ScrapeGraphAI/Scrapegraph-ai · error · ValueError

input_tokens is required for completion costs with tiered pr

Error message

input_tokens is required for completion costs with tiered pricing

What it means

ValueError from get_token_cost_for_model: for models with tiered (token-volume-dependent) pricing, the completion/output rate depends on the input token count, so input_tokens must be supplied to compute completion costs. Without it the correct tier cannot be selected.

Source

Thrown at scrapegraphai/utils/custom_callback.py:53

    Args:
        model_name: Name of the model
        num_tokens: Number of tokens.
        is_completion: Whether the model is used for completion or not.
            Defaults to False.
        input_tokens: Number of input tokens used to select a pricing tier.
        service_tier: Provider service tier. Defaults to standard.

    Returns:
        Cost in USD.
    """
    if (
        model_name not in MODEL_COST_PER_1K_TOKENS_INPUT
        and model_name not in MODEL_COST_TIERS_PER_1K_TOKENS
    ):
        return 0.0
    if input_tokens is None:
        if is_completion and model_name in MODEL_COST_TIERS_PER_1K_TOKENS:
            raise ValueError(
                "input_tokens is required for completion costs with tiered pricing"
            )
        input_tokens = num_tokens
    rate = get_model_cost_per_1k_tokens(
        model_name,
        input_tokens,
        is_completion=is_completion,
        service_tier=service_tier,
    )
    return rate * (num_tokens / 1000)


class CustomCallbackHandler(BaseCallbackHandler):
    """Callback Handler that tracks LLMs info."""

    total_tokens: int = 0
    prompt_tokens: int = 0
    completion_tokens: int = 0

View on GitHub (pinned to 532dfffbf6)

Solutions

Pass input_tokens (the prompt token usage from response.llm_output) alongside num_tokens when computing completion cost
Upgrade/patch the callback so on_llm_end extracts token_usage.prompt_tokens and forwards it
If tiering is irrelevant, use a model without tiered pricing

Example fix

# before
cost = get_token_cost_for_model(model, num_tokens=output_tokens, is_completion=True)
# after
cost = get_token_cost_for_model(model, num_tokens=output_tokens, input_tokens=prompt_tokens, is_completion=True)

Defensive patterns

Strategy: validation

Validate before calling

usage = response.llm_output.get("token_usage", {}) if response.llm_output else {}
input_tokens = usage.get("prompt_tokens")
if model in MODEL_COST_TIERS_PER_1K_TOKENS and is_completion and input_tokens is None:
    input_tokens = usage.get("total_tokens", 0)

Type guard

def has_input_tokens(response) -> bool:
    llm_out = getattr(response, "llm_output", None) or {}
    return bool((llm_out.get("token_usage") or {}).get("prompt_tokens") is not None)

Try / catch

try:
    cost = get_token_cost_for_model(model, n, input_tokens=inp, is_completion=True)
except ValueError:
    cost = 0.0  # skip untracked completion

Prevention

Always extract prompt_tokens from llm_output in on_llm_end
Default missing prompt token counts to 0, never omit
Unit-test cost callbacks against tiered-pricing models

When it happens

Trigger: Calling on_llm_end / get_token_cost_for_model with is_completion=True for a model listed in MODEL_COST_TIERS_PER_1K_TOKENS (e.g. MiniMax M3) without passing input_tokens.

Common situations: Custom LangChain callbacks tracking costs for tiered-pricing models where only the response (num_tokens) is available from the LMLOutput.

Understand the failure class

Background: Missing required parameter errors: what 'X is required' and 'the required X param is missing' mean, and how to fix them — this error's family across 27 libraries.

Related errors

AI-assisted analysis of ScrapeGraphAI/Scrapegraph-ai@532dfffbf6 (2026-08-28). Data as JSON: /api/errors/697be0a8eae0e467. Report an issue: GitHub.