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>

relay/ntfy.py

@@ -0,0 +1,59 @@
+"""ntfy.sh notifications.
+
+Topic is loaded from settings (auto-generated on first run). The topic
+is functionally a password — anyone subscribed to the topic URL
+receives notifications. We use cryptographically-random topics
+(secrets.token_urlsafe(12)) to make brute-force discovery impractical.
+
+Notifications are best-effort: a failure to deliver (network down,
+ntfy.sh outage) is logged but does NOT block the daemon's main loop.
+"""
+
+from __future__ import annotations
+
+import logging
+
+import requests
+
+logger = logging.getLogger(__name__)
+
+NTFY_BASE = "https://ntfy.sh"
+
+
+def topic_url(topic: str) -> str:
+    return f"{NTFY_BASE}/{topic}"
+
+
+def notify(
+    topic: str,
+    title: str,
+    message: str,
+    *,
+    priority: str = "default",
+    tags: list[str] | None = None,
+) -> bool:
+    """Post one notification. Returns True on HTTP 200, False otherwise.
+
+    Never raises on transport errors — this path runs from the daemon's
+    main loop and a failed notification should not stop work.
+    """
+    if not topic:
+        logger.warning("ntfy topic is empty; skipping notification: %s", title)
+        return False
+    headers = {
+        "Title": title,
+        "Priority": priority,
+    }
+    if tags:
+        headers["Tags"] = ",".join(tags)
+    try:
+        resp = requests.post(
+            topic_url(topic), data=message.encode("utf-8"), headers=headers, timeout=10
+        )
+    except requests.RequestException as exc:
+        logger.warning("ntfy delivery failed for topic <redacted>: %s", exc)
+        return False
+    if resp.status_code != 200:
+        logger.warning("ntfy non-200 (%s): %s", resp.status_code, resp.text[:200])
+        return False
+    return True