The admins moved the config back to llm.eltex.loc:4000, but the raw
/anthropic path is not set up there (401 "No Anthropic credential";
glm-5.3 is also 403 for the keys). The regular /v1/messages route with
anth-glm-5.3 (glm-5.3 on the Z.ai Anthropic endpoint) covers what we
need, so Claude Code talks to the proxy root again:
- images via Z.ai's server-side analyze_image (Read png ->
server_tool_use analyze_image -> correct answer, same as ai-glm);
- effort works: low has no thinking, medium/high/max do;
- a 309K-token request passes, [1m] keeps the 1M window;
- prompt cache: 94-98% of input is a cache read from the second call.
anth-glm-5.3 goes into every picker slot; zai-glm-5.3 and flash are
not used. Effort stays max by default via CLI --effort.
Checked on 21.10: all 6 pilot keys see the same models and answer on
anth-glm-5.3; the full checks were run on one key.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ai-setup.sh no longer installs Google Android CLI or runs android init,
so menu items 1 and 7 (setup / update) leave it alone. The install
block moved verbatim into scripts/android-cli-setup.sh, which menu item
8 now runs instead of the whole ai-setup.sh.
The final summary of ai-setup.sh points to item 8, README describes the
opt-in install, tests check the new script and that ai-setup.sh does
not install Android CLI any more.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude Code now talks to the raw /anthropic path of the Eltex proxy,
which forwards requests byte-for-byte to the Z.ai Anthropic endpoint,
the same one ai-glm uses directly:
- images work through Z.ai's server-side analyze_image (Read png ->
server_tool_use analyze_image -> correct answer, same as ai-glm);
the regular /v1/messages route converts to the OpenAI endpoint,
where Z.ai rejects images with 400;
- effort reaches Z.ai: low has no thinking, then medium < high < max;
- prompt cache: ~99% of input is a cache read from the second turn,
same as ai-glm.
The key check and model pick still use the proxy root /v1/models.
The model is glm-5.3 in every picker slot with [1m] for the 1M
window; the anth-glm-5.3 / zai-glm-5.3 / flash names are gone.
Effort stays max by default via CLI --effort.
Checked on 20.10 on llm-t2.eltex.loc:4000; admins are moving that
config to llm.eltex.loc:4000, which stays the default URL.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ai-qwen runs Claude Code through Alibaba Cloud Model Studio's
Anthropic-compatible Token Plan endpoint
(token-plan.ap-southeast-1.maas.aliyuncs.com/apps/anthropic). The shared
key is built into the generated script, so it starts right away without
asking for a key.
qwen3.8-max (the smartest model the key has) is in opus/sonnet and for
subagents, --model opus on each start; qwen3.8-flash is in haiku for
background tasks. Measured on 20.10: both models accept up to 983616
input tokens (CLAUDE_CODE_MAX_CONTEXT_TOKENS=1000000, autocompact fires
before the API limit) and see images. Qwen validates
output_config.effort (bogus -> 400), Max shows no trend from low to
max, Flash thinks ~2x longer on max; the launcher starts on --effort max
like ai-mimo, EFFORT_MAPPING.md records the measurements.
setup.sh status, statusline brand color, README, QUICK_START and
AGENTS.md list the new launcher.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Admins now give every key three models. ai-eltex maps them to the
picker slots:
- opus (default): anth-glm-5.3 - glm-5.3 on the Z.ai Anthropic
endpoint: images via analyze_image, effort works, 1M window;
- sonnet: zai-glm-5.3 - glm-5.3 on the OpenAI endpoint, no images or
effort, 200K. Two keys get it as zai-glm-5.3-2-2 (zai-glm-5.3 is 403
for them), so the name is matched exactly or by prefix;
- haiku: zai-glm-5.3-flash - fast, sees images itself, no effort,
200K; Claude Code also uses this slot for background tasks.
A missing zai/flash model falls back to the default one.
Model name is anth-glm-5.3 again: with anth-glm-5.3-5 the auto-logout
would drop every saved key, since no key has that name now.
Context window is per model instead of a global
CLAUDE_CODE_MAX_CONTEXT_TOKENS=1000000: [1m] only on anth-glm-5.3
(Claude Code strips the suffix before the API call), 200K for the
others, otherwise a long flash/zai session would hit "Prompt exceeds
max length" instead of auto-compacting.
