feat: add AUTO_CREATE_DATASET env var to control dataset auto-creation

When set to false, HuggingClaw will not auto-create the dataset repo
and will prompt the user to create it manually on HuggingFace.
Default: true (existing behavior preserved).

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

.env.example

@@ -49,6 +49,13 @@ HF_TOKEN=hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
 #
 OPENCLAW_DATASET_REPO=your-username/openclaw-data
 
+# Whether to auto-create the Dataset repo if it doesn't exist.
+# Set to false if you prefer to create the repo manually on HuggingFace.
+#
+# [OPTIONAL]  Default: true
+#
+# AUTO_CREATE_DATASET=true
+
 # How often (in seconds) to back up data to the Dataset repo.
 # Lower values = safer but more API calls to HuggingFace.
 #
@@ -57,26 +64,40 @@ OPENCLAW_DATASET_REPO=your-username/openclaw-data
 # SYNC_INTERVAL=120
 
 
-# ─── LLM PROVIDER ────────────────────────────────────────────────────────
+# ─── LLM / OPENAI-COMPATIBLE API ───────────────────────────────────────────
 #
-# OpenClaw needs at least one LLM provider to power AI conversations.
-# OpenRouter is recommended — one API key, 200+ models, free tier included.
+# OpenClaw supports any OpenAI-compatible API. Set the API key for the
+# provider(s) you use. See OpenClaw docs: https://openclawdoc.com/docs/reference/environment-variables
 #
-# OpenRouter API Key
-# Get one at: https://openrouter.ai/keys
-# Free-tier models: stepfun/step-3.5-flash:free, deepseek/deepseek-chat:free
+# OpenAI (or any OpenAI-compatible endpoint)
+# Use OPENAI_API_KEY alone for api.openai.com, or set OPENAI_BASE_URL for
+# compatible endpoints (e.g. OpenRouter, local LLM servers, Azure OpenAI).
 #
-# [RECOMMENDED]
+# [RECOMMENDED] At least one of the following for AI conversations
+#
+OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
+
+# Optional: base URL for OpenAI-compatible API (default: https://api.openai.com/v1)
+# Examples: https://openrouter.ai/api/v1, http://localhost:11434/v1 (Ollama), etc.
+#
+# OPENAI_BASE_URL=https://api.openai.com/v1
+
+# OpenRouter — one key, 200+ models, free tier: https://openrouter.ai/keys
 #
 OPENROUTER_API_KEY=sk-or-v1-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
 
-# Default model for new conversations.
-# Must be a model available from your configured provider.
-# Format: provider/model-id  (e.g., openrouter/deepseek/deepseek-chat:free)
+# Other providers (OpenClaw reads these from the environment)
+#
+# ANTHROPIC_API_KEY=sk-ant-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
+# GOOGLE_API_KEY=AIza...
+# MISTRAL_API_KEY=mis-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
+# COHERE_API_KEY=co-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
+
+# Default model for new conversations (must exist in your configured provider).
 #
-# [OPTIONAL]  Default: openrouter/stepfun/step-3.5-flash:free
+# [OPTIONAL]  Examples: openai/gpt-4o-mini, openrouter/deepseek/deepseek-chat:free
 #
-# OPENCLAW_DEFAULT_MODEL=openrouter/deepseek/deepseek-chat:free
+# OPENCLAW_DEFAULT_MODEL=openai/gpt-4o-mini
 
