tirth8205/code-review-graph · error · ValueError

Embedding refresh requires an explicit provider and model.

Error message

Embedding refresh requires an explicit provider and model.

What it means

refresh_embeddings() requires both provider and model explicitly because refresh must never guess which embedding backend to rebuild with — a wrong guess would silently migrate the index to incompatible vectors.

Source

Thrown at code_review_graph/embeddings.py:1341

    *,
    provider: str,
    model: str,
) -> dict[str, int] | None:
    """Refresh a previously embedded graph under one exact provider identity.

    This function is deliberately not called by default build paths.  Callers
    must supply both provider and model explicitly.  A graph with no existing
    vectors returns before provider resolution, so routine builds cannot load
    a local model, contact a cloud service, or incur API cost.

    Existing vectors must all use the identity resolved from the requested
    provider/model (including the endpoint for OpenAI-compatible providers).
    Refresh never silently migrates an index to another model or endpoint.
    """
    provider = provider.strip().lower()
    model = model.strip()
    if not provider or not model:
        raise ValueError(
            "Embedding refresh requires an explicit provider and model.",
        )

    has_table = graph_store._conn.execute(
        "SELECT 1 FROM sqlite_master "
        "WHERE type = 'table' AND name = 'embeddings'",
    ).fetchone()
    if has_table is None:
        return None
    has_rows = graph_store._conn.execute(
        "SELECT 1 FROM embeddings LIMIT 1",
    ).fetchone()
    if has_rows is None:
        return None
    try:
        rows = graph_store._conn.execute(
            "SELECT DISTINCT provider FROM embeddings ORDER BY provider",
        ).fetchall()

View on GitHub (pinned to b58668751a)

Solutions

1. Pass explicit non-empty provider and model, e.g. refresh_embeddings(gs, provider='voyage', model='voyage-3-lite')
2. Fix the upstream config so empty strings don't reach this API
3. Use the same provider/model recorded in the embeddings table (SELECT DISTINCT provider FROM embeddings)

Example fix

# before
refresh_embeddings(gs, provider=os.environ.get('P') or '', model=os.environ.get('M') or '')
# after
refresh_embeddings(gs, provider='voyage', model='voyage-3-lite')

provider = (provider or "").strip()
model = (model or "").strip()
if not provider or not model:
    raise ValueError("refresh requires explicit provider and model")

try:
    refresh_embeddings(gs, provider=p, model=m)
except ValueError as e:
    if "requires an explicit provider and model" in str(e):
        p, m = "local", "minilm-l12-v2"  # sane defaults
        refresh_embeddings(gs, provider=p, model=m)
    else:
        raise

When it happens

Trigger: Calling refresh_embeddings(graph_store, provider='', model='') or omitting/blanking either argument (values are stripped before the check, so whitespace-only also fails).

Common situations: Calling refresh from a wrapper that reads unset env vars and passes empty strings, or assuming refresh inherits the provider from the stored index.

Understand the failure class

Background: "Invalid configuration value" and "Unsupported/Unknown setting value" errors: why libraries reject your config strings, numbers, and types — this error's family across 30 libraries.

Related errors

• OpenAI API returned empty data
• Unknown embedding provider '{name}'. Valid: local, openai, g
• Missing required environment variable(s) for the OpenAI embe
• MINIMAX_API_KEY environment variable is required for the Min
• VOYAGE_API_KEY environment variable is required for the Voya

AI-assisted analysis of tirth8205/code-review-graph@b58668751a (2026-08-28). Data as JSON: /api/errors/d1783c2fa111c267. Report an issue: GitHub.