microsoft/graphrag · error · ValueError

At least one of requests_per_period or tokens_per_period mus

Error message

At least one of requests_per_period or tokens_per_period must be specified for Sliding Window rate limit.

What it means

Sliding Window rate limiting needs something to limit: RateLimitConfig validation demands at least one of requests_per_period or tokens_per_period. A config with neither makes the limiter a no-op, which is rejected.

Source

Thrown at packages/graphrag-llm/graphrag_llm/config/rate_limit_config.py:45

    requests_per_period: int | None = Field(
        default=None,
        description="The maximum number of requests allowed per period. (default: None, no limit).",
    )

    tokens_per_period: int | None = Field(
        default=None,
        description="The maximum number of tokens allowed per period. (default: None, no limit).",
    )

    def _validate_sliding_window_config(self) -> None:
        """Validate Sliding Window rate limit configuration."""
        if self.period_in_seconds is not None and self.period_in_seconds <= 0:
            msg = "period_in_seconds must be a positive integer for Sliding Window rate limit."
            raise ValueError(msg)

        if not self.requests_per_period and not self.tokens_per_period:
            msg = "At least one of requests_per_period or tokens_per_period must be specified for Sliding Window rate limit."
            raise ValueError(msg)

        if self.requests_per_period is not None and self.requests_per_period <= 0:
            msg = "requests_per_period must be a positive integer for Sliding Window rate limit."
            raise ValueError(msg)

        if self.tokens_per_period is not None and self.tokens_per_period <= 0:
            msg = "tokens_per_period must be a positive integer for Sliding Window rate limit."
            raise ValueError(msg)

    @model_validator(mode="after")
    def _validate_model(self):
        """Validate the rate limit configuration based on its type."""
        if self.type == RateLimitType.SlidingWindow:
            self._validate_sliding_window_config()
        return self

View on GitHub (pinned to f40e9a26ce)

Solutions

Add requests_per_period (e.g. 100) or tokens_per_period (e.g. 1000000) to the rate_limit block
Check YAML indentation so the keys land under rate_limit and are actually parsed

Example fix

# before
rate_limit:
  type: sliding_window
  period_in_seconds: 60
# after
rate_limit:
  type: sliding_window
  period_in_seconds: 60
  requests_per_period: 100

Defensive patterns

Strategy: validation

Validate before calling

rl = cfg.get("rate_limit") or {}
if rl.get("type") == "sliding_window" and not (rl.get("requests_per_period") or rl.get("tokens_per_period")):
    raise ValueError("set requests_per_period or tokens_per_period")

Prevention

Treat the rate_limit block as all-or-nothing in config templates
Lint YAML blocks for required sibling keys

When it happens

Trigger: RateLimitConfig(type=RateLimitType.SlidingWindow, period_in_seconds=60) with both requests_per_period and tokens_per_period omitted/None.

Common situations: Enabling rate_limit in settings.yaml but only setting the period; assuming a default limit exists; YAML indentation putting the limit keys under the wrong block.

Related errors

AI-assisted analysis of microsoft/graphrag@f40e9a26ce (2026-08-27). Data as JSON: /api/errors/f04c85a3bee0b5ae. Report an issue: GitHub.