 
 # ─── PERFORMANCE ──────────────────────────────────────────────────────────
@@ -97,3 +118,20 @@ OPENROUTER_API_KEY=sk-or-v1-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
 # [OPTIONAL]  Default: UTC
 #
 # TZ=Asia/Shanghai
+
+
+# ─── OPENCLAW (from official docs) ─────────────────────────────────────────
+#
+# Optional overrides. See https://openclawdoc.com/docs/reference/environment-variables
+#
+# OPENCLAW_HOME=~/.openclaw
+# OPENCLAW_STATE_DIR=~/.openclaw
+# OPENCLAW_CONFIG_PATH=~/.openclaw/openclaw.json
+# OPENCLAW_LOG_LEVEL=info
+# OPENCLAW_API_PORT=8080
+# OPENCLAW_WS_PORT=8081
+# OPENCLAW_HOST=0.0.0.0
+# OLLAMA_HOST=http://localhost:11434
+# OPENCLAW_HTTP_PROXY=
+# OPENCLAW_HTTPS_PROXY=
+# OPENCLAW_NO_PROXY=

README.md

@@ -8,6 +8,26 @@ pinned: false
 license: mit
 short_description: Deploy OpenClaw on HuggingFace Spaces
 app_port: 7860
+tags:
+  - chatbot
+  - llm
+  - openclaw
+  - ai-assistant
+  - whatsapp
+  - telegram
+  - text-generation
+  - openai-api
+  - huggingface-spaces
+  - docker
+  - deployment
+  - persistent-storage
+  - agents
+  - multi-channel
+  - openai-compatible
+  - free-tier
+  - one-click-deploy
+  - self-hosted
+  - messaging-bot
 ---
 
 <div align="center">
@@ -63,7 +83,11 @@ Go to **Settings → Repository secrets** and configure:
 | `OPENCLAW_PASSWORD` | Recommended | Password for the Control UI (default: `huggingclaw`) |
 | `HF_TOKEN` | **Required** | HF Access Token with write permission ([create one](https://huggingface.co/settings/tokens)) |
 | `OPENCLAW_DATASET_REPO` | **Required** | Dataset repo for backup, e.g. `your-name/openclaw-data` |
-| `OPENROUTER_API_KEY` | Recommended | [OpenRouter](https://openrouter.ai) API key for LLM access |
+| `OPENAI_API_KEY` | Recommended | OpenAI (or any [OpenAI-compatible](https://openclawdoc.com/docs/reference/environment-variables)) API key for LLM |
+| `OPENROUTER_API_KEY` | Optional | [OpenRouter](https://openrouter.ai) API key (200+ models, free tier) |
+| `ANTHROPIC_API_KEY` | Optional | Anthropic Claude API key |
+| `GOOGLE_API_KEY` | Optional | Google / Gemini API key |
+| `OPENCLAW_DEFAULT_MODEL` | Optional | Default model, e.g. `openai/gpt-4o-mini` or `openrouter/deepseek/deepseek-chat:free` |
 
 > For the full list of environment variables, see [`.env.example`](.env.example).
 
@@ -80,8 +104,8 @@ HuggingClaw is configured entirely through **environment variables**. A fully do
 | Category | Variables | Purpose |
 |----------|-----------|---------|
 | **Security** | `OPENCLAW_PASSWORD` | Protect the Control UI with a password |
-| **Persistence** | `HF_TOKEN`, `OPENCLAW_DATASET_REPO`, `SYNC_INTERVAL` | Auto-backup to HF Dataset |
-| **LLM** | `OPENROUTER_API_KEY`, `OPENCLAW_DEFAULT_MODEL` | Power AI conversations |
+| **Persistence** | `HF_TOKEN`, `OPENCLAW_DATASET_REPO`, `AUTO_CREATE_DATASET`, `SYNC_INTERVAL` | Auto-backup to HF Dataset |
+| **LLM (OpenAI-compatible)** | `OPENAI_API_KEY`, `OPENAI_BASE_URL`, `OPENROUTER_API_KEY`, `ANTHROPIC_API_KEY`, `GOOGLE_API_KEY`, `MISTRAL_API_KEY`, `COHERE_API_KEY`, `OPENCLAW_DEFAULT_MODEL` | Power AI conversations ([OpenClaw env reference](https://openclawdoc.com/docs/reference/environment-variables)) |
 | **Performance** | `NODE_MEMORY_LIMIT` | Tune Node.js memory usage |
 | **Locale** | `TZ` | Set timezone for logs |
 

scripts/sync_hf.py

@@ -62,20 +62,27 @@ TELEGRAM_BOT_TOKEN  = os.environ.get("TELEGRAM_BOT_TOKEN", "")
 TELEGRAM_BOT_NAME   = os.environ.get("TELEGRAM_BOT_NAME", "")
 TELEGRAM_ALLOW_USER = os.environ.get("TELEGRAM_ALLOW_USER", "")
 
