v0.20.0

0.20.0 (2026-10-20)

Features

• hub: decision_config /2 for a tuned model's repository (tuned and defaults sections, tuned-config and check commands) (0f0c092)
• hub: hub_fp8 merge - a LoRA adapter merged into the base in fp32 and quantized once to FP8 (3ca0e5e)
• serve: --over-limit truncate - answer a state over the prompt ceiling from its start (never certified, reported in the response) (ad16bbd)
• serve: a tuned repository's defaults are served - temperatures, choice bands and the commit setting, with the flags above them (62b3d13)
• serve: a tuned repository's defaults are served (#244) (62b3d13)

Upgrading to 0.20.0. Nothing changes for the default base or a decisio.decision_config/1 repository. A tuned model can now ship as a decisio.decision_config/2 repository (the merged weights and the temperatures, choice bands and commit setting it is served at, no flag needed); an older decisio refuses it by its schema. --over-limit truncate (also a request field and a repository default) answers a state over the prompt ceiling from its start; the response says so and no answer of a truncated state is certified. The default stays refuse. docs/upgrading.md has the table of who is affected. Certify certified tasks again after any upgrade.

Container image

ghcr.io/aminry/decisio:0.20.0

ghcr.io/aminry/decisio@sha256:df3cb3165d963593075e765fad570477c9caf2f2bf913365777756983972abd2

Verify: gh attestation verify oci://ghcr.io/aminry/decisio@sha256:df3cb3165d963593075e765fad570477c9caf2f2bf913365777756983972abd2 --repo aminry/decisio