Checked on all 6 pilot keys (19.10): default model reads a png via
Read + analyze_image; sonnet/haiku answer with 200K windows; effort
low on anth-glm-5.3 = 68 output tokens (max ~3500).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Z.ai maps xhigh to native high, so this trims the thinking budget for
everyday ai-glm launches; AI_EFFORT=max still forces the old behavior.
Co-Authored-By: Claude Code <noreply@anthropic.com>
ai-mimo runs Claude Code through Xiaomi's Anthropic-compatible API
(Token Plan or pay-as-you-go). The key's endpoint is found by probing
token-plan-sgp, token-plan-ams and api.xiaomimimo.com (a key only works
in its own region) and is saved to mimo_base_url; MIMO_BASE_URL
overrides. Unreachable endpoints are skipped, token-plan-cn is left out
of the probe since it times out from our network, and a 402 on the
pay-as-you-go endpoint explains that a tp- Token Plan key is needed.
Always the smartest model: mimo-v2.6-pro in every slot and for
subagents, --model opus and --effort max on each start, 128000 output
tokens (Claude Code's ceiling), real 1M context window. Measured on
13.10: MiMo ignores output_config.effort (even "bogus" is accepted, no
length trend from low to max) and treats thinking adaptive like
enabled, so reasoning is always on; EFFORT_MAPPING.md records this.
/logout-mimo is registered as a command and handled by the logout hook
(removes mimo_key and mimo_base_url). setup.sh status, statusline brand
color, README, QUICK_START and AGENTS.md list the new launcher.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ai-eltex-qwen: new launcher for the Eltex proxy on Qwen models
(Coder-Senior, then Flash-Next), own key eltex_qwen_key, config dir and
/logout-eltex-qwen hook branch, 200K window.
Effort: ai-glm and ai-eltex start on max via CLI --effort (AI_EFFORT
overrides), CLAUDE_CODE_MAX_OUTPUT_TOKENS=64000 so long GLM answers are
not cut at 32000. EFFORT_MAPPING notes the proxy passes effort for
anth-glm-5.3 since 09.10.
ai-eltex model pick: no longer falls back to the first /v1/models entry,
which twice picked something unusable (zai-glm-5.3 routed to flash, and
the access-group name ZAI-Test-Group -> 400). Without anth-glm-5.3 the
launcher exits with a clear message; ELTEX_MODEL still overrides.
/logout-eltex: Claude Code answered "Unknown command" and never called
the logout hook (the built-in /logout bypasses hooks too). ai-eltex now
registers commands/logout-eltex.md in its config dir, so the hook
receives the command and deletes eltex_key (checked end to end).
ru-bypass: resolve exception hosts via ahostsv4 first.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ctx: in ai-eltex filled up too fast: the launcher capped the window at
200K (CLAUDE_CODE_AUTO_COMPACT_WINDOW from the admins' doc plus
CLAUDE_CODE_DISABLE_1M_CONTEXT), while anth-glm-5.3 is the same glm-5.3
on the Z.ai Anthropic endpoint that ai-glm runs with a 1M window.
A 598K-token request went through the proxy fine (glm-5.3 answered
correctly from the last line). Claude Code now reports contextWindow
1000000.
The 200K cap also made Z.ai's analyze_image turns dangerous: their
usage covers both internal model passes (~x2), e.g. 82.7K -> 166.7K
(83% of 200K), which could trigger an early auto-compact.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Admins moved the pilot to anth-glm-5.3: proxy deployments on the Z.ai
Anthropic endpoint. Checked on all 6 pilot keys: 48/48 requests served
by glm-5.3 (no flash), images recognised via Z.ai analyze_image in
Claude Code 8/8. zai-glm-5.3 goes through an auto-router that swaps in
glm-5.3-flash, so ai-eltex now prefers anth-glm-5.3 when the key has it
(ELTEX_MODEL still overrides).
ENABLE_TOOL_SEARCH is no longer forced: on the anthropic path GLM
searched for Read/Bash via ToolSearch and decided they were missing.
