Gonka Proxy Admin

Attempts

0

Checkout / invoice starts

Successful

0

0% conversion

Failed / errors

0

Webhook fails + checkout errors

Revenue (paid)

$0

From confirmed payments

By front & provider

Funnel split: front × provider (Stripe / crypto / balance) × kind. Respects the date filter above.

Front	Provider	Kind	Attempts	Success	Failed	Conv %	Revenue $

Top failure reasons

Front	Provider	Code	Reason	Count

Recent payment events

Newest first — attempts, results and errors across Stripe + crypto.

Date & time	Stage	Status	Front	Provider	Kind	Plan	Amount $	User	Error

Invited users

0

— in last 30 days

Active inviters

0

— users have a code

Signup bonus credited

$0

inviter $0 · invited $0

Commission available

$0

Matured, not yet withdrawn

Commission pending

$0

Still in hold period

Commission lifetime / paid out

$0 / $0

Total accrued / withdrawn

Program defaults

$5 · 5/mo · 20% · 30d

bonus · cap · commission · hold (per-partner overridable)

Set terms for any user (even with no invites):

Invited users by frontend

Which front (mingo / gonka) the invited users signed up on — from users.brand of accounts with referred_by.

Frontend	Invited users	Share

Inviter leaderboard

Everyone who invited ≥1 user. Bonus $ = signup bonus paid (credited to the user's balance). Available = matured commission ready to pay out. Use Terms to customise a partner and Withdraw to record a payout.

Inviter	Code	Invited	By frontend	Bonus $	Comm. lifetime	Available	Pending	Terms

Activity log

Recent referral signups, newest first (one row per credited referral event).

Date & time	Invited user	Frontend	Inviter	Code used

Create a promo code

Codes do nothing until active. kind: balance (credit USD) · discount (% off next top-up) · free_sub (N days of a plan).

Code Kind Amount USD Valid until (optional) Max uses (0=∞) Per user Note

Promo codes

Code	Kind	Value	Window	Uses	Active
Loading…

Prepaid USD inference balance per user. Spent is the all-time total from proxy request logs; subscription usage is % of the plan's money allowance (period · 5h window · week).

Body logging defaults: request off, response off

Balance > 0

User	Created	User id	Balance (USD)	Spent (USD)	Last active	Subscription	Body policy

User	Requests	Errors	USD Spent	Tokens

API Key ID	Requests	Errors	USD Spent	Tokens

Errors by category

Category	Status	Count

Current routing

Time filters below apply to statistics tables only; live pool state is always current.

—

Availability timeline

Each strip is an upstream (bold) or one model on it. Cell = share of ~1-min health samples in the bucket where it was available: green = up, red = down, orange = partly, grey = no data or model not served by that upstream. Hover for details.

Loading…

Gonka upstreams (devshards)

Each row is an upstream Gonka inference server (a “devshard”). Health = last background probe of /v1/models; an unhealthy devshard is skipped so user requests don't wait on its timeout. Models = catalog used for model-based routing (live from the devshard, or from static config). Pool = whether it's tried next. 429 cooldown = this upstream returned HTTP 429/5xx and is resting. Cooldowns and failure streaks are tracked per (upstream, model) — the colored dots under the model count show each model's own state, so one broken model doesn't pull the devshard out of the pool for the rest. When every devshard is cooling/unhealthy the proxy still retries all in priority order rather than dropping the request. Note: X-Devshard-ID in logs is a per-request escrow id from the Gonka gateway — a different thing from these upstream servers.

Priority	Devshard	Host	Health	Models	Key (stats)	Pool	429 cooldown	429@up	Test

ML nodes

Host	Key (stats)	Health check	Last error

Traffic by node

From proxy logs; backend_node_key is the node that produced the final response. If Gonka returned 429 on node A and succeeded on node B, the request counts toward B for errors/success — use 429@upstream (all-time, from cluster) to see 429s per URL. paid nodes (e.g. openbroker) are metered — watch their tokens & cost. Tokens/cost/last-used respect the time filter above.

Node	Requests	Errors	Err %	Input tok	Output tok	Total tok	Cost USD	Last used	429@up

Error mix by node

Node key	Category	Client status	Upstream	Count

—

Database (data + indexes)

—

storage + index size (approx.)

