feat(deploy): add Render + HF Inference API cloud deploy support

.env.example

@@ -0,0 +1,3 @@
+# Copy to .env and fill in your values — never commit .env
+HF_TOKEN=your_huggingface_write_token_here
+HF_DEFAULT_MODEL=Qwen/Qwen2.5-72B-Instruct

.gitignore

@@ -58,6 +58,7 @@ tools/test_batch*.json
 # Secrets / credentials
 .env
 .env.*
+!.env.example
 *.key
 *.pem
 credentials*

Procfile

@@ -0,0 +1 @@
+web: python prompt_expert_enhance.py web --port $PORT --no-browser

prompt_expert_enhance.py

@@ -65,6 +65,13 @@
 OLLAMA_URL = "http://localhost:11434/api/generate"
 BASE_DIR = Path(__file__).resolve().parent
 
+# Cloud deploy: set HF_TOKEN env var to switch to HF Inference API backend.
+# All Ollama checks are skipped automatically when this is set.
+_HF_TOKEN: str = os.environ.get("HF_TOKEN", "")
+_CLOUD_MODE: bool = bool(_HF_TOKEN)
+_HF_INFERENCE_URL = "https://api-inference.huggingface.co/v1/chat/completions"
+_HF_DEFAULT_MODEL = os.environ.get("HF_DEFAULT_MODEL", "Qwen/Qwen2.5-72B-Instruct")
+
 
 def _is_frozen() -> bool:
     """True when running inside a PyInstaller-compiled app, not from source."""
@@ -324,6 +331,8 @@ def is_ollama_installed() -> bool:
 
 
 def is_ollama_running() -> bool:
+    if _CLOUD_MODE:
+        return True
     try:
         requests.get(f"{OLLAMA_API_BASE}/api/tags", timeout=3)
         return True
@@ -439,7 +448,20 @@ def start_ollama_serve():
         print(f"  [!] Could not start Ollama: {e}")
 
 
+_HF_AVAILABLE_MODELS = [
+    "Qwen/Qwen2.5-72B-Instruct",
+    "Qwen/Qwen2.5-Coder-32B-Instruct",
+    "meta-llama/Llama-3.3-70B-Instruct",
+    "mistralai/Mistral-7B-Instruct-v0.3",
+    "mistralai/Mixtral-8x7B-Instruct-v0.1",
+    "google/gemma-3-27b-it",
+    "microsoft/Phi-4-reasoning",
+]
+
+
 def list_local_models() -> List[Dict[str, str]]:
+    if _CLOUD_MODE:
+        return [{"name": m, "size": "cloud", "modified": ""} for m in _HF_AVAILABLE_MODELS]
     base = _BACKEND_API_BASE or OLLAMA_API_BASE
     if _BACKEND_TYPE == "openai_compatible":
         try:
@@ -554,6 +576,8 @@ def pick_model_interactive(label: str, current: str) -> str:
 
 
 def ensure_ollama_ready():
+    if _CLOUD_MODE:
+        return
     if not is_ollama_installed():
         if not install_ollama_interactive():
             print("\n  [!] Ollama is not available. Generation will fail.")
@@ -1283,8 +1307,9 @@ def query_ollama(
     if _BACKEND_TYPE == "openai_compatible":
         is_chat = _is_chat_endpoint(ollama_url)
         payload = _openai_compatible_payload(model, prompt, temperature, num_predict, ollama_url, stream=False)
+        headers = {"Authorization": f"Bearer {_HF_TOKEN}"} if _HF_TOKEN else {}
         try:
-            resp = requests.post(ollama_url, json=payload, timeout=timeout)
+            resp = requests.post(ollama_url, json=payload, headers=headers, timeout=timeout)
             resp.raise_for_status()
             data = resp.json()
             result = _openai_compatible_extract_text(data["choices"][0], is_chat) or "[Empty response]"
@@ -1353,9 +1378,10 @@ def query_ollama_stream(
     if _BACKEND_TYPE == "openai_compatible":
         is_chat = _is_chat_endpoint(ollama_url)
         payload = _openai_compatible_payload(model, prompt, temperature, num_predict, ollama_url, stream=True)
+        headers = {"Authorization": f"Bearer {_HF_TOKEN}"} if _HF_TOKEN else {}
         full_text = []
         try:
-            with requests.post(ollama_url, json=payload, timeout=timeout, stream=True) as resp:
+            with requests.post(ollama_url, json=payload, headers=headers, timeout=timeout, stream=True) as resp:
                 resp.raise_for_status()
                 for raw_line in resp.iter_lines():
                     if not raw_line:

render.yaml

@@ -0,0 +1,12 @@
+services:
+  - type: web
+    name: wild-root-prompt
+    runtime: python
+    buildCommand: pip install -r requirements.txt
+    startCommand: python prompt_expert_enhance.py web --port $PORT --no-browser
+    envVars:
+      - key: HF_TOKEN
+        sync: false        # set this in Render dashboard — never commit the value
+      - key: HF_DEFAULT_MODEL
+        value: Qwen/Qwen2.5-72B-Instruct
+    healthCheckPath: /api/status

web_server.py

@@ -30,6 +30,7 @@
     load_settings, save_settings, PRE_PROCESSOR_TIMEOUT,
     is_ollama_running, ensure_ollama_ready,
     set_backend_type, set_backend_api_base,
+    _CLOUD_MODE, _HF_INFERENCE_URL,
     PROMPT_TEMPLATES,
 )
 
@@ -458,7 +459,9 @@
     const r = await fetch('/api/status')
     const d = await r.json()
     $('ollama-dot').className = 'dot ' + (d.ollama ? 'green' : 'red')
-    $('ollama-status').textContent = d.ollama ? 'Ollama running — ready' : 'Ollama not running — start it first'
+    $('ollama-status').textContent = d.ollama
+      ? (d.cloud ? 'HF Inference API — ready' : 'Ollama running — ready')
+      : 'Ollama not running — start it first'
   } catch { $('ollama-dot').className = 'dot red'; $('ollama-status').textContent = 'Server error' }
 }
 
@@ -918,7 +921,7 @@ def index():
 
 @app.route("/api/status")
 def api_status():
-    return jsonify({"ollama": is_ollama_running()})
+    return jsonify({"ollama": is_ollama_running(), "cloud": _CLOUD_MODE})
 
 
 @app.route("/api/models")
@@ -1096,8 +1099,12 @@ def generate_sse():
 
 def run_web_server(port: int = 7860, open_browser: bool = True):
     startup_settings = load_settings()
-    set_backend_type(startup_settings.get("backend_type", "ollama"))
-    set_backend_api_base(startup_settings.get("ollama_url", OLLAMA_URL))
+    if _CLOUD_MODE:
+        set_backend_type("openai_compatible")
+        set_backend_api_base(_HF_INFERENCE_URL)
+    else:
+        set_backend_type(startup_settings.get("backend_type", "ollama"))
+        set_backend_api_base(startup_settings.get("ollama_url", OLLAMA_URL))
 
     print(f"\n  Wild_Root_Prompt Web UI")
     print(f"  ─────────────────────────────────────")