orka.nodes.memory_writer_node module

class orka.nodes.memory_writer_node.MemoryWriterNode(node_id: str, **kwargs: Any)[source]

Bases: BaseNode

A node that stores information in OrKa’s memory system with intelligent classification.

The MemoryWriterNode is responsible for persisting information in OrKa’s memory backend (RedisStack by default) with automatic memory type classification, vector embeddings for semantic search, and configurable metadata.

Key Features:

Automatic memory type classification (short_term vs long_term)
Vector embeddings for 100x faster semantic search
Configurable metadata and namespaces
Intelligent decay management
Importance-based retention

namespace

Memory namespace for organizing stored data

Type:: str

vector

Whether to enable vector embeddings for semantic search

Type:: bool

metadata

Additional metadata to store with the memory

Type:: dict

memory_type

Override for automatic memory type classification

Type:: str

importance_score

Override for automatic importance scoring

Type:: float

Example:

- id: memory_store
  type: memory-writer
  namespace: knowledge_base
  params:
    vector: true
    metadata:
      source: "user_input"
      confidence: "high"
      timestamp: "{{ now() }}"
  prompt: |
    Store this information:
    {{ input }}

    With metadata:
    Category: {{ previous_outputs.classifier }}
    Confidence: {{ previous_outputs.validator.confidence }}

The node automatically:

Classifies memory as short_term or long_term based on importance
Generates vector embeddings if enabled
Adds system metadata (timestamp, memory type, etc.)
Configures decay based on memory type
Stores in the specified namespace with RedisStack HNSW indexing

async run(context: dict[str, Any]) → dict[str, Any][source]: Write to memory using RedisStack memory logger.