fix: restore foreground sync + stdout logging for HF runtime log API

- Restore foreground sync (init runs before port 7860 opens) so HF
captures sync progress in its runtime log SSE stream
- Switch all log() functions from stderr to stdout via tee (writes to
both stdout and logfile simultaneously)
- Add startup_duration_timeout: 1h to README.md YAML to prevent HF
from timing out during init/restore
- Remove diagnostic echo/python test lines from entrypoint.sh
- Add scripts/stream_logs.py for JWT-based log streaming
- Update scripts/curl_logs.sh with cached token fallback

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

README.md

@@ -7,6 +7,7 @@ sdk: docker
 pinned: false
 license: mit
 app_port: 7860
+startup_duration_timeout: 1h
 tags:
   - docker
   - huggingface-spaces

scripts/curl_logs.sh

@@ -1,7 +1,7 @@
-#!/bin/sh
-# 直接用你给的指令查看远端：Bearer token + curl run/build 日志
-# 用法: ./scripts/curl_logs.sh [run|build]  默认拉 run
-# 需要 .env 中有 HF_TOKEN 或环境变量 HF_TOKEN
+#!/bin/bash
+# Stream HF Space runtime/build logs via JWT-based SSE endpoint
+# Usage: ./scripts/curl_logs.sh [run|build]  (default: run)
+# Requires HF_TOKEN in .env or environment
 
 set -e
 cd "$(dirname "$0")/.."
@@ -9,6 +9,11 @@ if [ -f .env ]; then
   HF_TOKEN="${HF_TOKEN:-$(awk -F= '/^HF_TOKEN=/ {print $2; exit}' .env | tr -d '\r\n" ')}"
   export HF_TOKEN
 fi
+# Fallback to cached token
+if [ -z "$HF_TOKEN" ] && [ -f ~/.cache/huggingface/token ]; then
+  HF_TOKEN="$(cat ~/.cache/huggingface/token)"
+  export HF_TOKEN
+fi
 if [ -z "$HF_TOKEN" ]; then
   echo "HF_TOKEN not set (add to .env or export)" >&2
   exit 1
@@ -17,8 +22,12 @@ fi
 SPACE_ID="${SPACE_ID:-tao-shen/HuggingRun}"
 LOG="${1:-run}"
 
-# Get container logs (SSE)
-# Get build logs (SSE)
-curl -N \
+echo "=== Streaming ${LOG} logs for ${SPACE_ID} ===" >&2
+echo "=== Method 1: Direct API (huggingface.co) ===" >&2
+echo "=== Press Ctrl+C to stop ===" >&2
+echo "" >&2
+
+# Method 1: Direct HF API (works for most cases)
+curl -sN \
   -H "Authorization: Bearer $HF_TOKEN" \
   "https://huggingface.co/api/spaces/$SPACE_ID/logs/$LOG"

scripts/entrypoint.sh

@@ -1,23 +1,17 @@
 #!/bin/bash
 # ─────────────────────────────────────────────────────────────────────
 # HuggingRun Entrypoint
-# 1. Start services FIRST (so HF doesn't timeout on port 7860)
-# 2. Run init (download + restore) in BACKGROUND
-# 3. Start sync-loop daemon in BACKGROUND
+# 1. Init: download dataset + restore filesystem (foreground, logs visible)
+# 2. Start sync-loop daemon in background
+# 3. exec user command (start-server.sh opens port 7860)
 # ─────────────────────────────────────────────────────────────────────
 set -e
 
-# Diagnostic: test if HF captures ANY container output
-echo "=== [stdout] ENTRYPOINT PID=$$ $(date -u) ==="
-echo "=== [stderr] ENTRYPOINT PID=$$ $(date -u) ===" >&2
-python3 -c "import sys; print('=== [py-stdout] PYTHON TEST ===', flush=True); print('=== [py-stderr] PYTHON TEST ===', file=sys.stderr, flush=True)"
-
 LOGFILE="/var/log/huggingrun.log"
 touch "$LOGFILE"
 
 log() {
-    echo "$@" >> "$LOGFILE"
-    echo "$@" >&2
+    echo "$@" | tee -a "$LOGFILE"
 }
 
 log "========================================"
@@ -77,20 +71,20 @@ export PERSIST_PATH="${PERSIST_PATH:-/data}"
 ENVEOF
 log "[entrypoint] Wrote /etc/huggingrun.env"
 
-# Background: init (download + restore) then start sync-loop
-# This runs AFTER services start so HF doesn't timeout
-# NO redirects here — stderr must flow to container stderr for HF log API
-(
-    echo "[entrypoint:bg] Starting init in background ..." >&2
-    python3 -u /opt/git_sync_daemon.py init
-    echo "[entrypoint:bg] Init done, starting sync-loop ..." >&2
-    exec python3 -u /opt/git_sync_daemon.py sync-loop
-) &
-BG_PID=$!
-log "[entrypoint] Background init+sync PID=${BG_PID}"
+# Step 1: Init — download dataset + restore filesystem (foreground so logs are visible)
+log "[entrypoint] ── Step 1: Init (download + restore) ──"
+python3 -u /opt/git_sync_daemon.py init
+log "[entrypoint] ── Init done ──"
+
+# Step 2: Start background sync-loop daemon
+log "[entrypoint] ── Step 2: Starting sync-loop daemon ──"
+python3 -u /opt/git_sync_daemon.py sync-loop &
+SYNC_PID=$!
+log "[entrypoint] Sync daemon PID=${SYNC_PID}"
 
