PROMINENT AI - ILLUSTRATIVE RAG COST CALCULATOR

Requires Python 3; uses only its standard library. No credentials, API calls,
network access or third-party packages are used.

Run:
python rag_cost_calculator.py rag-cost-assumptions.json

All example rates are fictional currency units, not vendor quotes.
Replace rates with current, applicable prices and volumes with measured data.
Keep all rates in one currency. Per-million prices apply to token counts.
The included fixture calculates 729.24 currency units per month.

Model boundaries:
- Cache hits skip query embedding, retrieval, reranking and generation.
- Document embedding, fixed platform, cache and operations costs remain.
- Input tokens include all prompt/history/context tokens on an uncached request.
- Do not also add per-request retrieval costs already covered by fixed capacity.
- Add omitted OCR, enrichment, network, taxes, support, backup, retry and
  implementation costs where applicable. This is a planning aid, not a quote.
- Prompt caching and answer caching are different. The example is answer caching.
- Valid permissions and source freshness are prerequisites for cache reuse.

Source article: https://prominent.tech/insights/blog/
