oobabooga/textgen · error · InvalidRequestError

messages: invalid content item format

Error message

messages: invalid content item format

What it means

When a message's 'content' is a list (multimodal OpenAI format), each item must be a dict containing a 'type' key. Items that are not dicts, or dicts without 'type', raise InvalidRequestError (400, param='messages') before the content is flattened for the model.

Source

Thrown at modules/api/completions.py:515

        if 'role' not in m:
            raise InvalidRequestError(message="messages: missing role", param='messages')
        elif m['role'] == 'function':
            raise InvalidRequestError(message="role: function is not supported.", param='messages')

        # Handle multimodal content validation
        content = m.get('content')
        if content is None:
            # OpenAI allows content: null on assistant messages when tool_calls is present
            if m['role'] == 'assistant' and m.get('tool_calls'):
                m['content'] = ''
            else:
                raise InvalidRequestError(message="messages: missing content", param='messages')

        # Validate multimodal content structure
        if isinstance(content, list):
            for item in content:
                if not isinstance(item, dict) or 'type' not in item:
                    raise InvalidRequestError(message="messages: invalid content item format", param='messages')
                if item['type'] not in ['text', 'image_url']:
                    raise InvalidRequestError(message="messages: unsupported content type", param='messages')
                if item['type'] == 'text' and 'text' not in item:
                    raise InvalidRequestError(message="messages: missing text in content item", param='messages')
                if item['type'] == 'image_url' and ('image_url' not in item or 'url' not in item['image_url']):
                    raise InvalidRequestError(message="messages: missing image_url in content item", param='messages')

    # Chat Completions
    object_type = 'chat.completion' if not stream else 'chat.completion.chunk'
    created_time = int(time.time())
    cmpl_id = "chatcmpl-%d" % (int(time.time() * 1000000000))
    resp_list = 'data' if is_legacy else 'choices'

    # generation parameters
    generate_params = process_parameters(body, is_legacy=is_legacy)
    if stop_event is not None:
        generate_params['stop_event'] = stop_event
    continue_ = body['continue_']

View on GitHub (pinned to ed888c71f2)

Solutions

  1. Wrap every list item with an explicit type: {"type": "text", "text": ...} or {"type": "image_url", "image_url": {"url": ...}}.
  2. For plain text, prefer a string content instead of a single-element list.
  3. Validate content-list shape client-side before sending.

Example fix

# before
{"role": "user", "content": ["describe this", {"image_url": {"url": u}}]}

# after
{"role": "user", "content": [{"type": "text", "text": "describe this"}, {"type": "image_url", "image_url": {"url": u}}]}
Defensive patterns

Strategy: validation

Validate before calling

def valid_content_items(content: list) -> bool:
    return all(isinstance(i, dict) and isinstance(i.get('type'), str) for i in content)

Type guard

def is_content_item(i) -> bool:
    return isinstance(i, dict) and 'type' in i and i['type'] in ('text', 'image_url')

Try / catch

try:
    resp = client.chat.completions.create(model=m, messages=msgs)
except openai.BadRequestError as e:
    if 'invalid content item format' in str(e):
        msgs = [{'role': m['role'], 'content': repair_items(m['content']) if isinstance(m.get('content'), list) else m.get('content', '')} for m in msgs]
        resp = client.chat.completions.create(model=m, messages=msgs)
    else:
        raise

Prevention

When it happens

Trigger: POST /v1/chat/completions with {"role": "user", "content": ["hello"]} (bare string item) or {"role": "user", "content": [{"text": "hello"}]} (missing 'type').

Common situations: Hand-building multimodal arrays and forgetting the {'type': 'text', ...} wrapper; mixing OpenAI vision format with Anthropic/Llama-style content blocks; old client libs that emitted text-only arrays.

Related errors

AI-assisted analysis of oobabooga/textgen@ed888c71f2 (2026-08-15). Data as JSON: /api/errors/016a31b4b9e461a7. Report an issue: GitHub.