mem0ai/mem0 · error · ValueError

Invalid memory action: {memory_action}

Error message

Invalid memory action: {memory_action}

What it means

Raised by VertexAIEmbedding.embed when memory_action is a value not in the class's embedding_types mapping (which maps mem0 actions like 'add'/'search' to Vertex task types such as RETRIEVAL_DOCUMENT/QUESTION_ANSWERING). Passing None is allowed (defaults to SEMANTIC_SIMILARITY); any other unknown string is rejected before the API call.

Source

Thrown at mem0/embeddings/vertexai.py:57

                    "Google application credentials JSON is not provided. Please provide a valid JSON path or set the 'GOOGLE_APPLICATION_CREDENTIALS' environment variable."
                )

        self.model = TextEmbeddingModel.from_pretrained(self.config.model)

    def embed(self, text, memory_action: Optional[Literal["add", "search", "update"]] = None):
        """
        Get the embedding for the given text using Vertex AI.

        Args:
            text (str): The text to embed.
            memory_action (optional): The type of embedding to use. Must be one of "add", "search", or "update". Defaults to None.
        Returns:
            list: The embedding vector.
        """
        embedding_type = "SEMANTIC_SIMILARITY"
        if memory_action is not None:
            if memory_action not in self.embedding_types:
                raise ValueError(f"Invalid memory action: {memory_action}")

            embedding_type = self.embedding_types[memory_action]

        text_input = TextEmbeddingInput(text=text, task_type=embedding_type)
        embeddings = self.model.get_embeddings(texts=[text_input], output_dimensionality=self.config.embedding_dims)

        return embeddings[0].values

    def embed_batch(self, texts, memory_action="add"):
        if not texts:
            return []
        embedding_type = "SEMANTIC_SIMILARITY"
        if memory_action is not None:
            if memory_action not in self.embedding_types:
                raise ValueError(f"Invalid memory action: {memory_action}")
            embedding_type = self.embedding_types[memory_action]
        all_embeddings = []
        for i in range(0, len(texts), 250):

View on GitHub (pinned to 001c235229)

Solutions

1. Pass only None or the actions defined in the provider's embedding_types mapping (add, search, update)
2. Upgrade mem0 so the provider's mapping matches the actions used by the memory layer
3. If you call embed() directly for generic embedding, omit memory_action entirely

Example fix

// before
vec = embedder.embed("hello", memory_action="index")  # ValueError

# after
vec = embedder.embed("hello", memory_action="add")
# or omit: vec = embedder.embed("hello")

VALID_ACTIONS = {None, "add", "search", "update"}  # provider's embedding_types keys
assert memory_action in VALID_ACTIONS, f"unsupported memory_action: {memory_action!r}"

from typing import Optional

VALID_ACTIONS = {"add", "search", "update"}

def is_valid_action(a) -> bool:
    return a is None or a in VALID_ACTIONS

try:
    vec = embedder.embed(text, memory_action=action)
except ValueError as e:
    if "Invalid memory action" in str(e):
        vec = embedder.embed(text)  # fall back to default task type
    else:
        raise

When it happens

Trigger: Calling embed(text, memory_action="delete") or a custom string; a mem0-internal caller passing a new action type this provider never mapped; user code invoking the embedder directly with an arbitrary label.

Common situations: Direct use of VertexAIEmbedding outside Memory; version skew where a newer mem0 passes an action this provider version does not know; typos in the action string.

Related errors

• Vertex AI embed_batch() returned {len(all_embeddings)} embed
• Invalid memory action: ${memoryAction}
• `model` parameter is required
• `model` must be an instance of Embeddings
• LM Studio embed_batch() returned {len(embeddings)} embedding

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/59fefdcc750aeb11. Report an issue: GitHub.