-# OpenRouter API key for LLM access
+# OpenAI-compatible API (OpenAI, OpenRouter, or any compatible endpoint)
+OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY", "")
+OPENAI_BASE_URL = os.environ.get("OPENAI_BASE_URL", "https://api.openai.com/v1").rstrip("/")
+
+# OpenRouter API key (optional; alternative to OPENAI_API_KEY + OPENAI_BASE_URL)
 OPENROUTER_API_KEY = os.environ.get("OPENROUTER_API_KEY", "")
 
 # Gateway password (override via HF Secret OPENCLAW_PASSWORD)
 OPENCLAW_PASSWORD = os.environ.get("OPENCLAW_PASSWORD", "huggingclaw")
 
-# Default model for new conversations
-OPENCLAW_DEFAULT_MODEL = os.environ.get("OPENCLAW_DEFAULT_MODEL", "openrouter/stepfun/step-3.5-flash:free")
+# Default model for new conversations (infer from provider if not set)
+OPENCLAW_DEFAULT_MODEL = os.environ.get("OPENCLAW_DEFAULT_MODEL") or (
+    "openai/gpt-4o-mini" if OPENAI_API_KEY else "openrouter/stepfun/step-3.5-flash:free"
+)
 
 # HF Spaces built-in env vars (auto-set by HF runtime)
 SPACE_HOST = os.environ.get("SPACE_HOST", "")   # e.g. "tao-shen-huggingclaw.hf.space"
 SPACE_ID   = os.environ.get("SPACE_ID", "")      # e.g. "tao-shen/HuggingClaw"
 
 SYNC_INTERVAL = int(os.environ.get("SYNC_INTERVAL", "120"))
+AUTO_CREATE_DATASET = os.environ.get("AUTO_CREATE_DATASET", "true").lower() in ("true", "1", "yes")
 
 # Setup logging
 log_dir = OPENCLAW_HOME / "workspace"
@@ -107,12 +114,17 @@ class OpenClawFullSync:
     # ── Repo management ────────────────────────────────────────────────
 
     def _ensure_repo_exists(self):
-        """Check if dataset repo exists; auto-create if not."""
+        """Check if dataset repo exists; auto-create if AUTO_CREATE_DATASET is enabled."""
         try:
             self.api.repo_info(repo_id=HF_REPO_ID, repo_type="dataset")
             print(f"[SYNC] Dataset repo found: {HF_REPO_ID}")
             return True
         except Exception:
+            if not AUTO_CREATE_DATASET:
+                print(f"[SYNC] Dataset repo NOT found: {HF_REPO_ID}")
+                print(f"[SYNC] AUTO_CREATE_DATASET is disabled. Please create the dataset repo manually.")
+                print(f"[SYNC]   → https://huggingface.co/new-dataset")
+                return False
             print(f"[SYNC] Dataset repo NOT found: {HF_REPO_ID} - creating...")
             try:
                 self.api.create_repo(
@@ -273,12 +285,17 @@ class OpenClawFullSync:
                 if "gateway" in cfg and "auth" in cfg["gateway"]:
                     if cfg["gateway"]["auth"].get("password") == "__OPENCLAW_PASSWORD__":
                         cfg["gateway"]["auth"]["password"] = OPENCLAW_PASSWORD
+                if OPENAI_API_KEY and "models" in cfg and "providers" in cfg["models"] and "openai" in cfg["models"]["providers"]:
+                    cfg["models"]["providers"]["openai"]["apiKey"] = OPENAI_API_KEY
+                    if OPENAI_BASE_URL:
+                        cfg["models"]["providers"]["openai"]["baseUrl"] = OPENAI_BASE_URL
+                elif "models" in cfg and "providers" in cfg["models"]:
+                    if not OPENAI_API_KEY:
+                        cfg["models"]["providers"].pop("openai", None)
                 if OPENROUTER_API_KEY:
-                    # Replace placeholder with actual key
                     if "models" in cfg and "providers" in cfg["models"] and "openrouter" in cfg["models"]["providers"]:
                         cfg["models"]["providers"]["openrouter"]["apiKey"] = OPENROUTER_API_KEY
                 else:
-                    # No API key: remove provider entirely to avoid config validation error
                     if "models" in cfg and "providers" in cfg["models"]:
                         cfg["models"]["providers"].pop("openrouter", None)
                     print("[SYNC] No OPENROUTER_API_KEY — removed openrouter provider from config")
@@ -368,8 +385,16 @@ class OpenClawFullSync:
             data.setdefault("agents", {}).setdefault("defaults", {}).setdefault("model", {})
             data.setdefault("session", {})["scope"] = "global"
 
