parse_vision_messages fails to work correctly when encountering an array content.

[growmuye]

🐛 Describe the bug

在非vision模式下，使用AsyncMemory await memory.add(messages) messages内容如下

[ {
  "role" : "user",
  "content" : "你好"
}, {
  "role" : "user",
  "content" : [ {
    "type" : "text",
    "text" : "For context:"
  }, {
    "type" : "text",
    "text" : "xxxxxxx"
  } ]
} ]

在openai官网，他可能是个 text array

https://platform.openai.com/docs/api-reference/chat/create

[parshvadaftari]

Hey @growmuye sorry for the trouble caused, can you share when and how exactly this error occurs?

[Anand0295]

Cause:

• The parse_vision_messages function doesn't handle array-type message content (e.g. [{'type': 'text', 'text': ...}] as allowed by OpenAI API). Instead, it expects a simple string, so parsing fails and downstream output files like memories/{idx}.txt are never created.

Solution:

• Update the parsing logic to flatten/join array-type content into a single string:

if isinstance(content, list):
    content_str = ''.join([ch['text'] for ch in content if 'text' in ch])
else:
    content_str = content

• After this change, the pipeline should process messages correctly and create the expected files.

Let me know if you want a PR or code sample for the fix!