Quickstart¶
from mimir import Mimir
model = Mimir.from_pretrained("Mythologic/MIMIR-1")
result = model.choose(
"My card was charged twice for the same order.",
"Which team should handle this ticket?",
options={"billing": "Billing: payments, refunds", "security": "Security: account access"},
)
The first call downloads the model from the Hugging Face Hub at the revision this package version pins, verifies the manifest's Sigstore signature, checks every file against the SHA-256 in the manifest, and loads it. Nothing is read until every check passes.
result.status # Status.DECIDED, Status.ABSTAINED or Status.DEFERRED
result.answer # "billing", or None when no option applies
result.probabilities # calibrated probability of each option id
result.relevant_context # the parts of the context the answer relied on, most relevant first
result.certificate # the certified threshold the decision was checked against
answer is what the model thinks. status is what you may do with it:
DECIDED— act onanswer; it is certified at the requested risk level.ABSTAINED— no listed option applies, and that conclusion is certified.DEFERRED— do not act; escalate.result.deferral.reasonsays why:below_threshold,out_of_distribution, orno_certified_threshold.
Next¶
- Decisions — yes/no questions, claim verification, rankings, ratings, and estimates, over passages, tables, and JSON.
- The certificate — what the risk level guarantees, and how to certify thresholds on your own data.
- Python — batching, async, offline use, and the HTTP client.