Usage

ChatBotKit tracks your usage across several key dimensions. Understanding these metrics helps you manage costs, monitor performance, and make informed decisions about your subscription plan.

Understanding the Metrics

Tokens

Tokens represent the computational units consumed by your AI models. Each model has different input and output token costs.

How token consumption is calculated:

ChatBotKit records the actual token counts reported by the upstream model provider (such as OpenAI or Anthropic) and applies the formula:

Each model has separate input and output token ratios. For example, if a model has an input token ratio of 0.0893 and output token ratio of 0.5556:

• A request with 400 input tokens and 600 output tokens consumes: (400 × 0.0893) + (600 × 0.5556) = 369.08 CBK tokens

Key details about token billing:

• Upstream provider reporting: ChatBotKit bills based on the actual token counts reported by the model provider, not estimates
• Cached tokens excluded: When a provider supports prompt caching, only non-cached input tokens are counted. Cached tokens are not billed
• Per-conversation accumulation: Within a single conversation, tokens accumulate across all message exchanges. You can control how many prior messages are included in each request using the model's "Interaction Max Messages" setting to manage token growth
• Usage log details: Each conversation provides a line-item breakdown showing input tokens, output tokens, and other components. Summing all line items matches your total usage
• Model-specific pricing: Visit the platform models page to see exact input and output ratios for each supported model

Conversations and Messages

Conversations and messages track interaction volume on your account.

• Conversations: Each unique conversation thread between a user and a bot counts as one conversation
• Messages: Each individual message (user or bot) increments the message count by one
• Monthly allocation: These metrics are allocated monthly and reset at the start of each billing cycle

Additional Factors

Token consumption can increase when using:

• Datasets: Searching datasets adds tokens for retrieval and context inclusion
• Skillsets: Actions like web fetching, text generation, or vision analysis consume additional tokens
• Model settings: Using higher "Max Tokens" or "Interaction Max Messages" settings increases consumption
• File attachments: Conversations with attached files consume tokens for processing

Monitoring Your Usage

Visit the Usage dashboard to view your metrics in real-time. The dashboard shows:

• Current period usage: Tokens, conversations, and messages used in the current billing cycle
• Plan limits: Your subscription's allocation and how close you are to limits
• Historical trends: Usage patterns over time
• Detailed logs: Download raw usage data for analysis and record-keeping

All metrics update in real-time with no delay.

Optimizing Token Usage

Here are practical strategies to reduce token consumption:

Choose more efficient models: Compare token ratios on the models page. Smaller, more efficient models often perform well for specific tasks at lower cost.

Limit conversation history: Lower the "Interaction Max Messages" setting to include fewer prior messages in each request. This is most impactful for long conversations.

Use strategic dataset linking: Connect only necessary datasets to each bot. Only active dataset searches consume tokens.

Minimize skillset actions: Use skillset abilities judiciously. Each action (fetch, text generation, image analysis) adds token consumption.

Monitor context size: Reduce "Max Tokens" settings where appropriate for your use case. Balancing quality against cost matters.

Test before deploying: Use the chat feature on bot and skillset pages to test interactions and understand expected token usage before production deployment.

Plan Considerations

• Approaching limits: When you exceed 80% of your monthly allocation, check your usage patterns and consider your needs
• Consistent overages: If you regularly exceed your allocation, upgrading to a higher tier may be more cost-effective than managing overages
• Scaling gradually: As usage grows, monitor trends and adjust your plan proactively

By understanding these metrics and following optimization practices, you can maximize the value of your ChatBotKit subscription while maintaining the conversational quality your users expect.