Effort still has no effect: Claude Code sends output_config.effort, but
LiteLLM drops it (direct Z.ai low/max = 81/3712 output tokens, via the
proxy ~3-5K at any level).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ai-eltex runs Claude Code through the Eltex LiteLLM proxy
(http://llm.eltex.loc:4000, team GLM subscription). The personal key
from admins is checked via /v1/models and stored in
~/.config/ai-setup/eltex_key; the model is taken from the key's
allowed list (ELTEX_MODEL / ELTEX_BASE_URL override). The proxy
ignores effort/thinking, so no effort forcing. /logout-eltex and
/logout in ai-eltex drop the key; setup.sh status shows it.
ru-bypass: hosts block now has an end marker and only that block is
rewritten, manual entries outside it survive; llm.eltex.loc/llm2 added
since the VPN DNS does not resolve it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Z.ai Anthropic endpoint reads output_config.effort and accepts all 5
Claude Code levels, but GLM 5.3 has only low/high/max natively:
medium -> low, xhigh -> high (measured, not documented by Z.ai).
ai-glm now starts on xhigh instead of max.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude Code не знает моделей glm-* и для glm-5.3 без суффикса считал окно
= 200K: счётчик ctx: врал в ~5 раз, автокомпакт жал сессию в 200K.
Окно теперь задаётся CLAUDE_CODE_MAX_CONTEXT_TOKENS=1000000 (env применяется
к не-claude моделям независимо от суффикса), поэтому [1m] у Haiku-слота
стал дублирующим и убран — все три слота пикера теперь glm-5.3.
Co-Authored-By: Claude <noreply@anthropic.com>
glm-5.3 доступна на api.z.ai (вышла 2026-08-14, 1M контекст): opus/sonnet =
glm-5.3, haiku = glm-5.3[1m], subagent = glm-5.3, health-check ключа тоже
на 5.3. Секция OpenRouter остаётся на z-ai/glm-5.2 — там 5.3 ещё нет.
Co-Authored-By: Claude <noreply@anthropic.com>
Логин Claude.ai подвисал навсегда: claude auth login ждёт код из браузера в stdin, а хук запускал его в фоне с </dev/null, куда код вставить невозможно. Теперь /add-account поднимает логин со stdin из fifo и печатает ссылку, а /add-account <код> отдаёт код, дожидается обмена и сохраняет новый аккаунт по email из токена. Провал логина больше не маскируется под успешное сохранение текущего аккаунта, а фоновые процессы гасятся строго по своим PID вместо широкого pkill -f. Поведение закреплено пятью тестами в tests/test_fixes.sh.
Удалены все кастомные скиллы из home-configs/claude/skills и логика
их деплоя (секция 6.6 ai-setup.sh для Claude и Gemini). Из README
убраны упоминания про деплой скиллов, из test_fixes.sh — тест наличия
android-скилла. Хуки аккаунтов, команда android init, правило #5 про
reusable skills и симлинк общей папки ~/.claude/skills оставлены
нетронутыми — скиллы теперь устанавливаются из отдельного репозитория.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Пользователь перешёл на Amnezia VPN — встроенный VLESS-прокси (xray + proxychains4)
и отдельный server-installer fuck-rkn.sh (telemt + xray) больше не нужны. Убран
весь код в ai-setup.sh/setup.sh, конфиги с VPN-кредами, неактуальный тест и
упоминания в документации.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Генерация нативного лаунчера ai-cursor для Cursor CLI (agent) с
автоустановкой через cursor.com/install, интеграцией proxychains4 в
VLESS-режиме, обновлением ~/.cursorrules и флагом --force (--yolo).
Очистка старого ai-cursor из ~/.local/bin при нестандартном BIN_DIR.
Добавлены тест генерации и упоминания в EFFORT_MAPPING/QUICK_START/README.
Co-Authored-By: Claude <noreply@anthropic.com>
Новый лаунчер ai-glm запускает Claude Code через официальный Anthropic-совместимый
API Zhipu AI (api.z.ai). Модели: glm-5.2 (opus/sonnet), glm-5.2[1m] (haiku).
Обновлены документация, статус, тесты и EFFORT_MAPPING.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Установка android CLI из dl.google.com для linux/mac/windows
- Запуск android init для деплоя skills в агентов
- Кастомный skill android-cli в home-configs/claude/skills/
- Пункт меню 8) Android CLI в setup.sh
- Тесты и документация
Co-Authored-By: Claude <noreply@anthropic.com>
Раньше все ai-* лаунчеры делили один ~/.claude и общий settings.json, из-за
чего кастомная модель (openai/gpt-5.5) из ai-openrouter протекала в пикер
ai-claude. Теперь каждый сторонний провайдер изолирован в своём
CLAUDE_CONFIG_DIR (~/.config/ai-setup/cfg/<launcher>) - свои settings.json и
.claude.json, ноль протечек. ai-claude остаётся на ~/.claude (нативный логин).
