Telegram Context

Telegram Context

Provides conversational continuity by fetching recent Telegram messages at session start. When enabled, the agent automatically retrieves message history to maintain context across disconnected sessions.

Credits

Created by @fourthdensity

Commands

• /telegram-context on — Enable automatic history fetching
• /telegram-context off — Disable automatic fetching
• /telegram-context status — Show current settings
• /telegram-context fetch [n] — Manually fetch last n messages (default: 20)

Setup

1. The skill auto-detects Telegram channels — no configuration needed
2. State is stored in memory/telegram-context.json
3. Only activates when the current channel is Telegram

How It Works

When a session starts in Telegram:

1. Check memory/telegram-context.json for enabled state
2. If enabled, fetch recent messages via message tool
3. Provide history as context for the conversation
4. Update lastFetch timestamp

State File

memory/telegram-context.json:

{
  "enabled": true,
  "fetchCount": 20,
  "lastFetch": "2025-01-15T10:30:00Z"
}

Implementation Notes

Command Handlers

Enable/disable:

// Read current state
read: memory/telegram-context.json

// Update state
write: memory/telegram-context.json
{
  "enabled": true/false,
  "fetchCount": 20,
  "lastFetch": "2025-01-15T10:30:00Z"
}

Manual fetch:

message: {
  action: "list",
  limit: 20  // or user-specified count
}
// Provide results as context summary

Session Start Behavior

At the start of each Telegram session:

1. Check if memory/telegram-context.json exists
2. If enabled, call message tool with action: "list"
3. Summarize recent messages for context window
4. Acknowledge continuity to user (optional)

Implementation Scope

The skill uses OpenClaw's built-in message tool with:

• action: "list" — limited to the current Telegram chat only
• No access to other chats, channels, or external Telegram accounts
• Requires the OpenClaw gateway to have Telegram channel permissions already configured

Privacy & Security

Data Handling:

• Only fetches from the current chat (never cross-chats or other Telegram conversations)
• Message content is included in the agent's context window and sent to the configured LLM provider
• No message content is stored in telegram-context.json — only settings and timestamps
• Message content may appear in OpenClaw session logs (depends on your logging configuration)

Recommendations for Sensitive Conversations:

• Use manual fetch (/telegram-context fetch) instead of auto-fetch
• Set low fetchCount (5-10) to minimize context exposure
• Disable the skill entirely when discussing sensitive topics: /telegram-context off
• Be aware that fetched messages become part of the conversation history sent to AI models

User Control:

• Full toggle on/off anytime — no persistence beyond your control
• No external credentials or API keys required
• No binaries installed — pure instruction-based skill

Limitations

• Telegram-only (other channels not supported)
• Requires appropriate message permissions via OpenClaw gateway
• Large history may need summarization to fit context window
• Fetched messages are sent to your configured LLM provider — review your threat model for sensitive conversations

Tips

• Set fetchCount to 10-30 for most use cases (balance context vs. tokens)
• Use /telegram-context fetch 50 when you need deep context for a specific task
• Works best alongside MEMORY.md for long-term persistence