-            # Force OpenRouter provider
+            # OpenAI-compatible provider (OPENAI_API_KEY + optional OPENAI_BASE_URL)
             data.setdefault("models", {}).setdefault("providers", {})
+            if OPENAI_API_KEY:
+                data["models"]["providers"]["openai"] = {
+                    "baseUrl": OPENAI_BASE_URL,
+                    "apiKey": OPENAI_API_KEY,
+                    "api": "openai-completions",
+                }
+                print(f"[SYNC] Set OpenAI-compatible provider (baseUrl={OPENAI_BASE_URL})")
+            # OpenRouter provider (optional)
             if OPENROUTER_API_KEY:
                 data["models"]["providers"]["openrouter"] = {
                     "baseUrl": "https://openrouter.ai/api/v1",
@@ -380,9 +405,9 @@ class OpenClawFullSync:
                         {"id": "deepseek/deepseek-chat:free", "name": "DeepSeek V3 (Free)"}
                     ]
                 }
-            else:
-                print("[SYNC] WARNING: OPENROUTER_API_KEY not set, skipping provider config")
-            # Remove old gemini provider if present
+                print("[SYNC] Set OpenRouter provider")
+            if not OPENAI_API_KEY and not OPENROUTER_API_KEY:
+                print("[SYNC] WARNING: No OPENAI_API_KEY or OPENROUTER_API_KEY set, LLM features may not work")
             data["models"]["providers"].pop("gemini", None)
             data["agents"]["defaults"]["model"]["primary"] = OPENCLAW_DEFAULT_MODEL
 
@@ -492,13 +517,15 @@ class OpenClawFullSync:
         # Open log file
         log_fh = open(log_file, "a")
 
-        # Prepare environment with required variables
+        # Prepare environment (all API keys passed through for OpenClaw)
         env = os.environ.copy()
+        if OPENAI_API_KEY:
+            env["OPENAI_API_KEY"] = OPENAI_API_KEY
+            env["OPENAI_BASE_URL"] = OPENAI_BASE_URL
         if OPENROUTER_API_KEY:
             env["OPENROUTER_API_KEY"] = OPENROUTER_API_KEY
-            print(f"[SYNC] Setting OPENROUTER_API_KEY environment variable")
-        else:
-            print(f"[SYNC] WARNING: OPENROUTER_API_KEY not set, LLM features may not work")
+        if not OPENAI_API_KEY and not OPENROUTER_API_KEY:
+            print(f"[SYNC] WARNING: No OPENAI_API_KEY or OPENROUTER_API_KEY set, LLM features may not work")
         try:
             # Use Popen without shell to avoid pipe issues
             # auth disabled in config — no token needed