BerriAI/litellm · error · HTTPException

Set `'STORE_MODEL_IN_DB='True'` in your env to enable this f

Error message

Set `'STORE_MODEL_IN_DB='True'` in your env to enable this feature.

What it means

Saving cache settings via POST /cache/settings additionally requires the STORE_MODEL_IN_DB feature flag: proxy_server reads it with get_secret_bool("STORE_MODEL_IN_DB", ...) and this endpoint refuses with HTTP 500 when it is not exactly True. The gate exists because settings must round-trip through the database, which is only enabled when the deployment opts into DB-stored config. The odd quoting in the message ('STORE_MODEL_IN_DB='True'') is cosmetic in the source string.

Source

Thrown at litellm/proxy/management_endpoints/cache_settings_endpoints.py:601

    This endpoint:
    1. Encrypts sensitive fields (passwords, etc.)
    2. Saves to LiteLLM_CacheConfig table
    3. Reinitializes cache with new settings
    """
    from litellm.proxy.proxy_server import (
        prisma_client,
        proxy_config,
        store_model_in_db,
    )

    if prisma_client is None:
        raise HTTPException(
            status_code=500,
            detail={"error": "Database not connected. Please connect a database."},
        )

    if store_model_in_db is not True:
        raise HTTPException(
            status_code=500,
            detail={"error": "Set `'STORE_MODEL_IN_DB='True'` in your env to enable this feature."},
        )

    try:
        # Read the stored row first: its decrypted values back any credential the
        # caller echoed back redacted, and its key set drives the audit diff.
        existing_row: Final = await _cache_config_table(prisma_client).find_unique(where={"id": "cache_config"})
        before_settings: dict[str, object] | None = None
        saved_settings: dict[str, object] = {}
        if existing_row is not None and existing_row.cache_settings:
            before_settings = _parse_stored_settings(existing_row.cache_settings)
            saved_settings = proxy_config._decrypt_db_variables(variables_dict=before_settings)
        action: Final[AUDIT_ACTIONS] = "updated" if existing_row is not None else "created"

        # Preserve stored secrets behind any redacted or omitted credential, then
        # resolve the url-vs-discrete-fields precedence.
        cache_settings: Final = _resolve_cache_url_precedence(_merge_over_saved(request.cache_settings, saved_settings))

View on GitHub (pinned to 77b7c6c40c)

Solutions

  1. Set the env var: export STORE_MODEL_IN_DB="True" (or add it to the container/deployment env)
  2. Equivalently enable it in the config under general_settings: store_model_in_db: true, then restart the proxy
  3. Confirm both requirements hold: DATABASE_URL set AND STORE_MODEL_IN_DB=True, then retry POST /cache/settings

Example fix

# before
export DATABASE_URL="postgresql://..."
litellm  # POST /cache/settings -> 500 Set 'STORE_MODEL_IN_DB=True'

# after
export DATABASE_URL="postgresql://..."
export STORE_MODEL_IN_DB="True"
litellm  # POST /cache/settings -> 200
Defensive patterns

Strategy: validation

Validate before calling

import os

# The flag must be exactly True (string 'True' from env) on the proxy side
flag = os.getenv("STORE_MODEL_IN_DB", "False")
assert flag.lower() == "true", f"STORE_MODEL_IN_DB={flag!r}; set it to True to save cache settings"

Type guard

def store_model_in_db_enabled(value: object) -> bool:
    return isinstance(value, str) and value.strip().lower() == "true"

Try / catch

import httpx

try:
    httpx.post(f"{PROXY_URL}/cache/settings", json=settings, headers=hdrs).raise_for_status()
except httpx.HTTPStatusError as e:
    if e.response.status_code == 500 and "STORE_MODEL_IN_DB" in e.response.text:
        raise RuntimeError("enable STORE_MODEL_IN_DB=True on the proxy, then retry") from e
    raise

Prevention

When it happens

Trigger: POST /cache/settings on a proxy where DATABASE_URL is set but STORE_MODEL_IN_DB is unset/false; enabling the flag only in config but not env (or vice versa depending on how the deployment reads secrets); forgetting the flag after a redeploy.

Common situations: Standing up the Admin UI cache editor for the first time; migrations of docker-compose files that drop env entries; flag set as string "True" vs boolean in YAML causing get_secret_bool to parse false.

Understand the failure class

Background: "environment variable is not set" and "Missing keys in environment" errors: what missing required env var messages mean and how to fix them — this error's family across 28 libraries.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/dd09f26da9854979. Report an issue: GitHub.