microsoft/autogen · error · NotImplementedError

Error: {query.mime_type} is not supported. Only MemoryMimeTy

Error message

Error: {query.mime_type} is not supported. Only MemoryMimeType.TEXT, MemoryMimeType.JSON, MemoryMimeType.MARKDOWN are currently supported.

What it means

On the semantic (non-sequential) query path, RedisMemory must turn the query into a string prompt for embedding. It accepts MemoryContent only with TEXT, MARKDOWN, or JSON mime types; anything else (IMAGE, BINARY) raises NotImplementedError listing the supported types.

Source

Thrown at python/packages/autogen-ext/src/autogen_ext/memory/redis/_redis_memory.py:316

            raise ValueError(
                "Non-sequential queries cannot be run with an underlying sequential RedisMemory. Set sequential=False in RedisMemoryConfig to enable semantic memory querying."
            )
        elif sequential or self.config.sequential:
            results = self.message_history.get_recent(
                top_k=top_k,
                raw=False,
            )
        else:
            # get the query string, or raise an error for unsupported MemoryContent types
            if isinstance(query, str):
                prompt = query
            elif isinstance(query, MemoryContent):
                if query.mime_type in (MemoryMimeType.TEXT, MemoryMimeType.MARKDOWN):
                    prompt = str(query.content)
                elif query.mime_type == MemoryMimeType.JSON:
                    prompt = serialize(query.content)
                else:
                    raise NotImplementedError(
                        f"Error: {query.mime_type} is not supported. Only MemoryMimeType.TEXT, MemoryMimeType.JSON, MemoryMimeType.MARKDOWN are currently supported."
                    )
            else:
                raise TypeError("'query' must be either a string or MemoryContent")

            results = self.message_history.get_relevant(  # type: ignore
                prompt=prompt,  # type: ignore[reportArgumentType]
                top_k=top_k,
                distance_threshold=distance_threshold,
                raw=False,
            )

        memories: List[MemoryContent] = []
        for result in results:  # type: ignore[reportUnkownVariableType]
            metadata = deserialize(result["metadata"])  # type: ignore[reportArgumentType]
            mime_type = MemoryMimeType(metadata.pop("mime_type"))
            if mime_type in (MemoryMimeType.TEXT, MemoryMimeType.MARKDOWN):
                memory_content = result["content"]  # type: ignore[reportArgumentType]

View on GitHub (pinned to 027ecf0a37)

Solutions

1. Query with a plain string or a TEXT/MARKDOWN/JSON MemoryContent containing the text to embed.
2. For image-based retrieval, first produce a textual description and use that as the query.
3. Centralize a to_query_text() helper so all query inputs are normalized to strings.

Example fix

# before
result = await memory.query(MemoryContent(content=img, mime_type=MemoryMimeType.IMAGE))  # NotImplementedError

# after
result = await memory.query('notes about the architecture diagram')

from autogen_core.memory import MemoryContent, MemoryMimeType

QUERYABLE = (MemoryMimeType.TEXT, MemoryMimeType.MARKDOWN, MemoryMimeType.JSON)

def query_ok(q) -> bool:
    return isinstance(q, str) or (isinstance(q, MemoryContent) and q.mime_type in QUERYABLE)

def is_queryable_content(q) -> bool:
    return isinstance(q, MemoryContent) and q.mime_type in (MemoryMimeType.TEXT, MemoryMimeType.MARKDOWN, MemoryMimeType.JSON)

try:
    res = await memory.query(q)
except NotImplementedError as e:
    raise ValueError(f'convert query to text first: {e}') from e

When it happens

Trigger: await redis_memory.query(MemoryContent(content=Image(...), mime_type=MemoryMimeType.IMAGE)) on a sequential=False memory; querying with a binary MemoryContent; passing a custom mime enum value.

Common situations: Passing the agent's last multimodal message directly as the query; building query MemoryContent from uploaded files without converting to text; mixing sequential and semantic code paths where the sequential path never validates mime type.

Related errors

• Error: {content.mime_type} is not supported. Only MemoryMime
• Unsupported content type: {mime_type}
• Unsupported content type: {mime_type}
• Non-sequential queries cannot be run with an underlying sequ
• 'query' must be either a string or MemoryContent

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/bb85298c4b74f697. Report an issue: GitHub.