langchain-ai/deepagents · error · ValueError

RewarmEstimate incremental cost {self.incremental_cost_usd!r

Error message

RewarmEstimate incremental cost {self.incremental_cost_usd!r} cannot exceed the cold cost {self.cold_cost_usd!r}

What it means

RewarmEstimate requires incremental_cost_usd <= cold_cost_usd: re-warming a cache can never cost more than a full cold rebuild. If it would, the estimate is inconsistent and the library rejects construction.

Source

Thrown at libs/code/deepagents_code/cold_cache.py:219

            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.

    Constructed only after every gate has passed -- a policy resolved, the
    prefix cleared the provider's cache minimum, and the priced delta reached
    the configured threshold -- so the modal renders it without re-deciding
    anything.
    """

    policy: PromptCachePolicy
    estimate: RewarmEstimate
    context_tokens: int

    age_seconds: float | None
    """Idle time since the last successful turn.

View on GitHub (pinned to a1af029e6e)

Solutions

1. Recompute both estimates from the same token counts and pricing source so incremental <= cold holds.
2. Clamp the incremental estimate to the cold estimate (min()) if clamping is a defensible approximation for your use case.
3. Audit which pricing snapshot each number came from and refresh them together.

Example fix

// before
est = RewarmEstimate(cold_cost_usd=cold_old_pricing, incremental_cost_usd=inc_new_pricing)
// after
cold = estimate_cold(tokens, pricing_latest)
inc = min(estimate_incremental(tokens, pricing_latest), cold)
est = RewarmEstimate(cold_cost_usd=cold, incremental_cost_usd=inc)

if inc > cold:
    raise ValueError(f"incremental {inc!r} exceeds cold {cold!r}; recheck estimates")
est = RewarmEstimate(cold_cost_usd=cold, incremental_cost_usd=inc)

def is_valid_cost_pair(cold: float, inc: float) -> bool:
    return math.isfinite(cold) and math.isfinite(inc) and 0 <= inc <= cold

try:
    est = RewarmEstimate(cold_cost_usd=cold, incremental_cost_usd=inc)
except ValueError:
    est = RewarmEstimate(cold_cost_usd=cold, incremental_cost_usd=min(inc, cold))

When it happens

Trigger: Constructing RewarmEstimate(cold_cost_usd=1.0, incremental_cost_usd=1.5), typically because the two estimates came from different scopes, different models, or stale vs refreshed pricing.

Common situations: Cold cost computed for a smaller token count than the incremental path, mixing prices from two model versions, or double-counting fees into the incremental estimate.

Related errors

• RewarmEstimate costs must be finite, got cold={self.cold_cos
• RewarmEstimate costs must be non-negative, got cold={self.co
• modes can only be provided when agent is a factory
• models can only be provided when agent is a factory
• -32602

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/d26fcd7c757d7406. Report an issue: GitHub.