hermes-agent/hermes_time.py
Teknium 0800af0b8a
perf(cli): TTFT round 2 — live reasoning by default, partial-line streaming, prompt-build cache, stale budget-warning docs (#59389)
Follow-up to #59332 targeting the remaining PERCEIVED first-token latency
(the wire streaming was already per-token; these fix what the user sees):

1. display.show_reasoning default ON. On thinking models the reasoning
   phase streams for tens of seconds; with the display off users stare
   at a spinner the whole time and read it as a stall. Flipped in
   DEFAULT_CONFIG, load_cli_config defaults, tui_gateway raw-YAML
   fallbacks, and the hermes setup status line (all four read sites kept
   in sync). Gateway per-platform defaults intentionally stay off —
   messaging chats shouldn't fill with thinking text. /reasoning hide
   still turns it off and persists.

2. Response box force-flushes long partial lines. _emit_stream_text only
   painted on newline, so a response opening with a long paragraph
   stayed invisible until the first \n — seconds of blank box. Now
   partial lines wrap at terminal width and paint as tokens arrive
   (mirrors the reasoning box's 80-char force-flush that existed since
   day one). Table blocks remain batch-aligned; no content loss at wrap
   boundaries (regression tests added).

3. hermes_time timezone resolution uses read_raw_config (mtime-cached +
   libyaml C loader) instead of a raw yaml.safe_load of config.yaml
   (~110-140ms measured) inside the FIRST system prompt build. First
   build drops 320ms -> ~155ms on a 200-skill install.

4. Stale docs: configuration.md (en+zh) still documented the 70%/90%
   [BUDGET WARNING] tool-result injections. Those were removed in April
   2026 (c8aff7463) precisely because they hurt task completion; current
   behavior is exhaustion-message + one grace call, no mid-loop
   injection, no cache impact. Docs now describe reality.

Verified: token-count compression decisions already use API-reported
last_prompt_tokens (rough estimators are preflight-only and cost ~1.7ms
even on 1.7MB histories — not worth touching).
2026-07-06 00:16:38 -07:00

135 lines
4.4 KiB
Python

"""
Timezone-aware clock for Hermes.
Provides a single ``now()`` helper that returns a timezone-aware datetime
based on the user's configured IANA timezone (e.g. ``Asia/Kolkata``).
Resolution order:
1. ``HERMES_TIMEZONE`` environment variable
2. ``timezone`` key in ``~/.hermes/config.yaml``
3. Falls back to the server's local time (``datetime.now().astimezone()``)
Invalid timezone values log a warning and fall back safely — Hermes never
crashes due to a bad timezone string.
"""
import logging
import os
from datetime import datetime
from hermes_constants import get_config_path
from typing import Optional
logger = logging.getLogger(__name__)
try:
from zoneinfo import ZoneInfo
except ImportError:
# Python 3.8 fallback (shouldn't be needed — Hermes requires 3.9+)
from backports.zoneinfo import ZoneInfo # type: ignore[no-redef]
# Cached state — resolved once, reused on every call.
# Call reset_cache() to force re-resolution (e.g. after config changes).
_cached_tz: Optional[ZoneInfo] = None
_cached_tz_name: Optional[str] = None
_cache_resolved: bool = False
def _resolve_timezone_name() -> str:
"""Read the configured IANA timezone string (or empty string).
This does file I/O when falling through to config.yaml, so callers
should cache the result rather than calling on every ``now()``.
"""
# 1. Environment variable (highest priority — set by Supervisor, etc.)
tz_env = os.getenv("HERMES_TIMEZONE", "").strip()
if tz_env:
return tz_env
# 2. config.yaml ``timezone`` key
try:
# Prefer the shared cached raw-config reader (mtime/size-keyed cache +
# libyaml C loader) — a direct yaml.safe_load of a large config.yaml
# costs ~100ms+ and this used to run inside the FIRST system prompt
# build, on the time-to-first-token critical path.
try:
from hermes_cli.config import read_raw_config
cfg = read_raw_config() or {}
except Exception:
import yaml
config_path = get_config_path()
if config_path.exists():
with open(config_path, encoding="utf-8") as f:
cfg = yaml.safe_load(f) or {}
else:
cfg = {}
if cfg:
# Managed scope: an administrator can pin ``timezone`` too. Overlay
# via the shared helper (fail-open) since this reads config.yaml directly.
try:
from hermes_cli import managed_scope
cfg = managed_scope.apply_managed_overlay(cfg)
except Exception:
pass
tz_cfg = cfg.get("timezone", "")
if isinstance(tz_cfg, str) and tz_cfg.strip():
return tz_cfg.strip()
except Exception:
pass
return ""
def _get_zoneinfo(name: str) -> Optional[ZoneInfo]:
"""Validate and return a ZoneInfo, or None if invalid."""
if not name:
return None
try:
return ZoneInfo(name)
except (KeyError, Exception) as exc:
logger.warning(
"Invalid timezone '%s': %s. Falling back to server local time.",
name, exc,
)
return None
def get_timezone() -> Optional[ZoneInfo]:
"""Return the user's configured ZoneInfo, or None (meaning server-local).
Resolved once and cached. Call ``reset_cache()`` after config changes.
"""
global _cached_tz, _cached_tz_name, _cache_resolved
if not _cache_resolved:
_cached_tz_name = _resolve_timezone_name()
_cached_tz = _get_zoneinfo(_cached_tz_name)
_cache_resolved = True
return _cached_tz
def reset_cache() -> None:
"""Clear the cached timezone so the next call re-resolves it.
Call this after the configured timezone may have changed (e.g. after a
config edit or ``HERMES_TIMEZONE`` update) to force ``get_timezone()`` /
``now()`` to read the new value instead of the value cached at first use.
"""
global _cached_tz, _cached_tz_name, _cache_resolved
_cached_tz = None
_cached_tz_name = None
_cache_resolved = False
def now() -> datetime:
"""
Return the current time as a timezone-aware datetime.
If a valid timezone is configured, returns wall-clock time in that zone.
Otherwise returns the server's local time (via ``astimezone()``).
"""
tz = get_timezone()
if tz is not None:
return datetime.now(tz)
# No timezone configured — use server-local (still tz-aware)
return datetime.now().astimezone()