mirror of
https://github.com/samvallad33/vestige.git
synced 2026-06-14 20:55:14 +02:00
Release v2.1.23 Receipt Lock hardening
Hardens Sanhedrin Receipt Lock for model-agnostic use, adds fail-open telemetry and receipt docs, fixes smart_ingest batch safety, wires opt-in CUDA Qwen3 device selection, and refreshes dashboard/release assets. Fixes #58 Fixes #60
This commit is contained in:
Sam Valladares
parent
a8550410b0
commit
5edb163157
161 changed files with 1775 additions and 262 deletions
|
|
@ -12,7 +12,7 @@ load_vestige_sanhedrin_env() {
|
|||
command -v python3 >/dev/null 2>&1 || return 0
|
||||
while IFS="$(printf '\t')" read -r key value; do
|
||||
case "$key" in
|
||||
VESTIGE_SANHEDRIN_ENABLED|VESTIGE_SANHEDRIN_MODEL|VESTIGE_SANHEDRIN_ENDPOINT|VESTIGE_SANHEDRIN_CLAIM_MODE|VESTIGE_SANHEDRIN_OUTPUT|VESTIGE_SANHEDRIN_PYTHON|VESTIGE_DASHBOARD_PORT)
|
||||
VESTIGE_SANHEDRIN_ENABLED|VESTIGE_SANHEDRIN_MODEL|VESTIGE_SANHEDRIN_ENDPOINT|VESTIGE_SANHEDRIN_BACKEND|VESTIGE_SANHEDRIN_CLAIM_MODE|VESTIGE_SANHEDRIN_OUTPUT|VESTIGE_SANHEDRIN_PYTHON|VESTIGE_DASHBOARD_PORT)
|
||||
export "$key=$value"
|
||||
;;
|
||||
esac
|
||||
|
|
@ -24,6 +24,7 @@ allowed = {
|
|||
"VESTIGE_SANHEDRIN_ENABLED",
|
||||
"VESTIGE_SANHEDRIN_MODEL",
|
||||
"VESTIGE_SANHEDRIN_ENDPOINT",
|
||||
"VESTIGE_SANHEDRIN_BACKEND",
|
||||
"VESTIGE_SANHEDRIN_CLAIM_MODE",
|
||||
"VESTIGE_SANHEDRIN_OUTPUT",
|
||||
"VESTIGE_SANHEDRIN_PYTHON",
|
||||
|
|
@ -79,9 +80,10 @@ if [ -f "$SANHEDRIN_ENV" ]; then
|
|||
load_vestige_sanhedrin_env "$SANHEDRIN_ENV" || true
|
||||
fi
|
||||
|
||||
SANHEDRIN_ENDPOINT="${VESTIGE_SANHEDRIN_ENDPOINT:-${MLX_ENDPOINT:-http://127.0.0.1:8080/v1/chat/completions}}"
|
||||
SANHEDRIN_ENDPOINT="${VESTIGE_SANHEDRIN_ENDPOINT:-${MLX_ENDPOINT:-}}"
|
||||
SANHEDRIN_ENDPOINT="${SANHEDRIN_ENDPOINT%/}"
|
||||
SANHEDRIN_MODELS_URL="${SANHEDRIN_ENDPOINT%/chat/completions}/models"
|
||||
SANHEDRIN_MODELS_URL=""
|
||||
[ -n "$SANHEDRIN_ENDPOINT" ] && SANHEDRIN_MODELS_URL="${SANHEDRIN_ENDPOINT%/chat/completions}/models"
|
||||
SANHEDRIN_CLAIM_MODE="${VESTIGE_SANHEDRIN_CLAIM_MODE:-0}"
|
||||
SANHEDRIN_OUTPUT="${VESTIGE_SANHEDRIN_OUTPUT:-text}"
|
||||
|
||||
|
|
@ -161,10 +163,12 @@ if [ "$CHECK_SANHEDRIN" -eq 1 ]; then
|
|||
command -v mlx_lm.server >/dev/null && ok "mlx-lm" || warn "mlx-lm — uv tool install mlx-lm"
|
||||
command -v hf >/dev/null && ok "huggingface_hub CLI" || warn "hf — uv tool install 'huggingface_hub[cli]'"
|
||||
|
||||
MODEL="${VESTIGE_SANHEDRIN_MODEL:-${VESTIGE_SANDWICH_MODEL:-mlx-community/Qwen3.6-35B-A3B-4bit}}"
|
||||
MODEL="${VESTIGE_SANHEDRIN_MODEL:-${VESTIGE_SANDWICH_MODEL:-}}"
|
||||
HF_HOME_DEFAULT="${HF_HOME:-$HOME/.cache/huggingface}"
|
||||
ENC_MODEL="models--$(printf '%s' "$MODEL" | sed 's|/|--|g')"
|
||||
if [ -d "$HF_HOME_DEFAULT/hub/$ENC_MODEL" ]; then
|
||||
if [ -z "$MODEL" ]; then
|
||||
info "No local MLX model configured; choose any OpenAI-compatible Sanhedrin model or preset."