-# Start services immediately (so HF sees port 7860)
+# Step 3: Run user command (opens port 7860)
 CMD="${RUN_CMD:-python3 /app/demo_app.py}"
-log "[entrypoint] Starting services: ${CMD}"
+log "[entrypoint] ── Step 3: exec user command ──"
+log "[entrypoint] CMD=${CMD}"
 log "========================================"
 exec $CMD

scripts/stream_logs.py

@@ -0,0 +1,105 @@
+#!/usr/bin/env python3
+"""Stream HF Space runtime logs via JWT-authenticated SSE endpoint.
+
+Usage:
+    python3 scripts/stream_logs.py [run|build]
+
+This uses the same endpoint as the HF web UI's "Container logs" tab:
+  1. Fetch JWT from https://huggingface.co/api/spaces/{id}/jwt
+  2. Stream SSE from https://api.hf.space/v1/{id}/logs/{level}
+"""
+import json
+import os
+import sys
+
+def get_token():
+    """Get HF token from env, .env file, or cached token."""
+    token = os.environ.get("HF_TOKEN", "")
+    if not token:
+        env_file = os.path.join(os.path.dirname(__file__), "..", ".env")
+        if os.path.exists(env_file):
+            for line in open(env_file):
+                if line.startswith("HF_TOKEN="):
+                    token = line.split("=", 1)[1].strip().strip('"').strip("'")
+                    break
+    if not token:
+        cached = os.path.expanduser("~/.cache/huggingface/token")
+        if os.path.exists(cached):
+            token = open(cached).read().strip()
+    return token
+
+def stream_logs(space_id, level="run"):
+    import requests
+
+    token = get_token()
+    if not token:
+        print("ERROR: HF_TOKEN not found", file=sys.stderr)
+        sys.exit(1)
+
+    print(f"=== Streaming {level} logs for {space_id} ===", file=sys.stderr)
+
+    # Step 1: Get JWT token
+    print("Fetching JWT token...", file=sys.stderr)
+    resp = requests.get(
+        f"https://huggingface.co/api/spaces/{space_id}/jwt",
+        headers={"Authorization": f"Bearer {token}"},
+    )
+    if resp.status_code != 200:
+        print(f"JWT fetch failed ({resp.status_code}): {resp.text}", file=sys.stderr)
+        print("Falling back to direct API...", file=sys.stderr)
+        # Fallback to direct API
+        resp = requests.get(
+            f"https://huggingface.co/api/spaces/{space_id}/logs/{level}",
+            headers={"Authorization": f"Bearer {token}"},
+            stream=True,
+        )
+        for line in resp.iter_lines():
+            if line:
+                line = line.decode("utf-8", errors="replace")
+                if line.startswith("data: "):
+                    try:
+                        event = json.loads(line[6:])
+                        ts = event.get("timestamp", "")
+                        data = event.get("data", "")
+                        print(f"[{ts}] {data}", end="", flush=True)
+                    except json.JSONDecodeError:
+                        print(line, flush=True)
+                else:
+                    print(line, flush=True)
+        return
+
+    jwt = resp.json()["token"]
+    print(f"JWT obtained (expires in token payload)", file=sys.stderr)
+
+    # Step 2: Stream logs via SSE
+    print(f"Connecting to api.hf.space SSE stream...", file=sys.stderr)
+    print("---", file=sys.stderr)
+
+    resp = requests.get(
+        f"https://api.hf.space/v1/{space_id}/logs/{level}",
+        headers={"Authorization": f"Bearer {jwt}"},
+        stream=True,
+    )
+
+    if resp.status_code != 200:
+        print(f"SSE stream failed ({resp.status_code}): {resp.text}", file=sys.stderr)
+        sys.exit(1)
+
+    for line in resp.iter_lines():
+        if line:
+            line = line.decode("utf-8", errors="replace")
+            if line.startswith("data: "):
+                try:
+                    event = json.loads(line[6:])
+                    ts = event.get("timestamp", "")
+                    data = event.get("data", "")
+                    print(f"[{ts}] {data}", end="", flush=True)
+                except json.JSONDecodeError:
+                    print(line, flush=True)
+            else:
+                print(line, flush=True)
+
+if __name__ == "__main__":
+    space_id = os.environ.get("SPACE_ID", "tao-shen/HuggingRun")
+    level = sys.argv[1] if len(sys.argv) > 1 else "run"
+    stream_logs(space_id, level)

ubuntu-server/git_sync_daemon.py

@@ -46,7 +46,7 @@ LOGFILE = "/var/log/huggingrun.log"
 def log(msg):
     ts = time.strftime("%H:%M:%S", time.gmtime())
     line = f"[sync {ts}] {msg}"
-    print(line, file=sys.stderr, flush=True)
+    print(line, flush=True)
     try:
         with open(LOGFILE, "a") as f:
             f.write(line + "\n")

ubuntu-server/start-server.sh

@@ -10,11 +10,10 @@ LOGFILE="/var/log/huggingrun.log"
 export SSH_PORT="${SSH_PORT:-2222}"
 export TTYD_PORT="${TTYD_PORT:-7681}"
 
-# Log to file + stderr (HF may or may not capture stderr)
+# Log to file + stdout (HF captures container stdout/stderr)
 log() {
     local msg="$*"
-    echo "$msg" >> "$LOGFILE"
-    echo "$msg" >&2
+    echo "$msg" | tee -a "$LOGFILE"
 }
 
 # ── Start nginx FIRST so HF sees port 7860 ──────────────────────