scripts: add update_hf_env.sh and redeploy.sh -- close runbook gaps

update_hf_env.sh: sets the 6 HF Space variables (RIPRAP_LLM_PRIMARY,
RIPRAP_LLM_BASE_URL, RIPRAP_LLM_API_KEY, RIPRAP_ML_BACKEND,
RIPRAP_ML_BASE_URL, RIPRAP_ML_API_KEY) via huggingface_hub Python API
(add_space_variable / restart_space -- huggingface-cli space does not
exist in v0.36.x), then polls /api/backend up to 120 s for HTTP 200.

redeploy.sh: thin orchestrator -- generates a fresh token, calls
deploy_droplet.sh, calls update_hf_env.sh, calls probe_addresses.py,
prints a pass/fail summary. Exit 0 only when all three steps pass.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

scripts/redeploy.sh

@@ -0,0 +1,118 @@
+#!/usr/bin/env bash
+# Full redeploy to an existing AMD MI300X droplet.
+#
+#   1. Generate a fresh bearer token
+#   2. scripts/deploy_droplet.sh <ip> <token>   (bring up vLLM + riprap-models)
+#   3. scripts/update_hf_env.sh <ip> <token>    (update HF Space vars + restart)
+#   4. .venv/bin/python scripts/probe_addresses.py  (5/5 must pass)
+#
+# Usage: scripts/redeploy.sh <droplet-ip>
+#
+# Requires:
+#   HF_TOKEN  env var with write access to the HF Space
+#   .venv     Python virtual environment with probe_addresses.py deps
+#   SSH access to the droplet (ssh-agent or SSH_KEY env var)
+#
+# Exit codes:
+#   0  all three steps passed
+#   1  deploy_droplet.sh failed (HF Space NOT touched)
+#   1  update_hf_env.sh failed (droplet is up but HF Space NOT updated)
+#   1  probe_addresses.py failed (deploy + HF update succeeded; not rolled back)
+set -euo pipefail
+
+if [ "$#" -ne 1 ]; then
+    echo "Usage: $0 <droplet-ip>" >&2
+    exit 1
+fi
+
+IP="$1"
+
+if [ -z "${HF_TOKEN:-}" ]; then
+    echo "Error: HF_TOKEN env var is required (write access to the HF Space)" >&2
+    exit 1
+fi
+
+REPO_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
+START_SECONDS=$SECONDS
+
+DEPLOY_STATUS="FAIL"
+HF_STATUS="FAIL"
+PROBE_STATUS="FAIL"
+
+# ---- 1. Generate a fresh bearer token ------------------------------------
+# openssl rand -base64 24 produces 32 chars; strip +/= to keep URL-safe.
+TOKEN=$(openssl rand -base64 24 | tr -d '/+=')
+echo "==> Deploying to ${IP} with fresh token..."
+echo
+
+# ---- 2. deploy_droplet.sh ------------------------------------------------
+if bash "${REPO_ROOT}/scripts/deploy_droplet.sh" "$IP" "$TOKEN"; then
+    DEPLOY_STATUS="PASS"
+else
+    echo "deploy_droplet.sh failed" >&2
+    # Print summary before exiting so the caller sees partial state.
+    ELAPSED=$(( SECONDS - START_SECONDS ))
+    echo
+    echo "=== redeploy summary ==="
+    echo "Droplet IP : ${IP}"
+    echo "Token      : (not set — deploy failed before token was registered)"
+    echo "Deploy     : ${DEPLOY_STATUS}"
+    echo "HF Space   : ${HF_STATUS}"
+    echo "E2E probe  : ${PROBE_STATUS}"
+    printf "Total time : %dm%02ds\n" $(( ELAPSED / 60 )) $(( ELAPSED % 60 ))
+    exit 1
+fi
+
+echo
+echo "==> Deploy succeeded. Updating HF Space..."
+echo
+
+# ---- 3. update_hf_env.sh -------------------------------------------------
+if bash "${REPO_ROOT}/scripts/update_hf_env.sh" "$IP" "$TOKEN"; then
+    HF_STATUS="PASS"
+else
+    echo "update_hf_env.sh failed. HF Space NOT updated." >&2
+    ELAPSED=$(( SECONDS - START_SECONDS ))
+    echo
+    echo "=== redeploy summary ==="
+    echo "Droplet IP : ${IP}"
+    echo "Token      : (regenerated, see HF Space vars)"
+    echo "Deploy     : ${DEPLOY_STATUS}"
+    echo "HF Space   : ${HF_STATUS}"
+    echo "E2E probe  : ${PROBE_STATUS}"
+    printf "Total time : %dm%02ds\n" $(( ELAPSED / 60 )) $(( ELAPSED % 60 ))
+    exit 1
+fi
+
+echo
+echo "==> HF Space updated. Running end-to-end probe..."
+echo
+
+# ---- 4. probe_addresses.py -----------------------------------------------
+# probe_addresses.py exits 0 only when 5/5 pass (from docs/DROPLET-RUNBOOK.md).
+# Disable set -e for this step so we can capture the exit code and still
+# print the summary.
+set +e
+"${REPO_ROOT}/.venv/bin/python" "${REPO_ROOT}/scripts/probe_addresses.py"
+PROBE_EXIT=$?
+set -e
+
+if [ "$PROBE_EXIT" -eq 0 ]; then
+    PROBE_STATUS="PASS"
+else
+    PROBE_STATUS="FAIL"
+fi
+
+# ---- 5. Summary ----------------------------------------------------------
+ELAPSED=$(( SECONDS - START_SECONDS ))
+echo
+echo "=== redeploy summary ==="
+echo "Droplet IP : ${IP}"
+echo "Token      : (regenerated, see HF Space vars)"
+echo "Deploy     : ${DEPLOY_STATUS}"
+echo "HF Space   : ${HF_STATUS}"
+echo "E2E probe  : ${PROBE_STATUS}"
+printf "Total time : %dm%02ds\n" $(( ELAPSED / 60 )) $(( ELAPSED % 60 ))
+
+# Exit 1 if probe failed; deploy + HF update already succeeded, not rolling back.
+[ "$PROBE_STATUS" = "PASS" ]

scripts/update_hf_env.sh

@@ -0,0 +1,117 @@
+#!/usr/bin/env bash
+# Update HF Space env vars to point at a new droplet, restart the Space,
+# and poll until the agent endpoint returns HTTP 200.
+#
+# Usage: scripts/update_hf_env.sh <droplet-ip> <bearer-token>
+#
+# Requires:
+#   HF_TOKEN  env var with write access to the Space
+#   huggingface_hub >= 0.36 installed (provides the Python API used below;
+#   note: 'huggingface-cli space variables' does not exist in this version)
+#
+# Space slug: lablab-ai-amd-developer-hackathon/riprap-nyc
+# Variables set (from docs/DROPLET-RUNBOOK.md §Required secrets):
+#   RIPRAP_LLM_PRIMARY    vllm
+#   RIPRAP_LLM_BASE_URL   http://<ip>:8001/v1
+#   RIPRAP_LLM_API_KEY    <token>
+#   RIPRAP_ML_BACKEND     remote
+#   RIPRAP_ML_BASE_URL    http://<ip>:7860
+#   RIPRAP_ML_API_KEY     <token>
+set -euo pipefail
+
+if [ "$#" -ne 2 ]; then
+    echo "Usage: $0 <droplet-ip> <bearer-token>" >&2
+    exit 1
+fi
+
+IP="$1"
+TOKEN="$2"
+
+if [ -z "${HF_TOKEN:-}" ]; then
+    echo "Error: HF_TOKEN env var is required (write access to the Space)" >&2
+    exit 1
+fi
+
+SPACE_ID="lablab-ai-amd-developer-hackathon/riprap-nyc"
+SPACE_URL="https://lablab-ai-amd-developer-hackathon-riprap-nyc.hf.space"
+VLLM_PORT=8001
+MODELS_PORT=7860
+
+echo "==> Updating HF Space variables"
+echo "    space:        ${SPACE_ID}"
+echo "    droplet ip:   ${IP}"
+echo "    vLLM port:    ${VLLM_PORT}"
+echo "    models port:  ${MODELS_PORT}"
+echo
+
+# ---- 1. Set all six Space variables via the huggingface_hub Python API ----
+# huggingface-cli space variables does not exist in huggingface_hub 0.36.x;
+# add_space_variable is the documented programmatic interface.
+python3 -c "
+import sys, os
+try:
+    from huggingface_hub import HfApi
+except ImportError:
+    print('Error: huggingface_hub not installed', file=sys.stderr)
+    sys.exit(1)
+
+api = HfApi(token=os.environ['HF_TOKEN'])
+space_id = '${SPACE_ID}'
+ip = '${IP}'
+token = '${TOKEN}'
+vllm_port = ${VLLM_PORT}
+models_port = ${MODELS_PORT}
+
+variables = {
+    'RIPRAP_LLM_PRIMARY':  'vllm',
+    'RIPRAP_LLM_BASE_URL': f'http://{ip}:{vllm_port}/v1',
+    'RIPRAP_LLM_API_KEY':  token,
+    'RIPRAP_ML_BACKEND':   'remote',
+    'RIPRAP_ML_BASE_URL':  f'http://{ip}:{models_port}',
+    'RIPRAP_ML_API_KEY':   token,
+}
+
+for key, value in variables.items():
+    display = '<redacted>' if 'KEY' in key else value
+    print(f'    setting {key} = {display}')
+    api.add_space_variable(repo_id=space_id, key=key, value=value)
+
+print('[python] all 6 variables set')
+"
+echo
+
+# ---- 2. Restart the Space ------------------------------------------------
+echo "==> Restarting HF Space"
+python3 -c "
+import os
+from huggingface_hub import HfApi
+api = HfApi(token=os.environ['HF_TOKEN'])
+rt = api.restart_space(repo_id='${SPACE_ID}')
+print(f'    stage after restart request: {rt.stage}')
+"
+echo
+
+# ---- 3. Poll /api/backend until HTTP 200 (max 120 s) ---------------------
+# /api/backend is documented in docs/DROPLET-RUNBOOK.md §Destroy checklist
+# as the endpoint to verify the Space is serving.
+echo "==> Polling ${SPACE_URL}/api/backend (up to 120 s)..."
+DEADLINE=$((SECONDS + 120))
+HEALTHY=0
+while (( SECONDS < DEADLINE )); do
+    HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" \
+        --max-time 10 "${SPACE_URL}/api/backend" 2>/dev/null || echo "000")
+    if [ "$HTTP_CODE" = "200" ]; then
+        HEALTHY=1
+        break
+    fi
+    echo "    (${HTTP_CODE}) not ready yet — waiting 10 s..."
+    sleep 10
+done
+
+if [ "$HEALTHY" -ne 1 ]; then
+    echo "HF Space did not become healthy within 120s" >&2
+    exit 1
+fi
+
+echo
+echo "HF Space updated and healthy. IP=${IP}"