|
||||
elif [ -d "$HF_HOME_DEFAULT/hub/$ENC_MODEL" ]; then
|
||||
ok "Model cached: $MODEL"
|
||||
else
|
||||
info "Model not cached: $MODEL (local MLX path downloads ~19GB)"
|
||||
|
|
@ -179,7 +183,9 @@ if [ "$CHECK_SANHEDRIN" -eq 1 ]; then
|
|||
info "Skipping MLX/launchd checks on $OS_NAME $ARCH_NAME"
|
||||
fi
|
||||
|
||||
if curl -fsS -m 2 "$SANHEDRIN_MODELS_URL" >/dev/null 2>&1; then
|
||||
if [ -z "$SANHEDRIN_MODELS_URL" ]; then
|
||||
warn "Sanhedrin endpoint/model not configured yet; hook will fail open until configured"
|
||||
elif curl -fsS -m 2 "$SANHEDRIN_MODELS_URL" >/dev/null 2>&1; then
|
||||
ok "Sanhedrin model endpoint responding at $SANHEDRIN_MODELS_URL"
|
||||
else
|
||||
warn "Sanhedrin endpoint not responding at $SANHEDRIN_MODELS_URL"
|
||||
|
|
|
|||
|
|
@ -19,11 +19,12 @@ set -euo pipefail
|
|||
|
||||
VERSION="${VESTIGE_SANDWICH_VERSION:-v2.1.1}"
|
||||
REPO="samvallad33/vestige"
|
||||
MODEL_ID="${VESTIGE_SANHEDRIN_MODEL:-${VESTIGE_SANDWICH_MODEL:-mlx-community/Qwen3.6-35B-A3B-4bit}}"
|
||||
MODEL_ID="${VESTIGE_SANHEDRIN_MODEL:-${VESTIGE_SANDWICH_MODEL:-}}"
|
||||
DASHBOARD_PORT="${VESTIGE_DASHBOARD_PORT:-3927}"
|
||||
SANHEDRIN_ENDPOINT="${VESTIGE_SANHEDRIN_ENDPOINT:-${MLX_ENDPOINT:-http://127.0.0.1:8080/v1/chat/completions}}"
|
||||
SANHEDRIN_ENDPOINT="${VESTIGE_SANHEDRIN_ENDPOINT:-${MLX_ENDPOINT:-}}"
|
||||
SANHEDRIN_ENDPOINT="${SANHEDRIN_ENDPOINT%/}"
|
||||
SANHEDRIN_MODELS_URL="${SANHEDRIN_ENDPOINT%/chat/completions}/models"
|
||||
SANHEDRIN_MODELS_URL=""
|
||||
[ -n "$SANHEDRIN_ENDPOINT" ] && SANHEDRIN_MODELS_URL="${SANHEDRIN_ENDPOINT%/chat/completions}/models"
|
||||
SANHEDRIN_CLAIM_MODE="${VESTIGE_SANHEDRIN_CLAIM_MODE:-1}"
|
||||
SANHEDRIN_OUTPUT="${VESTIGE_SANHEDRIN_OUTPUT:-json}"
|
||||
MODEL_ID_FROM_INSTALLER=0
|
||||
|
|
@ -88,14 +89,28 @@ die() { printf '\033[1;31m[sandwich]\033[0m %s\n' "$*" >&2; exit 1; }
|
|||
OS_NAME="$(uname -s)"
|
||||
ARCH_NAME="$(uname -m)"
|
||||
say "platform: $OS_NAME $ARCH_NAME"
|
||||
if [ "$ENABLE_SANHEDRIN" -eq 1 ] && [ "$WITH_LAUNCHD" -eq 0 ]; then
|
||||
say "Sanhedrin enabled without launchd; using OpenAI-compatible endpoint: $SANHEDRIN_ENDPOINT"
|
||||
fi
|
||||
if [ "$WITH_LAUNCHD" -eq 1 ] && { [ "$OS_NAME" != "Darwin" ] || [ "$ARCH_NAME" != "arm64" ]; }; then
|
||||
warn "--with-launchd is Apple Silicon only; skipping local MLX autostart on $OS_NAME $ARCH_NAME"
|
||||
warn "Sanhedrin can still run on x86 via --sanhedrin-endpoint or VESTIGE_SANHEDRIN_ENDPOINT."
|
||||
WITH_LAUNCHD=0
|
||||
fi
|
||||
if [ "$WITH_LAUNCHD" -eq 1 ]; then
|
||||
if [ "$SANHEDRIN_ENDPOINT_FROM_INSTALLER" -eq 0 ]; then
|
||||
SANHEDRIN_ENDPOINT="${SANHEDRIN_ENDPOINT:-http://127.0.0.1:8080/v1/chat/completions}"
|
||||
SANHEDRIN_MODELS_URL=""
|
||||
[ -n "$SANHEDRIN_ENDPOINT" ] && SANHEDRIN_MODELS_URL="${SANHEDRIN_ENDPOINT%/chat/completions}/models"
|
||||
fi
|
||||
if [ "$MODEL_ID_FROM_INSTALLER" -eq 0 ]; then
|
||||
MODEL_ID="${MODEL_ID:-mlx-community/Qwen3.6-35B-A3B-4bit}"
|
||||
fi
|
||||
fi
|
||||
if [ "$ENABLE_SANHEDRIN" -eq 1 ] && [ "$WITH_LAUNCHD" -eq 0 ]; then
|
||||
if [ -n "$SANHEDRIN_ENDPOINT" ] && [ -n "$MODEL_ID" ]; then
|
||||
say "Sanhedrin enabled with custom OpenAI-compatible model: $MODEL_ID"
|
||||
else
|
||||
warn "Sanhedrin enabled with no verifier model configured yet; it will fail open until VESTIGE_SANHEDRIN_ENDPOINT and VESTIGE_SANHEDRIN_MODEL are set."
|
||||
fi
|
||||
fi
|
||||
|
||||
# --- Prereqs (warnings only, install proceeds) ---
|
||||
command -v jq >/dev/null || die "jq required: brew install jq"
|
||||
|
|
@ -146,7 +161,7 @@ fi
|
|||
|
||||
# --- Copy hooks ---
|
||||
copied=0; skipped=0
|
||||
for f in "$SCRIPT_DIR/hooks"/*.sh "$SCRIPT_DIR/hooks"/*.py; do
|
||||
for f in "$SCRIPT_DIR/hooks"/*.sh "$SCRIPT_DIR/hooks"/*.py "$SCRIPT_DIR/hooks"/sanhedrin-presets.json; do
|
||||
[ -f "$f" ] || continue
|
||||
base="$(basename "$f")"
|
||||
# load-all-memory.sh dumps every memory MD — opt-in only
|
||||
|
|
@ -158,7 +173,10 @@ for f in "$SCRIPT_DIR/hooks"/*.sh "$SCRIPT_DIR/hooks"/*.py; do
|
|||
skipped=$((skipped + 1))
|
||||
continue
|
||||
fi
|
||||
install -m 0755 "$f" "$HOOKS_DIR/$base"
|
||||
case "$base" in
|
||||
*.json) install -m 0644 "$f" "$HOOKS_DIR/$base" ;;
|
||||
*) install -m 0755 "$f" "$HOOKS_DIR/$base" ;;
|
||||
esac
|
||||
copied=$((copied + 1))
|
||||
done
|
||||
say "hooks: $copied installed, $skipped skipped (use --force to overwrite)"
|
||||
|
|
@ -183,7 +201,7 @@ load_vestige_sanhedrin_env() {
|
|||
[ -f "$1" ] || return 0
|
||||
while IFS="$(printf '\t')" read -r key value; do
|
||||
case "$key" in
|
||||
VESTIGE_SANHEDRIN_ENABLED|VESTIGE_SANHEDRIN_MODEL|VESTIGE_SANHEDRIN_ENDPOINT|VESTIGE_SANHEDRIN_CLAIM_MODE|VESTIGE_SANHEDRIN_OUTPUT|VESTIGE_SANHEDRIN_PYTHON|VESTIGE_DASHBOARD_PORT)
|
||||
VESTIGE_SANHEDRIN_ENABLED|VESTIGE_SANHEDRIN_MODEL|VESTIGE_SANHEDRIN_ENDPOINT|VESTIGE_SANHEDRIN_API_KEY|VESTIGE_SANHEDRIN_BACKEND|VESTIGE_SANHEDRIN_CLAIM_MODE|VESTIGE_SANHEDRIN_OUTPUT|VESTIGE_SANHEDRIN_PYTHON|VESTIGE_SANHEDRIN_ALLOW_LOOSE_LEDGER|VESTIGE_DASHBOARD_PORT)
|
||||
export "$key=$value"
|
||||
;;
|
||||
esac
|
||||
|
|
@ -195,9 +213,12 @@ allowed = {
|
|||
"VESTIGE_SANHEDRIN_ENABLED",
|
||||
"VESTIGE_SANHEDRIN_MODEL",
|
||||
"VESTIGE_SANHEDRIN_ENDPOINT",
|
||||
"VESTIGE_SANHEDRIN_API_KEY",
|
||||
"VESTIGE_SANHEDRIN_BACKEND",
|
||||
"VESTIGE_SANHEDRIN_CLAIM_MODE",
|
||||
"VESTIGE_SANHEDRIN_OUTPUT",
|
||||
"VESTIGE_SANHEDRIN_PYTHON",
|
||||
"VESTIGE_SANHEDRIN_ALLOW_LOOSE_LEDGER",
|
||||
"VESTIGE_DASHBOARD_PORT",
|
||||
}
|
||||
|
||||
|
|
@ -248,7 +269,8 @@ if [ "$ENABLE_SANHEDRIN" -eq 1 ]; then
|
|||
SANHEDRIN_ENDPOINT="${VESTIGE_SANHEDRIN_ENDPOINT:-$SANHEDRIN_ENDPOINT}"
|
||||
SANHEDRIN_ENDPOINT="${SANHEDRIN_ENDPOINT%/}"
|
||||
fi
|
||||
SANHEDRIN_MODELS_URL="${SANHEDRIN_ENDPOINT%/chat/completions}/models"
|
||||
SANHEDRIN_MODELS_URL=""
|
||||
[ -n "$SANHEDRIN_ENDPOINT" ] && SANHEDRIN_MODELS_URL="${SANHEDRIN_ENDPOINT%/chat/completions}/models"
|
||||
if [ "$SANHEDRIN_CLAIM_MODE_FROM_INSTALLER" -eq 1 ]; then
|
||||
SANHEDRIN_CLAIM_MODE="$INSTALLER_SANHEDRIN_CLAIM_MODE"
|
||||
else
|
||||
|
|
@ -260,6 +282,16 @@ if [ "$ENABLE_SANHEDRIN" -eq 1 ]; then
|
|||
SANHEDRIN_OUTPUT="${VESTIGE_SANHEDRIN_OUTPUT:-$SANHEDRIN_OUTPUT}"
|
||||
fi
|
||||
fi
|
||||
if [ "$WITH_LAUNCHD" -eq 0 ] \
|
||||
&& [ "$MODEL_ID_FROM_INSTALLER" -eq 0 ] \
|
||||
&& [ "$SANHEDRIN_ENDPOINT_FROM_INSTALLER" -eq 0 ] \
|
||||
&& [ "$MODEL_ID" = "mlx-community/Qwen3.6-35B-A3B-4bit" ] \
|
||||
&& [ "$SANHEDRIN_ENDPOINT" = "http://127.0.0.1:8080/v1/chat/completions" ]; then
|
||||
MODEL_ID=""
|
||||
SANHEDRIN_ENDPOINT=""
|
||||
SANHEDRIN_MODELS_URL=""
|
||||
warn "Cleared legacy implicit MLX/Qwen Sanhedrin default. Choose a preset or set VESTIGE_SANHEDRIN_ENDPOINT and VESTIGE_SANHEDRIN_MODEL."
|
||||
fi
|
||||
TMP_ENV="$(mktemp)"
|
||||
if [ -f "$SANHEDRIN_ENV" ]; then
|
||||
awk -F= '
|
||||
|
|
@ -365,7 +397,8 @@ cat <<EOF
|
|||
scripts/check-sandwich-prereqs.sh # from a checkout
|
||||
3. Optional hook layers:
|
||||
./scripts/install-sandwich.sh --enable-preflight
|
||||
./scripts/install-sandwich.sh --enable-sanhedrin --sanhedrin-endpoint=$SANHEDRIN_ENDPOINT --sanhedrin-model=$MODEL_ID
|
||||
./scripts/install-sandwich.sh --enable-sanhedrin --sanhedrin-endpoint=<url> --sanhedrin-model=<model>
|
||||
./scripts/install-sandwich.sh --enable-sanhedrin --with-launchd # explicit MLX/Qwen path
|
||||
On Apple Silicon with >20 GB free RAM, add --with-launchd to auto-start
|
||||
the local MLX Qwen server. On x86, point --sanhedrin-endpoint at vLLM,
|
||||
Ollama, llama.cpp, or another OpenAI-compatible /v1/chat/completions URL.
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue