mem0ai/mem0 · error · ImportError

spaCy is not installed. Install it with: pip install mem0ai[

Error message

spaCy is not installed. Install it with: pip install mem0ai[nlp]

What it means

Raised by _ensure_model_available in mem0.utils.spacy_models when spaCy itself is not importable. spaCy is an optional dependency shipped under the mem0ai[nlp] extra; the NLP-based memory processing paths call this guard before loading models, so hitting it means the NLP feature was enabled without the extra.

Source

Thrown at mem0/utils/spacy_models.py:26

import logging
import threading

logger = logging.getLogger(__name__)

_nlp_full = None
_nlp_lemma = None
_load_failed_full = False
_load_failed_lemma = False
_lock = threading.Lock()


def _ensure_model_available():
    """Download en_core_web_sm if spaCy is installed but model is missing."""
    try:
        import spacy
    except ImportError:
        raise ImportError(
            "spaCy is not installed. Install it with: pip install mem0ai[nlp]"
        )

    if not spacy.util.is_package("en_core_web_sm"):
        logger.info("Downloading spaCy model en_core_web_sm...")
        try:
            from spacy.cli import download

            download("en_core_web_sm")
            logger.info("spaCy model en_core_web_sm downloaded successfully")
        except Exception as e:
            raise RuntimeError(
                f"Failed to download spaCy model en_core_web_sm: {e}. "
                "Please install manually: python -m spacy download en_core_web_sm"
            ) from e


def get_nlp_full():

View on GitHub (pinned to 001c235229)

Solutions

pip install 'mem0ai[nlp]' (quotes matter in zsh; also try mem0ai[nlp,graph] if you use graph memory)
Or install spaCy directly: pip install spacy — then the guard will download the en_core_web_sm model on first use
For Docker: add the extra to the requirements file and rebuild
If you do not need NLP features, disable the processing step that invokes them

Example fix

# before: ImportError: spaCy is not installed ...
from mem0 import Memory
m = Memory.from_config(cfg_with_nlp)

# after
# pip install 'mem0ai[nlp]'
from mem0 import Memory
m = Memory.from_config(cfg_with_nlp)

Defensive patterns

Strategy: validation

Validate before calling

import importlib.util
if importlib.util.find_spec('spacy') is None:
    raise ConfigError('spaCy missing — pip install "mem0ai[nlp]" before enabling NLP processing')

Try / catch

try:
    from mem0.utils.spacy_models import get_nlp_full
    nlp = get_nlp_full()
except ImportError as e:
    if 'spaCy is not installed' in str(e):
        raise ConfigError('install mem0ai[nlp] extra') from e
    raise

Prevention

Install the nlp extra whenever NLP processing is enabled: pip install 'mem0ai[nlp]'
Keep Docker/CI images in parity with local extras
Gate NLP features behind a config flag so minimal deployments do not import them

When it happens

Trigger: Enabling the spaCy/NLP processing features (entity extraction pipelines that call get_nlp_full/get_nlp_lemma) without installing the extra; running in a slim Docker image built from bare mem0ai; CI environments installing only core requirements.

Common situations: Upgrading mem0ai to a version that introduced NLP features while keeping the old dependency list; local dev works (global spaCy) but the deployed image fails; pip resolver skipping the extra because it was quoted wrongly.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/22bf81fbfaa4561d. Report an issue: GitHub.