This repository has been archived on 2026-05-02. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
risv3-relay/tests/test_daemon_loop.py
ac 540b4f5b01 feat: relay daemon skeleton — queue, dispatch, conversation, ntfy (#1)
First-PR scope from #1. Single-process Python daemon that relays
between Claude Code instances and chat-Claude (Anthropic API).

Components:

* relay.config — .env + config.yaml loader. Auto-generates ntfy
  topic on first run and persists it back to .env.
* relay.state — atomic file I/O via tempfile + rename, advisory
  flock at state/.lock to enforce single-instance.
* relay.conversation — append-only history with summarization.
  Triggers a summarize call when total chars exceed
  HISTORY_CHAR_CAP (default 400k); replaces history with the
  summary plus the most recent 10 turns.
* relay.anthropic_client — SDK wrapper. Marks the system prompt
  cacheable (5-min ephemeral cache); concatenates text blocks;
  estimates per-call cost from the Anthropic price table with
  cache-write/read accounted for.
* relay.queue — JSON envelope intake; oldest-by-mtime;
  malformed envelopes moved to queue/.rejected/.
* relay.dispatch — one-input-at-a-time per session
  (dispatch/<session_id>/input.txt). Won't overwrite a pending
  dispatch; queues internally and waits for CC to delete.
* relay.ntfy — best-effort POST to https://ntfy.sh/<topic>;
  failures logged but never block the main loop.
* relay.daemon — main loop. Polls jc_input.txt (priority) then
  queue/. Detects [NEEDS-JC] in the first 200 chars of any
  response and pauses dispatch until JC writes jc_input.txt.
  JC override supports @session-N: prefix for direct dispatch
  without an API call.
* relay.__main__ — CLI: relay run / relay status / relay topic.

Tests: 57 unit tests pass (config, state, conversation, queue,
dispatch, anthropic_client, ntfy, full daemon loop with a fake
client). One real-API smoke test marked real_api, opt-in via
pytest -m real_api; skips cleanly on credit-balance errors.

Out of scope for this PR (deferred to follow-ups): Flask status
endpoint, multi-session config in production, exponential
backoff, systemd unit, cost-tracking aggregation.

Closes #1.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-02 15:24:47 +00:00