Пикеры /model приведены к требуемому виду:
- ai-deepseek: только DeepSeek V4 Pro (opus) и DeepSeek V4 Flash (haiku),
дефолт Pro; через availableModels + ANTHROPIC_DEFAULT_*_MODEL_NAME
- ai-kimi: только Kimi K2.6 (opus)
- ai-claude: только нативные модели Claude
Общие skills и CLAUDE.md шарятся симлинком из ~/.claude.
Persistence effort - гибрид:
- low/medium/high/xhigh живут нативно в settings.json лаунчера, /effort
внутри сессии работает свободно и уровень сохраняется
- max нельзя сохранить в settings.json (session-only), поэтому он
восстанавливается через CLAUDE_CODE_EFFORT_LEVEL; в такой max-сессии
/effort залочен (ограничение Claude Code), выход - AI_EFFORT=<lvl> ai-*
Текущий уровень ловит статусбар в ~/.cache/ai-setup/effort_<launcher>.
Удалён устаревший effort-save-hook (заменён нативным persistence + гибридом),
почищен из ~/.claude/settings.json и осиротевший кэш model_*.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- все скрипты перенесены в scripts/ (ai-setup.sh, ru-bypass.sh, ks-off.sh, ks-on.sh)
- setup.sh — новый мастер-скрипт с меню: шаги 1-2 отделены от опций 3-4
- пояснения к GATEWAY и DEV при выборе пункта 2
- ai-setup.sh: SCRIPT_DIR поднят на уровень выше (/../) чтобы находить home-configs/
- tests/test_fixes.sh, test_isolated.sh: пути обновлены на scripts/ai-setup.sh
- QUICK_START.md, README.md, home-configs/network/README.md: пути обновлены
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Move GLOBAL_RULES.md to home-configs/ as single source of truth
- Add el-review and el-review-heavy skills for GitLab-style branch diff review
- Update ai-setup.sh to deploy skills to ~/.claude/skills/
- Update README and tests for new paths
- Replace native kimi CLI launcher with Claude Code + Kimi backend
- Update README description and permission flags
- Update test_isolated.sh and tests/test_fixes.sh for new ai-kimi logic
- Add claude-openai-anthropic-proxy.py (Anthropic ↔ OpenAI translator)
- Rewrite claude_kimi to use native kimi CLI instead of Claude wrapper
- Add OpenAI-compatible API test helpers (_claude_test_openai_api, _handle_openai_api_response)
- Replace 127.0.0.1 with localhost for broader compatibility
- Add test for query-string handling in Kimi proxy
- Update README to reflect Artemox API usage
В Claude Code v2.x команда "claude auth login" запускает полный интерактивный TUI
(включая "Welcome to Claude Code" и "Select login method"), а затем функция дополнительно
вызывала claude "$@" — вторая сессия. Пользователь проходил выбор типа аккаунта дважды.
Исправление: ветка [L] теперь запускает claude "$@" напрямую с нужными моделями
и сразу делает return "$?", не допуская повторного вызова claude снизу функции.
Claude сам обрабатывает весь auth flow (браузер → OAuth → выбор аккаунта) за одно
взаимодействие.
Добавлен тест test_fix8_no_double_login (3 проверки), итого 24/24.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Fix 1: export ANTHROPIC_API_KEY при ручном вводе ключа в claude_anthropic ([K]-ветка);
без export subprocess claude не видел ключ и падал с ошибкой авторизации
- Fix 2: trap RETURN в claude_gpt убивает прокси при любом ранем return 1,
устраняя утечку фоновых процессов
- Fix 3: sleep 1 заменён на poll-цикл (10 попыток, curl exit 7 = connection refused);
connection refused теперь не маскируется под «нет сети»
- Fix 4: после codex auth login в claude_gpt добавлена повторная проверка _claude_test_api;
прежде claude запускался без подтверждения успешности reauth
- Fix 5: аналогичная re-валидация в claude_gemini после браузерной авторизации
(ветки 401/403 и 429)
- Fix 6: prompt [c/Q] → [C/q] в 429-обработчике claude_anthropic — заглавная буква
соответствует умолчанию (стандарт файла: CAPITAL = default)
- Fix 7: trap 'rm -rf "$TMP"' EXIT — одинарные кавычки + кавычки вокруг $TMP,
предотвращают word-split при путях с пробелами
Добавлены тесты: tests/test_fixes.sh (21 тест, все проходят)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
