llms.txt
# milliseconds.ai
> decision-machine-1 is a decisions API: send text and a question, get a typed decision back with probabilities. No chat, no generated prose. Most calls finish in under a second. Built by CloudRaker in Montréal.
Base URL: https://api.milliseconds.ai/v1/decision-machine-1. Every request needs an API key from https://console.milliseconds.ai/ sent as `Authorization: Bearer test_sk-...`.
## Capabilities
- POST /yes-no: Flag messages that need a faster response.
- POST /classify: Choose a destination from your support queues.
- POST /rate: Turn customer frustration into a sortable score.
- POST /answer: Locate the shipment destination in a status update.
- POST /extract: Map invoice text to the fields in your records.
- POST /entities: Identify people, organizations, and references in a claim note.
- POST /verify: Check a proposed deductible against the policy text.
- POST /classify-tree: walk a nested label tree in one request.
- POST /v1/chat/completions: OpenAI-compatible surface for json_schema extraction and tool calls with model decision-machine-1.
## Pricing
- Test keys (test_sk): 125 million free input tokens per month, shared by the organization. 30 inference requests/minute and 500,000 input TPM. Purchased credits do not refill this allowance.
- Production keys (prod_sk): purchased credits only at $0.04 per million input tokens; output tokens cost $0. No free allowance is deducted.
- Production access requires a paid credit bundle of at least $10. Auto top-up is optional; a saved credit card is not an ongoing access requirement.
- Tier 1: $10 lifetime paid, immediate, 1,000 inference requests/minute / 1M TPM. Tier 2: $50 and 14 days since first payment, 5,000 inference requests/minute / 1M TPM. Tier 3: $250 and 21 days, 5,000 inference requests/minute / unlimited TPM. Tier 4: $500 and 30 days, 10,000 inference requests/minute / unlimited TPM.
- Images: an image bills as a fixed number of input tokens set by `detail` — low 1,000 (512 px), medium 2,000 (768 px, default), high 4,000 (1024 px). It bills once per request; text in the same request bills as usual. Decisions (yes-no, classify, classify-tree, rate) use those counts: $0.04 / $0.08 / $0.16 per 1,000 images. answer multiplies them by 1.5: $0.06 / $0.12 / $0.24. extract, entities and verify multiply them by 2: $0.08 / $0.16 / $0.32.
- Images are JPEG, PNG or WebP up to 5 MB, sent as bytes (data URL or base64). No image URLs, nothing stored.
- Image extract, entities and verify are generative passes for a handful of fields off one document image, such as a receipt total or an ID number. They are not for long-form or full-page transcription; image extract accepts at most 5 fields per call. For whole pages and long documents, use the CloudRaker paperwork API at https://cloudraker.com, which parses and extracts full documents with no field cap.
- Each inference counts toward the rate limit. A batch of 24 texts with 3 statements counts as 72 inference requests.
- Need immediate assistance with a higher tier? Reach out to our support team: support@milliseconds.ai.
- Details: https://milliseconds.ai/pricing.md
## Links
- [Blog](https://milliseconds.ai/blog/): business and engineering articles. RSS at https://milliseconds.ai/blog/rss.xml.
- [Documentation](https://docs.milliseconds.ai/): guides, API reference, changelog. Machine-readable index at https://docs.milliseconds.ai/llms.txt; append .md to any docs page for Markdown.
- [Quickstart](https://docs.milliseconds.ai/quickstart): the first call in curl, Python and TypeScript.
- [OpenAPI spec](https://api.milliseconds.ai/openapi.json)
- [Developer console](https://console.milliseconds.ai/): sign up, keys, usage and billing.
- [Homepage](https://milliseconds.ai/): product overview and featured workflows.
- [Product](https://milliseconds.ai/product/): capabilities and recorded API examples.
- [What it does](https://milliseconds.ai/use-cases/): working business workflows.
- [Developers](https://milliseconds.ai/developers/): TypeScript, Python, and dm1 CLI.
- [Agent skills](https://milliseconds.ai/agents/): installation and workflow patterns.
- [Pricing](https://milliseconds.ai/pricing/): token pricing and usage estimator.
- [Demos](https://demo.milliseconds.ai/): stock examples are free to try; custom inputs require your key.
