Why are tokens consumed so quickly?
What features consume Tokens?
First, we need to know which features may consume Tokens.
In short, all AI-related features:
- Generating summaries
- Title translation
- Debug Prompt
For AI Summarize and Translate, you can enable or disable them globally or at the feed level.
How is Token consumption composed?
Essentially, it is divided into Input and Output. Input consists of your Prompt and article content. Output is the LLM’s output.
LLM Providers generally charge separately for Input Tokens and Output Tokens, with Output Tokens being several times more expensive than Input Tokens. In Tidyread, to simplify the concept, we do not make this distinction, but when calculating Token consumption, the Output part is multiplied by a price factor (i.e., the price of Output Tokens of the model used divided by Input Tokens).
Why are Tokens consumed so quickly?
After reading the above content, it is not difficult to find that if you feel that tokens are consumed quickly, the reasons might be:
- You are using a lot of AI-related features.
- The input content for AI is large. This could be due to a large article or a complex Prompt.
- The output content from AI is large. This is likely related to the requirements of the Prompt, and sometimes it can be due to model fluctuations (for such issues, we support setting a global maximum Output Token Limit for each call).
How to reduce usage consumption?
1. Pause Anytime When You Don’t Want to Read
We offer granular pause capabilities, allowing you to pause all of Tidyread:
Or pause a specific Recipe:
Or pause a specific Feed:
Note: Pausing the Builder will still fetch feed content in the background and perform AI summarization, but it will not push digests.
2. Enable AI features in a granular way
Enable AI features only for certain sources or for content that meets specific conditions (see Filter feature).
3. Enable Scrape Full Text Option in a granular way
Currently, this option can be set at the feed level, and when enabled, it will scrape the full content:
If you also enable the AI Summary feature, this will increase the Input provided to AI (but you can get higher-quality summaries as a result).
For some unimportant feeds, it is recommended to disable this option.
We are working on supporting more granular control of this option: Automation.
4. Summarize part of the original content
We support the Dynamic Prompt feature, allowing you to summarize only part of the content, thus reducing the input content for AI and thereby reducing usage consumption.
5. Choose different Prompts based on content length
The Dynamic Prompt feature can select different Prompts based on content length. For example, when the content is judged to be short, a simplified Prompt can be used to reduce usage consumption.
6. Usage limits
Currently, the following usage limit features are supported:
- Limit the maximum Input Token for a single AI call.
- Limit the maximum Output Token for a single AI call.
- Limit the total Token usage per month.
- Limit the total Token usage per month by Recipe.
- Limit the total Token usage per day.
- Limit the total Token usage per day by Recipe.
You can set these limits here:
Was this page helpful?