Jev Anti-Spam Bot
Delete only high-confidence Telegram spam while keeping exemptions, thresholds, history, and failure behavior explicit.
What Jev decides
Whether each spam pattern applies and which contiguous recent messages are linked to the current spam campaign.
The grammY bot asks Jev a fixed battery of spam questions for each eligible message, checks a local threshold, identifies the linked suffix of a multi-message campaign, and fails open when classification or deletion is unavailable.
- Jev role
- Estimate several kinds of spam and identify which recent messages belong to the same campaign.
- Input
- The current Telegram text or caption plus up to ten recent messages scoped to the same chat and sender.
- Output
- Noul probabilities for spam signals and recent-message linkage.
- Decision frequency
- One classification request per eligible message, followed by linkage decisions after a threshold match.
- Fallback
- Fail open and keep messages when Jev or Telegram is unavailable; exempt verified administrators and enforce deletion only after local threshold checks.
How it works
Overview
Jev Anti-Spam Bot moderates new and edited Telegram group messages and media captions. It skips verified group administrators and the receiving group own linked channel, while still analyzing external channel identities. One Jev request evaluates eighteen spam signals and one linkage question for each recent message. Short-lived per-chat and per-sender history lets the bot recognize campaigns split across several posts.
When any spam signal reaches the configured threshold, a second decision identifies the contiguous recent messages belonging to the same campaign, and the bot deletes only that suffix plus the current message. The model version and threshold are pinned in code. If Jev, Telegram, or optional statistics storage fails, messages remain untouched and the moderation path continues fail-open.
- Language
- TypeScript
- License
- MIT
- Author
- backmeupplz
Sources
Related projects
fast-jev-compaction
Prune stale coding-agent tool history without rewriting the context that remains.
What Jev decides
Whether each old tool call and its result should remain available.
typesafe-computer-use
A native macOS computer-use loop built from local OCR, accessibility data, and bounded Jev actions.
What Jev decides
Which computer action, visible item, site, or off-screen accessibility control to use next.
Jev Chat Assistant
An Android conversation copilot that uses Jev to interpret chats and rank reply candidates without sending for the user.
What Jev decides
The likely intent, conversation risk, reply need, best next action, unresolved tension, and which of three candidate replies fits best.