shareAI-lab/learn-claude-code · error · ValueError

Unknown memory type: {mem_type}

Error message

Unknown memory type: {mem_type}

What it means

Raised by write_memory_file() in s09_memory/code.py:153 when mem_type is not one of the allowed MEMORY_TYPES: 'user', 'feedback', 'project', 'reference'. Memory records are tagged with a type in their YAML front matter and the index/rebuild logic groups by it, so unknown types would silently fragment the store. The membership test is an exact, case-sensitive tuple check, so 'User' or 'USER' fail just like 'preference'.

Source

Thrown at s09_memory/code.py:153

        ) == normalized_description:
            return False
        if _normalized_memory_text(str(memory.get("body", ""))) == normalized_body:
            return False
    return True

def memory_document(name: str, mem_type: str, description: str, body: str) -> str:
    metadata = yaml.safe_dump(
        {"name": name, "description": description, "type": mem_type},
        sort_keys=False,
        allow_unicode=True,
    ).strip()
    return f"---\n{metadata}\n---\n\n{body.strip()}\n"

def write_memory_file(name: str, mem_type: str, description: str, body: str) -> Path:
    if not name.strip():
        raise ValueError("Memory name cannot be empty")
    if mem_type not in MEMORY_TYPES:
        raise ValueError(f"Unknown memory type: {mem_type}")
    if not description.strip() or not body.strip():
        raise ValueError("Memory description and body cannot be empty")

    MEMORY_DIR.mkdir(parents=True, exist_ok=True)
    path = memory_path(f"{memory_slug(name)}.md")
    path.write_text(memory_document(name, mem_type, description, body))
    rebuild_memory_index()
    return path

def rebuild_memory_index() -> None:
    MEMORY_DIR.mkdir(parents=True, exist_ok=True)
    lines = []
    for path in sorted(MEMORY_DIR.glob("*.md")):
        if path.name == MEMORY_INDEX.name:
            continue
        try:
            path = memory_path(path.name)
        except ValueError:

View on GitHub (pinned to 985456f4ad)

Solutions

1. Use one of the exact allowed values: 'user', 'feedback', 'project', 'reference' (import MEMORY_TYPES and choose from it).
2. Map free-form or capitalized types to the closest allowed type before writing, defaulting to 'reference'.
3. If a genuinely new category is needed, extend the MEMORY_TYPES tuple deliberately and rebuild the index.

Example fix

# before
write_memory_file(name, 'preference', description, body)

# after
from s09_memory.code import MEMORY_TYPES
mem_type = mem_type.lower() if mem_type in MEMORY_TYPES else 'reference'
write_memory_file(name, mem_type, description, body)

from s09_memory.code import MEMORY_TYPES

def pick_memory_type(candidate: str) -> str:
    return candidate if candidate in MEMORY_TYPES else 'reference'

def is_known_memory_type(mem_type) -> bool:
    from s09_memory.code import MEMORY_TYPES
    return isinstance(mem_type, str) and mem_type in MEMORY_TYPES

try:
    write_memory_file(name, mem_type, description, body)
except ValueError as e:
    if 'Unknown memory type' in str(e):
        write_memory_file(name, 'reference', description, body)
    else:
        raise

When it happens

Trigger: Calling write_memory_file(name, 'preference', ...) — 'preference' is not in the tuple; passing a capitalized variant like 'Feedback'; passing a type string obtained from an LLM response that invented a new category; passing None or a non-string.

Common situations: LLM-driven memory capture where the model free-forms a type outside the allowed vocabulary; hand-written scripts that assume an intuitive-but-wrong type name; version drift if you upgrade code that once accepted arbitrary types.

Related errors

• The memory index is not a memory record
• Memory name cannot be empty
• Memory description and body cannot be empty
• Invalid task status: {task.status}
• Max 20 todos allowed

AI-assisted analysis of shareAI-lab/learn-claude-code@985456f4ad (2026-08-14). Data as JSON: /api/errors/6092e754d8b5aee5. Report an issue: GitHub.