Skip to main content

InertialAI-0.1

InertialAI-0.1 is a chat model built for real-world, physical data: sensor streams, machine telemetry, operational metrics, market series, and health signals. Paste readings into the conversation, ask questions in plain language, and get answers grounded in the numbers — classification, anomaly triage, summarization, forecast narration, and general reasoning over your metrics.

It is served behind a fully OpenAI-compatible API, so anything that speaks the OpenAI Chat Completions protocol — the official SDKs, LangChain, LlamaIndex, your existing framework code — works by changing two lines: the base URL and the API key.

Base URL   https://inertialai.com/api/v1
Endpoint   POST /api/v1/chat/completions
Model id   inertialai-0.1
Auth       Authorization: Bearer iai_...

Why teams pick it​

  • Built for physical data. The model is trained on paired signal + text data, so it reads raw numbers the way a domain engineer does — trends, seasonality, spikes, drift — instead of treating them as opaque digits.
  • Token-level confidence you can gate on. Set logprobs: true and every answer token comes back with a probability you can act on. On live graded runs, right answers average 0.97 token confidence and wrong ones 0.71 — one threshold keeps 92% of the right answers and drops over half the wrong ones. Alert routing and human-in-the-loop escalation become one if statement. See Token-level confidence.
  • Drop-in. OpenAI-compatible request and response shapes, including token logprobs (streaming: coming soon). No new SDK to learn.
  • Simple pricing. Pay per token: $0.25 / 1M input, $1.25 / 1M output. No subscriptions or minimums. Details on the billing page.

A 60-second taste​

from openai import OpenAI

client = OpenAI(
    base_url="https://inertialai.com/api/v1",
    api_key="iai_...",  # or os.environ["INERTIALAI_API_KEY"]
)

resp = client.chat.completions.create(
    model="inertialai-0.1",
    messages=[
        {
            "role": "user",
            "content": (
                "Hourly pump vibration RMS (mm/s): "
                "2.1, 2.0, 2.2, 2.1, 4.8, 5.1, 5.0. "
                "Anything wrong? Answer with one word: normal or anomaly."
            ),
        }
    ],
    logprobs=True,
    top_logprobs=3,
)

print(resp.choices[0].message.content)          # "anomaly"
print(resp.choices[0].logprobs.content[0])      # token + its probability

Where to go next​

  • Quickstart — key, warmup ping, first request in under five minutes.
  • Chat Completions reference — every request and response field, streaming, and cold-start behavior.
  • Token-level confidence — turning logprobs into thresholds you can ship.
  • Need the model tuned to your fleet, your patients, your book? Chronicle finetunes on your data from $5, with an automatic before/after eval and a refund if it doesn't beat the base.