APEX/BrowserUse_and_ComputerUse_skills
5
0
Fork 0
Code Issues 6 Pull requests 1 Projects 1 Releases Packages Wiki Activity Actions

fix to_capcha

Browse source
This commit is contained in:
VladislavIlin7 2026-04-24 15:55:07 +03:00
parent f1f32d8366
commit 3b7d02ee81
3 changed files with 13 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
browser_env/entrypoint.sh
View file

@ -98,7 +98,7 @@ if [ ! -f /var/lib/dbus/machine-id ]; then
dbus-uuidgen > /var/lib/dbus/machine-id 2>/dev/null || true dbus-uuidgen > /var/lib/dbus/machine-id 2>/dev/null || true
fi fi
# Удаляем stale lock/socket от прошлых падений Xvfb на том же DISPLAY. # Удаляем stale lock/socket от прошлых падений Xvfb на том же DISPLAY.
rm -f "/tmp/.X${DISPLAY_NUM}-lock" "/tmp/.X11-unix/X${DISPLAY_NUM}" || true rm -f "/tmp/.X${DISPLAY_NUM}-lock" "/tmp/.X11-unix/X${DISPLAY_NUM}" || true
log "starting X stack on DISPLAY=${DISPLAY}" log "starting X stack on DISPLAY=${DISPLAY}"

9
hermes_code/agent/prompt_builder.py
View file

@ -154,6 +154,15 @@ SKILLS_GUIDANCE = (
"Skills that aren't maintained become liabilities." "Skills that aren't maintained become liabilities."
) )
BROWSER_CAPTCHA_GUIDANCE = (
"For browser tasks, do not pre-emptively refuse just because CAPTCHA may appear. "
"Start the task with internet_browser. "
"If the browser runtime reports status='awaiting_user_captcha', immediately use to_captcha "
"to hand control to the user for manual verification and then resume. "
"Important: you must never claim that you solved CAPTCHA yourself and you must not attempt bypass methods. "
"Your role is orchestration: run browser steps, request manual CAPTCHA completion from the user, verify, and continue."
)
PLATFORM_HINTS = { PLATFORM_HINTS = {
"whatsapp": ( "whatsapp": (
"You are on a text messaging communication platform, WhatsApp. " "You are on a text messaging communication platform, WhatsApp. "

3
hermes_code/run_agent.py
View file

@ -73,6 +73,7 @@ from hermes_constants import OPENROUTER_BASE_URL
from agent.prompt_builder import ( from agent.prompt_builder import (
DEFAULT_AGENT_IDENTITY, PLATFORM_HINTS, DEFAULT_AGENT_IDENTITY, PLATFORM_HINTS,
MEMORY_GUIDANCE, SESSION_SEARCH_GUIDANCE, SKILLS_GUIDANCE, MEMORY_GUIDANCE, SESSION_SEARCH_GUIDANCE, SKILLS_GUIDANCE,
BROWSER_CAPTCHA_GUIDANCE,
) )
from agent.model_metadata import ( from agent.model_metadata import (
fetch_model_metadata, fetch_model_metadata,
@ -2279,6 +2280,8 @@ class AIAgent:
tool_guidance.append(SESSION_SEARCH_GUIDANCE) tool_guidance.append(SESSION_SEARCH_GUIDANCE)
if "skill_manage" in self.valid_tool_names: if "skill_manage" in self.valid_tool_names:
tool_guidance.append(SKILLS_GUIDANCE) tool_guidance.append(SKILLS_GUIDANCE)
if "internet_browser" in self.valid_tool_names and "to_captcha" in self.valid_tool_names:
tool_guidance.append(BROWSER_CAPTCHA_GUIDANCE)
if tool_guidance: if tool_guidance:
prompt_parts.append(" ".join(tool_guidance)) prompt_parts.append(" ".join(tool_guidance))