HeyPuter/puter · error · HttpError

unknown_error

unknown_error

Error message

Error calculating cost for Replicate model ${selectedModel.id}

What it means

Thrown by ReplicateImageGenerationProvider.generate() (ReplicateImageGenerationProvider.ts:148) when #estimateCost() returns a value <= 0. This is the outer guard after #estimateCost; note #estimateCost itself throws more specific errors for missing per-image cost (line 407) or missing output_mp cost (line 418), so reaching the <= 0 guard means costs were present but computed to zero (e.g. zero run cost, zero output_mp, zero input). HTTP 400 unknown_error — effectively a billing-config defect for the model.

Source

Thrown at src/backend/drivers/ai-image/providers/replicate/ReplicateImageGenerationProvider.ts:148

        const allInputUrls = singleImage ? [singleImage] : inputImages;
        const inputMp =
            allInputUrls.length > 0
                ? await this.#measureInputMegapixels(allInputUrls)
                : 0;

        const outputMp = this.#resolveOutputMegapixels(
            params.output_megapixels as string | undefined,
        );

        const totalCostMicroCents = this.#estimateCost(
            selectedModel,
            outputMp,
            goFast,
            inputMp,
            generationMode,
        );
        if (totalCostMicroCents <= 0) {
            throw new HttpError(
                400,
                `Error calculating cost for Replicate model ${selectedModel.id}`,
                { legacyCode: 'unknown_error' },
            );
        }
        const usageAllowed = await this.#meteringService.hasEnoughCredits(
            actor,
            totalCostMicroCents,
        );
        if (!usageAllowed) {
            throw new HttpError(
                402,
                'Insufficient credits for image generation',
                {
                    legacyCode: 'insufficient_funds',
                },
            );
        }

View on GitHub (pinned to 908ec23eda)

Solutions

1. Fix the model's cost configuration in src/backend/drivers/ai-image/providers/replicate/models.ts: ensure output_mp (or output for per-image) is a positive number.
2. Verify the selected costs variant (base vs costs_go_fast vs costs_by_generation_mode) has valid non-zero rates.
3. If you are a consumer, try without go_fast or with a different model to isolate which costs branch is zero.
4. Audit #estimateCost inputs (outputMp, inputMp, generationMode) to ensure they are positive.

Example fix

// models.ts — before
costs: { run: 0, output_mp: 0, input_mp: 0 }

// after
costs: { run: 0.1, output_mp: 0.5, input_mp: 0.2 }

function assertPositiveCost(model: ReplicateImageModel, costMicroCents: number): void {
  if (costMicroCents <= 0) {
    throw new Error(`Cost config invalid for Replicate model ${model.id}`);
  }
}
// or validate the model's cost map directly before generate():
const c = model.costs;
if ((c.output_mp ?? 0) <= 0 && (c.output ?? 0) <= 0) {
  throw new Error(`Model ${model.id} has no valid cost`);
}

When it happens

Trigger: A Replicate model whose resolved costs object (base, costs_go_fast, or costs_by_generation_mode) has output_mp = 0 and run = 0 and input_mp = 0; a per-image model with output = 0 (though that throws earlier at line 407); selecting go_fast or a generation_mode that resolves to a costs object with all-zero values; a model whose costs map is empty.

Common situations: Maintainer adds a model with placeholder zero costs; a generation_mode key typo that falls back to a sparse costs object; go_fast flag selecting a costs_go_fast map that lacks output_mp.

Related errors

• insufficient_funds
• insufficient_funds
• unknown_error
• insufficient_funds
• unknown_error

AI-assisted analysis of HeyPuter/puter@908ec23eda (2026-08-12). Data as JSON: /api/errors/a739eb885a0fa7f4. Report an issue: GitHub.