Experimental features

Agent Studio includes the following experimental features.

These features are experimental and change regularly. Share your feedback. Algolia may promote features to stable APIs based on usage and your feedback.

Default configuration settings

Agent Studio uses default settings optimized for performance and cost-efficiency.

Cache enabled: reduces LLM costs. For more information, see data retention.
Usage metrics: on-demand.

Customize these settings for each agent in the config object.

Available features

Configure features in your agent’s config object from the API.

`date_aware`

Injects current date (YYYY-MM-DD) into the system prompt.

JSON

{
  "config": {
    "features": ["date_aware"]
  }
}

Use case: news, events, daily promotions where the agent must be aware of the current date.

`datetime_aware`

Injects current date and time (YYYY-MM-DD HH:MM) into the system prompt.

{
  "config": {
    "features": ["datetime_aware"]
  }
}

Use case: support agents at scale needing minute-level freshness. Helps reduce repeated queries within a 60-second caching window.

`citations`

Adds citation markers [1], [2] to responses based on search results from tool calls.

JSON

{
  "config": {
    "features": ["citations"]
  }
}

Works in non-streaming mode only. If you require advanced citation capabilities, contact the Algolia support team to inquire about early access to the Citations v2 feature.

Search prefetch

Searches your index with the latest user message, before the model runs. When the hits answer the question, such as “waterproof hiking boots for kids”, the model replies right away. That’s often faster. Unlike the features in the features array, it has its own searchPrefetch property.

JSON

{
  "config": {
    "searchPrefetch": true
  }
}

For use cases and configuration options, see Search prefetch.

Cache tradeoffs

Time awareness features affect caching. For more information, see Date awareness and caching.

Client-side time injection

For full control over format, granularity, and timezone, inject time yourself:

JavaScript

import { useChat } from '@ai-sdk/react';

const { append } = useChat({
  // ... transport config
});

const askWithTime = (question) => {
  const now = new Date();
  const context = `[Current time: ${now.toISOString()}]\n\n`;
  append({ role: 'user', content: context + question });
};

Search and Discovery platform

Optimization and Personalization

AI-powered experiences

Production and scale

Experimental features

Default configuration settings

Available features

`date_aware`

`datetime_aware`

`citations`

Search prefetch

Cache tradeoffs

Client-side time injection

​Default configuration settings

​Available features

​date_aware

​datetime_aware

​citations

​Search prefetch

​Cache tradeoffs

​Client-side time injection

Default configuration settings

Available features

`date_aware`

`datetime_aware`

`citations`

Search prefetch

Cache tradeoffs

Client-side time injection