common/hermes
2
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

feat: inject current time into goal judge prompt

Browse Source 
The goal judge only receives the goal text and the agent's last
response. It has no concept of the current time, making it
impossible to evaluate time-sensitive goals like 'keep working
until 5pm'.

This commit adds 'Current time' to both JUDGE_USER_PROMPT_TEMPLATE
and JUDGE_USER_PROMPT_WITH_SUBGOALS_TEMPLATE, computed from
datetime.now().astimezone() at judge call time.
This commit is contained in:
draplater
2026-05-16 23:02:46 -07:00
committed by Teknium
parent 6f50c26b2a
commit 6158964ff6
1 changed files with 6 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
hermes_cli/goals.py
View File

@@ -34,6 +34,7 @@ import logging
import re import re
import time import time
from dataclasses import dataclass, field, asdict from dataclasses import dataclass, field, asdict
from datetime import datetime, timezone
from typing import Any, Dict, List, Optional, Tuple from typing import Any, Dict, List, Optional, Tuple
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
@@ -110,6 +111,7 @@ JUDGE_SYSTEM_PROMPT = (
JUDGE_USER_PROMPT_TEMPLATE = ( JUDGE_USER_PROMPT_TEMPLATE = (
"Goal:\n{goal}\n\n" "Goal:\n{goal}\n\n"
"Agent's most recent response:\n{response}\n\n" "Agent's most recent response:\n{response}\n\n"
"Current time: {current_time}\n\n"
"Is the goal satisfied?" "Is the goal satisfied?"
) )
@@ -120,6 +122,7 @@ JUDGE_USER_PROMPT_WITH_SUBGOALS_TEMPLATE = (
"Additional criteria the user added mid-loop (all must also be " "Additional criteria the user added mid-loop (all must also be "
"satisfied for the goal to be DONE):\n{subgoals_block}\n\n" "satisfied for the goal to be DONE):\n{subgoals_block}\n\n"
"Agent's most recent response:\n{response}\n\n" "Agent's most recent response:\n{response}\n\n"
"Current time: {current_time}\n\n"
"Decision: For each numbered criterion above, find concrete " "Decision: For each numbered criterion above, find concrete "
"evidence in the agent's response that the criterion is " "evidence in the agent's response that the criterion is "
"satisfied. Do not accept generic phrases like 'all requirements " "satisfied. Do not accept generic phrases like 'all requirements "
@@ -415,6 +418,7 @@ def judge_goal(
# Build the prompt — pick the with-subgoals variant when applicable. # Build the prompt — pick the with-subgoals variant when applicable.
clean_subgoals = [s.strip() for s in (subgoals or []) if s and s.strip()] clean_subgoals = [s.strip() for s in (subgoals or []) if s and s.strip()]
current_time = datetime.now(tz=timezone.utc).astimezone().strftime("%Y-%m-%d %H:%M:%S %Z")
if clean_subgoals: if clean_subgoals:
subgoals_block = "\n".join( subgoals_block = "\n".join(
f"- {i}. {text}" for i, text in enumerate(clean_subgoals, start=1) f"- {i}. {text}" for i, text in enumerate(clean_subgoals, start=1)
@@ -423,11 +427,13 @@ def judge_goal(
goal=_truncate(goal, 2000), goal=_truncate(goal, 2000),
subgoals_block=_truncate(subgoals_block, 2000), subgoals_block=_truncate(subgoals_block, 2000),
response=_truncate(last_response, _JUDGE_RESPONSE_SNIPPET_CHARS), response=_truncate(last_response, _JUDGE_RESPONSE_SNIPPET_CHARS),
current_time=current_time,
) )
else: else:
prompt = JUDGE_USER_PROMPT_TEMPLATE.format( prompt = JUDGE_USER_PROMPT_TEMPLATE.format(
goal=_truncate(goal, 2000), goal=_truncate(goal, 2000),
response=_truncate(last_response, _JUDGE_RESPONSE_SNIPPET_CHARS), response=_truncate(last_response, _JUDGE_RESPONSE_SNIPPET_CHARS),
current_time=current_time,
) )
try: try: