--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/mrjunejune/inference_stack.sh	Wed Aug 05 09:19:41 2026 -0700
@@ -0,0 +1,111 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+server="$(realpath "$1")"
+sidecar_launcher="$(realpath "$2")"
+sidecar_zip="$(realpath "$3")"
+copilot_cli="$(realpath "$4")"
+litellm_zip="$(realpath "$5")"
+litellm_config="$(realpath "$6")"
+python="$(realpath "$7")"
+
+: "${LITELLM_MASTER_KEY:?LITELLM_MASTER_KEY must be set}"
+
+litellm_host="${LITELLM_HOST:-127.0.0.1}"
+litellm_port="${LITELLM_PORT:-4000}"
+default_state_root="${XDG_STATE_HOME:-$HOME/.local/state}/mrjunejune/inference"
+state_root="${MRJUNEJUNE_INFERENCE_STATE:-$default_state_root}"
+token_dir="${GITHUB_COPILOT_TOKEN_DIR:-$state_root/litellm-copilot}"
+mkdir -p "$state_root/copilot" "$token_dir"
+chmod 700 "$state_root" "$state_root/copilot" "$token_dir"
+
+export GITHUB_COPILOT_TOKEN_DIR="$token_dir"
+if [[ ! -s "$GITHUB_COPILOT_TOKEN_DIR/access-token" ]]; then
+  echo "GitHub Copilot is not authenticated for LiteLLM." >&2
+  echo "Run: bazel run //mrjunejune/inference:litellm_proxy -- --authenticate --token-dir \"$GITHUB_COPILOT_TOKEN_DIR\"" >&2
+  exit 1
+fi
+
+litellm_pid=
+server_pid=
+
+cleanup() {
+  for pid in "$server_pid" "$litellm_pid"; do
+    if [[ -n "$pid" ]]; then
+      kill -- "-$pid" 2>/dev/null || true
+    fi
+  done
+  for pid in "$server_pid" "$litellm_pid"; do
+    [[ -n "$pid" ]] || continue
+    for _ in $(seq 1 30); do
+      if ! kill -0 -- "-$pid" 2>/dev/null; then
+        break
+      fi
+      sleep 0.1
+    done
+    if kill -0 -- "-$pid" 2>/dev/null; then
+      kill -KILL -- "-$pid" 2>/dev/null || true
+    fi
+    wait "$pid" 2>/dev/null || true
+  done
+}
+trap cleanup EXIT INT TERM
+
+setsid "$python" "$litellm_zip" \
+  --config "$litellm_config" \
+  --host "$litellm_host" \
+  --port "$litellm_port" &
+litellm_pid=$!
+
+for _ in $(seq 1 150); do
+  if ! kill -0 "$litellm_pid" 2>/dev/null; then
+    echo "LiteLLM exited before becoming ready" >&2
+    exit 1
+  fi
+  if curl --fail --silent \
+      --connect-timeout 1 \
+      --max-time 2 \
+      "http://${litellm_host}:${litellm_port}/health/readiness" \
+      >/dev/null; then
+    break
+  fi
+  sleep 0.1
+done
+
+if ! curl --fail --silent \
+    --connect-timeout 1 \
+    --max-time 2 \
+    "http://${litellm_host}:${litellm_port}/health/readiness" \
+    >/dev/null; then
+  echo "LiteLLM readiness timed out" >&2
+  exit 1
+fi
+
+export MRJUNEJUNE_PYTHON_PATH="$python"
+export MRJUNEJUNE_SIDECAR_ZIP="$sidecar_zip"
+export MRJUNEJUNE_INFERENCE_SIDECAR_PATH="$sidecar_launcher"
+export MRJUNEJUNE_COPILOT_CLI_PATH="$copilot_cli"
+export COPILOT_SIDECAR_HOME="$state_root/copilot"
+export LITELLM_BASE_URL="http://${litellm_host}:${litellm_port}/v1"
+export LITELLM_MODEL="${LITELLM_MODEL:-jrpg-copilot}"
+export LITELLM_WIRE_API="${LITELLM_WIRE_API:-completions}"
+export LITELLM_API_KEY="$LITELLM_MASTER_KEY"
+export MRJUNEJUNE_ALLOW_ANONYMOUS_INFERENCE="${MRJUNEJUNE_ALLOW_ANONYMOUS_INFERENCE:-0}"
+
+setsid "$server" &
+server_pid=$!
+while true; do
+  if ! kill -0 "$server_pid" 2>/dev/null; then
+    set +e
+    wait "$server_pid"
+    server_status=$?
+    set -e
+    exit "$server_status"
+  fi
+  if ! kill -0 "$litellm_pid" 2>/dev/null; then
+    wait "$litellm_pid" 2>/dev/null || true
+    echo "LiteLLM exited while the public server was running" >&2
+    exit 1
+  fi
+  sleep 0.2
+done