Estimated monthly total
—For the workload belowINFRASTRUCTURE ECONOMICS / 01
What does a greenwashing review cost?
Shared monthly storage
—Bucket + vectors + passage metadataInitial ingestion
—Already incurred for the current corpusPlanning budget
—Includes a reserve, not a billed feeSet your workload
Inputs are local to your browser. They never modify Google Cloud or the shared editorial notes.
Legal library & storage
Advanced usage & allowances
Where the money goes
Calculated from the inputs and published list rates. No credits, discounts or tax applied.
Vertex AI is dynamic, token-based.
Input tokens include your prompt and retrieved law. Output tokens include reasoning. Assessment and verification are both counted. More content or repair calls mean more usage.
Show every formula
Same library, different workloads
Only page volume changes. All other current inputs are held constant.
03 / EVIDENCE & LIMITS
A planning model, with its assumptions visible.
Measured / configured
20 imported provisions · 3,072 dimensions · 1.897373 MB of local source files · 9,011 embedding characters. Sample: 9,328 input and 4,172 output + reasoning tokens across two calls.
Estimated / not yet verified
Metadata footprint, page complexity and query length are assumptions. Live bucket size and billable vector storage have not been refreshed. The sample is one synthetic paragraph, not a production page average.
Outside the core estimate
Crawling, app hosting, network transfer, logs and customer-result storage are excluded unless entered as other costs. No dedicated ANN serving capacity is included. KNN scanning gets more expensive as the corpus grows.
04 / RATE CARD
Published rates. Inspectable maths.
USD list-price snapshot, 14 September 2026. MB means 1,000,000 bytes; billed GiB means 1,073,741,824 bytes. Storage uses a normalized 730-hour month. Finland bucket rate is a planning assumption pending a billing SKU check; no automatic live price refresh. Embedding rate applies to gemini-embedding-001, not Embedding 2.