Cache Set¶

Store a value in an in-memory cache by key, using the image's video identifier as a namespace to enable data sharing between workflow steps, caching intermediate results, and avoiding redundant computations within the same workflow execution context.

How This Block Works¶

This block stores values in an in-memory cache that can be later retrieved using the Cache Get block. The block:

1. Receives image, cache key, and value to store:
2. Takes an input image to determine the cache namespace
3. Receives a cache key (string) identifying the cache entry
4. Receives a value (any data type) to store in the cache
5. Determines cache namespace:
6. Extracts video identifier from the image's video metadata
7. Uses the video identifier as the cache namespace (isolates cache entries per video/stream)
8. Falls back to "default" namespace if no video identifier is present
9. Stores value in cache:
10. Accesses the in-memory cache dictionary for the determined namespace
11. Stores the value using the specified key in the cache
12. Overwrites any existing value with the same key (cache keys are unique within a namespace)
13. Returns stored value:
14. Outputs the stored value as a pass-through (same value that was stored)
15. The output can be used by subsequent workflow steps

The cache is namespaced by video identifier, meaning different videos or streams have separate cache storage. This allows workflows processing multiple videos to maintain separate caches for each video. The cache is stored in memory and is cleared when the workflow execution completes or when the block is destroyed. Cache Set must be used in conjunction with Cache Get - values are stored with Cache Set and retrieved with Cache Get using the same key and namespace (determined by the same video identifier).

Common Use Cases¶

• Shared State Between Steps: Store intermediate results in one workflow step for retrieval in another step (e.g., store detection results for later analysis, cache classification predictions for filtering, save metadata for subsequent blocks), enabling state sharing workflows
• Avoid Redundant Computations: Cache expensive computation results for reuse across multiple workflow steps (e.g., cache model predictions, store processed images, save transformation results), enabling computation caching workflows
• Video Frame Context: Maintain context across video frames by storing frame-specific data (e.g., cache previous frame detections, store frame sequence metadata, save tracking state), enabling frame context workflows
• Conditional Workflow Logic: Store decision results or flags that control workflow execution in subsequent steps (e.g., cache filtering decisions, store validation results, save workflow state), enabling conditional execution workflows
• Data Aggregation: Accumulate data across workflow steps by storing values in cache (e.g., store detection counts, cache statistics, save result collections), enabling data aggregation workflows
• Temporary Storage: Use cache as temporary storage for values that need to be accessed by multiple workflow steps without passing through the workflow graph (e.g., store cross-step data, maintain temporary state, share non-linear workflow data), enabling temporary storage workflows

Connecting to Other Blocks¶

This block stores values in cache and passes through the stored value:

• Before Cache Get block to store values that will be retrieved later (e.g., store detections for retrieval, cache predictions for later use, save metadata for access), enabling cache storage workflows
• After model or processing blocks to cache their outputs for later use (e.g., cache model predictions, store processed results, save computation outputs), enabling result caching workflows
• In workflow branches to store shared values accessible from parallel or conditional execution paths (e.g., store shared state, cache common results, save branch data), enabling branch coordination workflows
• Before blocks that need cached data to store values that will be used by subsequent blocks (e.g., store inputs for later processing, cache data for filtering, save values for analysis), enabling cached data workflows
• In conditional logic workflows to store flags or decisions for later use (e.g., store validation results, cache decision flags, save conditional state), enabling conditional logic workflows
• With video processing workflows to maintain frame-specific or video-specific cache namespaces (e.g., store frame context, cache video-specific data, save stream-specific values), enabling video context workflows

Requirements¶

This block requires an input image (used to determine the cache namespace via video identifier), a cache key (string) to identify the cache entry, and a value (any data type) to store. The block only works in LOCAL execution mode - it will raise a NotImplementedError if used in other execution modes. Values stored in the cache can be retrieved later using the Cache Get block with the same key and namespace (same video identifier). The cache is stored in memory and is automatically cleared when the workflow execution completes. The cache is namespaced by video identifier, so different videos have separate cache storage. If a key already exists in the cache, storing a new value with the same key will overwrite the previous value. The stored value can be any data type (strings, numbers, lists, detections, images, etc.).

Type identifier¶

Use the following identifier in step "type" field: roboflow_core/cache_set@v1to add the block as as step in your workflow.

Properties¶

The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.

Available Connections¶

Input and Output Bindings¶

The available connections depend on its binding kinds. Check what binding kinds Cache Set in version v1 has.

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/cache_set@v1",
    "image": "$inputs.image",
    "key": "my_cache_key",
    "value": "any_value"
}