invoke-ai/InvokeAI · error · DuplicateModelException

Model key '{key}' already exists. Provide a different key to

Error message

Model key '{key}' already exists. Provide a different key to install this external model.

What it means

_register_external_model() checks the record store before registering an external (provider-API) model under a caller-supplied key. If a model with that key already exists, it raises DuplicateModelException asking for a different key. This guard exists because external model keys are caller-chosen, so collisions must be caught explicitly.

Source

Thrown at invokeai/app/services/model_install/model_install_default.py:1036

        existing_external = next(
            (
                model
                for model in self.record_store.search_by_attr(
                    base_model=BaseModelType.External, model_type=ModelType.ExternalImageGenerator
                )
                if isinstance(model, ExternalApiModelConfig)
                and model.provider_id == provider_id
                and model.provider_model_id == provider_model_id
            ),
            None,
        )

        if existing_external is not None:
            key = existing_external.key
        else:
            try:
                self.record_store.get_model(key)
                raise DuplicateModelException(
                    f"Model key '{key}' already exists. Provide a different key to install this external model."
                )
            except UnknownModelException:
                pass

        config = ExternalApiModelConfig(
            key=key,
            name=name,
            description=job.config_in.description,
            provider_id=provider_id,
            provider_model_id=provider_model_id,
            capabilities=capabilities,
            default_settings=default_settings,
            source=str(job.source),
            source_type=MODEL_SOURCE_TO_TYPE_MAP[job.source.__class__],
            path="",
            hash="",
            file_size=0,

View on GitHub (pinned to 0b6a024f2f)

Solutions

Omit the 'key' field in ModelRecordChanges and let the service generate a unique key.
Generate a fresh unique key (e.g. from the source id + hash) before installing.
If the model is intentionally the same, look it up first and skip re-installation.
Use install_or_update-style flows that update an existing model instead of registering a duplicate.

Example fix

// before
config = ModelRecordChanges(source=source, key='my-sdxl')  # collides
// after
config = ModelRecordChanges(source=source)  # let service assign a unique key
service.install(config)

Defensive patterns

Strategy: try-catch

Validate before calling

from invokeai.app.services.shared.sqlite.sqlite_common import UnknownModelException
key_collision = False
try:
    service.record_store.get_model(key)
    key_collision = True
except UnknownModelException:
    pass

Try / catch

try:
    service.install(ModelRecordChanges(source=src, key=key))
except DuplicateModelException:
    existing = service.record_store.get_model(key)  # already installed; update or skip

Prevention

Omit the 'key' field and let the service generate unique keys.
Treat installs as idempotent: look up before inserting.
Avoid hard-coded keys shared across scripts/environments.
Serialize external-model installs to avoid concurrent key races.

When it happens

Trigger: Calling _register_or_install/install with ModelRecordChanges containing a 'key' that already exists in the record store; installing the same external model twice with the same hard-coded key; restoring from a store that already contains the key.

Common situations: Scripts that install external models idempotently but reuse a fixed key without checking first; two concurrent installs with the same key; copying install commands between environments sharing a database.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/baf1c6d55c3357b5. Report an issue: GitHub.