230 lines
8.3 KiB
Python

"""End-to-end loop test with a fake AnthropicClient.
Drives one daemon `_tick` at a time so we can inspect state after each
event. Tests:
- queue entry → API call → dispatch
- response containing [NEEDS-JC] → state=needs_jc, no dispatch
- jc_input.txt with @session-N: prefix → direct dispatch, no API call
- jc_input.txt without prefix → next chat-side turn (API call), clears needs_jc
- summarization triggers when history exceeds the cap
- malformed queue file is rejected, doesn't break the loop
"""
from __future__ import annotations
import json
from dataclasses import replace
from pathlib import Path
from types import SimpleNamespace
from unittest.mock import patch
import pytest
from relay.anthropic_client import TurnResult
from relay.config import SessionConfig, Settings
from relay.daemon import Daemon
@pytest.fixture
def settings(tmp_path: Path) -> Settings:
s = Settings(
api_key="sk-fake",
model="claude-opus-4-7",
ntfy_topic="test-topic",
status_port=0,
history_char_cap=400_000,
repo_root=tmp_path,
queue_dir=tmp_path / "queue",
dispatch_dir=tmp_path / "dispatch",
state_dir=tmp_path / "state",
logs_dir=tmp_path / "logs",
system_prompt="SYS",
summarization_prompt="please summarize",
sessions=(SessionConfig(session_id="session-1"),),
)
for d in (s.queue_dir, s.dispatch_dir, s.state_dir, s.logs_dir):
d.mkdir(parents=True, exist_ok=True)
return s
def _fake_turn_result(text: str, *, cost: float = 0.001) -> TurnResult:
return TurnResult(
text=text,
input_tokens=100,
output_tokens=50,
cache_creation_input_tokens=0,
cache_read_input_tokens=0,
estimated_cost_usd=cost,
raw=SimpleNamespace(),
)
def _drop_queue(settings: Settings, name: str, content: str, session_id: str = "session-1") -> Path:
path = settings.queue_dir / name
path.write_text(
json.dumps(
{"session_id": session_id, "timestamp": "2026-05-02T00:00:00Z", "content": content}
)
)
return path
def _build_daemon(settings: Settings, *, response_text: str) -> Daemon:
daemon = Daemon(settings)
daemon.client.send = lambda **kwargs: _fake_turn_result(response_text) # type: ignore[assignment]
return daemon
def test_queue_entry_flows_through_to_dispatch(settings: Settings) -> None:
_drop_queue(settings, "a.json", "first message")
daemon = _build_daemon(settings, response_text="ok cool reply")
with patch("relay.daemon.notify"):
daemon._tick()
dispatched = settings.dispatch_dir / "session-1" / "input.txt"
assert dispatched.read_text() == "ok cool reply"
# Queue entry consumed
assert list(settings.queue_dir.glob("*.json")) == []
# History recorded
assert daemon.conversation.total_chars() > 0
def test_needs_jc_token_pauses_dispatch(settings: Settings) -> None:
_drop_queue(settings, "a.json", "i need a decision")
daemon = _build_daemon(settings, response_text="[NEEDS-JC] should I pick option A or B?")
with patch("relay.daemon.notify") as ntfy:
daemon._tick()
assert daemon.status.state == "needs_jc"
assert daemon.status.last_needs_jc_text is not None
# No dispatch happened
dispatched = settings.dispatch_dir / "session-1" / "input.txt"
assert not dispatched.exists()
# ntfy was called with the high priority alert
assert any("paused" in c.kwargs.get("title", "").lower() for c in ntfy.call_args_list)
def test_needs_jc_pause_blocks_subsequent_queue_entries(settings: Settings) -> None:
_drop_queue(settings, "a.json", "first")
daemon = _build_daemon(settings, response_text="[NEEDS-JC] need decision")
with patch("relay.daemon.notify"):
daemon._tick()
# New queue entry while paused
_drop_queue(settings, "b.json", "second")
with patch("relay.daemon.notify"):
daemon._tick()
# Second queue entry NOT consumed
pending = list(settings.queue_dir.glob("*.json"))
assert len(pending) == 1
def test_jc_input_prefix_dispatches_directly_without_api_call(settings: Settings) -> None:
daemon = _build_daemon(settings, response_text="should not be called")
api_calls: list[object] = []
daemon.client.send = lambda **kwargs: (api_calls.append(kwargs), _fake_turn_result("x"))[1] # type: ignore[assignment]
(settings.state_dir / "jc_input.txt").write_text("@session-1: do this thing")
with patch("relay.daemon.notify"):
daemon._tick()
dispatched = settings.dispatch_dir / "session-1" / "input.txt"
assert dispatched.read_text() == "do this thing"
assert api_calls == [] # no API call for direct dispatch
assert not (settings.state_dir / "jc_input.txt").exists()
def test_jc_input_clears_needs_jc(settings: Settings) -> None:
daemon = _build_daemon(settings, response_text="[NEEDS-JC] question")
_drop_queue(settings, "a.json", "trigger needs_jc")
with patch("relay.daemon.notify"):
daemon._tick()
assert daemon.status.state == "needs_jc"
# JC writes a non-prefix response — treated as next chat turn.
daemon.client.send = lambda **kwargs: _fake_turn_result("post-jc reply") # type: ignore[assignment]
(settings.state_dir / "jc_input.txt").write_text("here's the answer: do A")
with patch("relay.daemon.notify"):
daemon._tick()
assert daemon.status.state == "running"
dispatched = settings.dispatch_dir / "session-1" / "input.txt"
assert dispatched.read_text() == "post-jc reply"
def test_summarization_triggers_at_cap(settings: Settings) -> None:
tiny = replace(settings, history_char_cap=200)
daemon = Daemon(tiny)
# Pre-load history to exceed the cap
daemon.conversation.append("user", "u" * 100)
daemon.conversation.append("assistant", "a" * 200)
assert daemon.conversation.needs_summarization(200)
daemon.client.send = lambda **kwargs: _fake_turn_result("CONDENSED SUMMARY") # type: ignore[assignment]
_drop_queue(tiny, "a.json", "next user turn")
with patch("relay.daemon.notify"):
daemon._tick()
# First turn after cap was hit invokes summarization, then sends the user
# turn through the same call. The summary turn is now the head.
assert daemon.conversation.turns[0].meta == "summary"
assert daemon.conversation.turns[0].content == "CONDENSED SUMMARY"
def test_malformed_queue_file_does_not_crash_loop(settings: Settings) -> None:
bad = settings.queue_dir / "broken.json"
bad.write_text("{not json")
daemon = _build_daemon(settings, response_text="x")
with patch("relay.daemon.notify"):
daemon._tick()
# Bad file moved to .rejected, no crash
assert not bad.exists()
assert (settings.queue_dir / ".rejected" / "broken.json").exists()
def test_dispatch_queues_when_session_input_pending(settings: Settings) -> None:
"""Two queue entries for the same session and the first dispatch hasn't been consumed yet."""
daemon = _build_daemon(settings, response_text="reply 1")
_drop_queue(settings, "a.json", "msg 1")
with patch("relay.daemon.notify"):
daemon._tick()
assert (settings.dispatch_dir / "session-1" / "input.txt").read_text() == "reply 1"
daemon.client.send = lambda **kwargs: _fake_turn_result("reply 2") # type: ignore[assignment]
_drop_queue(settings, "b.json", "msg 2")
with patch("relay.daemon.notify"):
daemon._tick()
# First dispatch still present (CC hasn't consumed); second is queued.
assert (settings.dispatch_dir / "session-1" / "input.txt").read_text() == "reply 1"
assert daemon.dispatch.has_pending("session-1")
# CC consumes
(settings.dispatch_dir / "session-1" / "input.txt").unlink()
with patch("relay.daemon.notify"):
daemon._tick()
assert (settings.dispatch_dir / "session-1" / "input.txt").read_text() == "reply 2"
def test_status_persists_to_disk(settings: Settings) -> None:
daemon = _build_daemon(settings, response_text="ok")
_drop_queue(settings, "a.json", "hi")
with patch("relay.daemon.notify"):
daemon._tick()
daemon._persist_status()
status_path = settings.state_dir / "status.json"
payload = json.loads(status_path.read_text())
assert payload["state"] == "running"
assert payload["last_dispatch_session"] == "session-1"
assert payload["history_turns"] >= 2 # user + assistant