Uncompressed data

—

Proxy log entries

—

One document per API call

Avg request / response size

— / —

From request_size_bytes / response_size_bytes

Proxy logs storage

Bodies are optional and truncated in DB; byte sizes reflect full HTTP payload where logged.

Collections

Per-collection document counts and sizes from MongoDB collStats.

Collection	Documents	Data	Storage	Indexes	Avg obj

Priority Waitlist

Submissions from /pricing & the LK Priority signup. Newest first.

Date & time	Kind	Contact	Use case	Account	IP

SMTP

—

Sent (24h)

Sent (7d)

Due now

In progress

Scheduled (future)

Next email

—

Failed

Cancelled

Unsubscribed

Due now and In progress are the numbers that matter — they should both sit near 0 when the worker is keeping up. Scheduled (future) is just upcoming drip mail waiting for its send day, not a backlog, so a large number here is normal.

Drip queue

Scheduled / fired drip jobs across all funnels.

Due at	Template	Status	User	Brand	Attempts	Key

Recently sent (email_log)

Idempotency log — every send that hit SMTP.

Sent at	To	Subject	Key

Delivery attempts (email_deliveries)

Every attempt, including the ones that failed. Verification and reset mail shows up here and nowhere else.

When	To	Kind	Status	Subject	Error

Import a bounce report

For when IMAP is off. Paste the provider's failure notice, drop in a downloaded .eml, or just list addresses one per line. Preview first — nothing changes until you apply.

Address	Severity	Code	Known	Result
Nothing parsed yet.

Suppression list

Unsubscribed / bounced addresses. Drip + lifecycle emails skip these.

Email	Reason	When

Send a test email

Verifies SMTP + DNS auth end-to-end.

Support chat sessions

Website widget conversations. Open one to see the full thread + AI summary, take over from the bot, and reply (the visitor sees it live).

Auto-refresh

Select a session on the left.

Очередь одобрений

Исходящие черновики, ждущие проверки оператора. Отредактируйте текст, затем «Одобрить и отправить» или «Отклонить».

Сводка воронки — —

Сделки по стадиям выбранного сценария.

Канбан

Одна колонка на стадию. Кликните карточку, чтобы открыть детали и историю сообщений.

Найденные каналы

Каналы и группы, найденные агентом; ранжированы по соответствию аудитории.

Название	Юзернейм	Тип	Участники	Fit-скор	Статус	Найден через

Контакты

Люди из каналов — админы, партнёры и прямые контакты. Ссылка Telegram открывает чат, строка — карточку CRM.

Имя	Telegram	Роль	Сделки	Открыть

Воронки сценариев

Одна воронка на сценарий аутрича. Инструкции — ровно то, чему следует агент в сделках этой воронки. Отредактируйте и сохраните, чтобы изменить поведение агента.

Плейбуки (FSM)

Каждый плейбук — детерминированная машина шагов: движок решает, какой шаг следующий, когда и от какой роли/тира персоны, чтобы агент не терял нить. Отредактируйте JSON и сохраните.

Воронки по сценариям

Сделки по стадиям, одна воронка на сценарий.

Расход лимитов за сегодня

Квоты аутрича (с учётом прогрева). Израсходовано / лимит по типам отправки.

Последние события

Поток событий воронки и аудита. Новые сверху.

Когда	Тип	Сделка	Контакт	Данные

Импорт аккаунта

Загрузите <id>_telethon.session продавца + <id>.json (tgdata). Файл _pyrogram.session не нужен. Аккаунт сохраняет свои api_id / устройство / прокси.

ID аккаунта (slug) Статус Файл сессии Telethon (обязателен) Файл tgdata JSON (опционально) Прокси (опционально, важнее JSON)

…или вставьте tgdata JSON текстом (используется, если файл выше не выбран)

Аккаунты Telegram

Пул юзерботов: статус, расход прогрева и прокси. Включайте/выключайте аккаунты и снимайте временные ограничения.

Аккаунт	Статус	Telegram	Телефон	Связь	ЛС сегодня	Группы сегодня	Ограничен до	Прокси	Действия