ypollak2/llm-router

Team Dashboard + Multi-Channel Push

New Features

Team Dashboard — See your team's LLM cost savings across projects with multi-channel push notifications:

llm-router team report              # Print savings dashboard
llm-router team push                # Push to Slack/Discord/Telegram
llm-router team setup               # Interactive webhook wizard

New MCP Tools

• llm_team_report — ASCII savings table with call counts, free-tier %, top models
• llm_team_push — Push to configured webhook (37 total tools)

Multi-Channel Push (auto-detected from URL)

• 🟣 Slack: Block Kit with header, fields, model list
• 🟦 Discord: Rich embed with color and fields
• 📱 Telegram: MarkdownV2 with proper escaping
• 🔗 Generic: Raw JSON POST for custom webhooks

Ollama Always Injected
Ollama is now always prepended to routing chains when OLLAMA_BASE_URL is set — it's free, so there's no reason to ever skip it.

Smart Enforcement Mode (default for new users)

• Q&A tasks (query, research, generate, analyze) → hard block until llm_* called
• Code tasks → soft (file tools are allowed for actual editing)

llm-router set-enforce hard    # Block all tools until routing used
llm-router set-enforce smart   # Default: hard for Q&A, soft for code
llm-router set-enforce soft    # Suggest routing but never block
llm-router set-enforce off     # Disable enforcement entirely

Upgrade

pip install --upgrade claude-code-llm-router
llm-router install

Configuration

# Set team webhook (Slack, Discord, Telegram, or any HTTP endpoint)
export LLM_ROUTER_TEAM_ENDPOINT="https://hooks.slack.com/..."
export LLM_ROUTER_TEAM_CHAT_ID="..."  # Telegram only