Upgrade to EdgeQuake v0.31.0

From: v0.30.0 · To: v0.31.0 · CD: GHCR (edgequake, edgequake-frontend, edgequake-postgres, edgequake-keycloak)

This minor release adds decision extraction (SPEC-160) as a preview. The schema moves from 165 to 166. Run migrate before /ready returns 200.

From 0.29.0 or older, migrate to 165 first (upgrade-to-0.30.0.md), then apply this release.

The default extractor is still llm. Decision mode is chosen per upload or as a workspace default. The two new tables are additive, and existing decision graph rows stay in place.

demo.edgequake.com: this cut installs the 0.31.0 API and WebUI. Decision mode is a preview with uncalibrated gates. Password auth remains the demo default. SSO stays opt-in from 0.30.0.

Highlights

Area	What changed
Schema	166: `decision_cache` and `decision_review` tables
Extraction	`decision` extraction mode on upload and per workspace; the chat LLM is unchanged
Status	`GET /api/v1/decision/status` and `GET /api/v1/decision/models` (3 s probe)
Benchmark	Same 2026-08-15 medical-mid attestation as 0.30.0; the default `llm` path was not re-scored and decision mode is unscored

How the decision mode is chosen

%%{init: {"theme":"base","themeVariables":{"primaryColor":"#E0E7FF","primaryBorderColor":"#6366F1","primaryTextColor":"#1E1B4B","secondaryColor":"#D1FAE5","secondaryBorderColor":"#10B981","secondaryTextColor":"#064E3B","tertiaryColor":"#FEF3C7","tertiaryBorderColor":"#F59E0B","tertiaryTextColor":"#6B7A90","lineColor":"#7A889C","clusterBkg":"rgba(99,102,241,0.07)","clusterBorder":"#A5B4FC","noteBkgColor":"#FEF9C3","noteTextColor":"#422006","textColor":"#6B7A90","titleColor":"#6B7A90","signalColor":"#7A889C","signalTextColor":"#6B7A90","loopTextColor":"#6B7A90","edgeLabelBackground":"#F1F5F9","actorLineColor":"#94A3B8"}}}%%
%% eq-theme:v1
flowchart TB
    A["Upload or workspace sets extraction mode"] --> B{"Mode is decision?"}
    B -->|"no (default llm)"| C["Standard LLM extraction, unchanged"]
    B -->|yes| D{"EDGEQUAKE_DECISION_ENABLED"}
    D -->|"0: locked off"| E["New decision uploads are refused"]
    D -->|"workspace: opt-in"| F["Only workspaces that opted in can use decision mode"]
    D -->|"1: enabled"| G["Decision backend runs (default ollama_system_one)"]
%% eq-classes
classDef eqLlm fill:#FEF3C7,stroke:#F59E0B,color:#451A03
classDef eqBad fill:#FEE2E2,stroke:#EF4444,color:#7F1D1D
class C,G eqLlm
class E eqBad

Notice that the environment setting gates every workspace opt-in.

Upgrade sequence

# Already on 0.30.0 (schema 165)
EDGEQUAKE_VERSION=0.31.0 docker compose pull
# migrate job or `edgequake migrate` (applies 166), then the API
EDGEQUAKE_VERSION=0.31.0 docker compose up -d

Verify

curl -sf localhost:8080/health | jq '{version, schema}'
# expect version "0.31.0", schema.latest_version 166 and pending_count 0
curl -sf localhost:8080/api/v1/decision/status
# shows the host and model fields; the probe never waits past 3 s

Residuals (not release blockers)

The openai_logprobs backend is not available in this release and is refused by settings validation.
There is no review screen. Review rows are stored and counted.
Gate presets are uncalibrated. The two-document probe is in w8-report.md. Treat tev1:0.8b at balanced as a smoke-test option only.
SPEC-001 Acc was not re-run for this cut.

Guide: Decision extraction.