langchain-ai/deepagents · error · ValueError

RewarmEstimate costs must be finite, got cold={self.cold_cos

Error message

RewarmEstimate costs must be finite, got cold={self.cold_cost_usd!r}, incremental={self.incremental_cost_usd!r}

What it means

RewarmEstimate is a frozen dataclass validated in __post_init__ to ensure its two cost fields are finite numbers. This error means cold_cost_usd or incremental_cost_usd was NaN or +/-inf, which would corrupt any downstream cost comparison or budget math. The library fails fast at construction rather than propagating non-finite costs.

Source

Thrown at libs/code/deepagents_code/cold_cache.py:206

        """Enforce the documented finiteness, ordering, and sign invariants.

        Raises:
            ValueError: When either figure is non-finite or negative, or the
                delta exceeds the total it is a part of.
        """
        # Checked first, and separately: `NaN` satisfies neither comparison
        # below (`nan < 0` and `nan > nan` are both `False`), so it would slide
        # past both guards and reach `format_cost_estimate`, where the
        # magnitude arithmetic raises far from the value's real origin.
        if not math.isfinite(self.cold_cost_usd) or not math.isfinite(
            self.incremental_cost_usd
        ):
            msg = (
                f"RewarmEstimate costs must be finite, got "
                f"cold={self.cold_cost_usd!r}, "
                f"incremental={self.incremental_cost_usd!r}"
            )
            raise ValueError(msg)
        if self.cold_cost_usd < 0 or self.incremental_cost_usd < 0:
            msg = (
                f"RewarmEstimate costs must be non-negative, got "
                f"cold={self.cold_cost_usd!r}, "
                f"incremental={self.incremental_cost_usd!r}"
            )
            raise ValueError(msg)
        if self.incremental_cost_usd > self.cold_cost_usd:
            msg = (
                f"RewarmEstimate incremental cost {self.incremental_cost_usd!r} "
                f"cannot exceed the cold cost {self.cold_cost_usd!r}"
            )
            raise ValueError(msg)


@dataclass(frozen=True, slots=True)
class ColdCacheWarning:
    """Validated data needed to render one advisory warning.

View on GitHub (pinned to a1af029e6e)

Solutions

  1. Check both cost inputs with math.isfinite() before constructing RewarmEstimate and substitute a real estimate for the missing value.
  2. Fix the upstream pricing/telemetry data source so it no longer yields NaN or infinity.
  3. If 'unknown cost' must be represented, use None or a separate flag at the call site instead of inf, before the dataclass boundary.

Example fix

// before
est = RewarmEstimate(cold_cost_usd=total / count, incremental_cost_usd=0.02)  # count==0 -> inf
// after
import math
cold = total / count if count else 0.0
assert math.isfinite(cold), "cold cost estimate not computable"
est = RewarmEstimate(cold_cost_usd=cold, incremental_cost_usd=0.02)
Defensive patterns

Strategy: validation

Validate before calling

import math
if not (math.isfinite(cold) and math.isfinite(inc)):
    raise ValueError(f"cost estimates must be finite: cold={cold!r}, inc={inc!r}")
est = RewarmEstimate(cold_cost_usd=cold, incremental_cost_usd=inc)

Type guard

def is_finite_cost(v: object) -> bool:
    return isinstance(v, (int, float)) and math.isfinite(v)

Try / catch

try:
    est = RewarmEstimate(cold_cost_usd=cold, incremental_cost_usd=inc)
except ValueError as e:
    logger.warning("invalid rewarm estimate: %s", e)
    est = None

Prevention

When it happens

Trigger: Constructing RewarmEstimate(cold_cost_usd=..., incremental_cost_usd=...) where either value is math.nan, math.inf, or -math.inf (e.g. computed by dividing by zero or from a missing pricing entry).

Common situations: Estimating costs from a pricing table with a missing model entry (0/0 division), loading cached estimates from corrupted JSON, or passing float('inf') as a sentinel for 'unknown cost'.

Related errors

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/773f80e18c9552f4. Report an issue: GitHub.