-8",
|
|
)
|
|
(root / "pnpm-lock.yaml").write_text("", encoding="utf-8")
|
|
|
|
|
|
def _make_project(root: Path) -> None:
|
|
root.mkdir()
|
|
_node_project(root)
|
|
|
|
|
|
def test_verify_on_stop_default_is_on(monkeypatch):
|
|
monkeypatch.delenv("HERMES_VERIFY_ON_STOP", raising=False)
|
|
assert verify_on_stop_enabled({"agent": {}}) is True
|
|
|
|
|
|
def test_verify_on_stop_env_can_disable(monkeypatch):
|
|
monkeypatch.setenv("HERMES_VERIFY_ON_STOP", "0")
|
|
assert verify_on_stop_enabled({"agent": {"verify_on_stop": True}}) is False
|
|
|
|
|
|
def test_verify_on_stop_config_can_disable(monkeypatch):
|
|
monkeypatch.delenv("HERMES_VERIFY_ON_STOP", raising=False)
|
|
assert verify_on_stop_enabled({"agent": {"verify_on_stop": False}}) is False
|
|
|
|
|
|
def test_no_nudge_after_fresh_pass(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("HERMES_HOME", str(tmp_path / ".hermes"))
|
|
_node_project(tmp_path)
|
|
changed = str(tmp_path / "src" / "app.ts")
|
|
|
|
record_terminal_result(
|
|
command="pnpm test",
|
|
cwd=tmp_path,
|
|
session_id="s1",
|
|
exit_code=0,
|
|
output="green",
|
|
)
|
|
|
|
assert build_verify_on_stop_nudge(session_id="s1", changed_paths=[changed]) is None
|
|
|
|
|
|
def test_nudge_checks_all_edited_workspaces(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("HERMES_HOME", str(tmp_path / ".hermes"))
|
|
project_a = tmp_path / "a"
|
|
project_b = tmp_path / "b"
|
|
_make_project(project_a)
|
|
_make_project(project_b)
|
|
changed_a = str(project_a / "src" / "app.ts")
|
|
changed_b = str(project_b / "src" / "app.ts")
|
|
|
|
record_terminal_result(
|
|
command="pnpm test",
|
|
cwd=project_a,
|
|
session_id="s1",
|
|
exit_code=0,
|
|
output="green",
|
|
)
|
|
mark_workspace_edited(session_id="s1", cwd=project_b, paths=[changed_b])
|
|
|
|
nudge = build_verify_on_stop_nudge(
|
|
session_id="s1",
|
|
changed_paths=[changed_a, changed_b],
|
|
)
|
|
|
|
assert nudge is not None
|
|
assert "fresh passing verification evidence" in nudge
|
|
|
|
|
|
def test_nudge_after_unverified_edit_with_known_command(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("HERMES_HOME", str(tmp_path / ".hermes"))
|
|
_node_project(tmp_path)
|
|
changed = str(tmp_path / "src" / "app.ts")
|
|
mark_workspace_edited(session_id="s1", cwd=tmp_path, paths=[changed])
|
|
|
|
nudge = build_verify_on_stop_nudge(session_id="s1", changed_paths=[changed])
|
|
|
|
assert nudge is not None
|
|
assert "fresh passing verification evidence" in nudge
|
|
assert "`pnpm run test`" in nudge
|
|
assert changed in nudge
|
|
|
|
|
|
def test_nudge_includes_failed_output_summary(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("HERMES_HOME", str(tmp_path / ".hermes"))
|
|
_node_project(tmp_path)
|
|
changed = str(tmp_path / "src" / "app.ts")
|
|
|
|
record_terminal_result(
|
|
command="pnpm test",
|
|
cwd=tmp_path,
|
|
session_id="s1",
|
|
exit_code=1,
|
|
output="expected 1 got 2",
|
|
)
|
|
|
|
nudge = build_verify_on_stop_nudge(session_id="s1", changed_paths=[changed])
|
|
|
|
assert nudge is not None
|
|
assert "failed" in nudge
|
|
assert "expected 1 got 2" in nudge
|
|
assert "repair the code" in nudge
|
|
|
|
|
|
def test_no_nudge_without_canonical_verify_command(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("HERMES_HOME", str(tmp_path / ".hermes"))
|
|
(tmp_path / "package.json").write_text("{}", encoding="utf-8")
|
|
changed = str(tmp_path / "src" / "app.ts")
|
|
|
|
assert build_verify_on_stop_nudge(session_id="s1", changed_paths=[changed]) is None
|
|
|
|
|
|
def test_nudge_attempts_are_bounded(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("HERMES_HOME", str(tmp_path / ".hermes"))
|
|
_node_project(tmp_path)
|
|
changed = str(tmp_path / "src" / "app.ts")
|
|
mark_workspace_edited(session_id="s1", cwd=tmp_path, paths=[changed])
|
|
|
|
assert build_verify_on_stop_nudge(
|
|
session_id="s1",
|
|
changed_paths=[changed],
|
|
attempts=2,
|
|
max_attempts=2,
|
|
) is None
|