import subprocess import json import time import httpx import logging import os from concurrent.futures import ThreadPoolExecutor, as_completed from typing import List, Dict, Any, Optional from urllib.parse import urlparse logger = logging.getLogger(__name__) # Cache for discovered hosts _hosts_cache: List[str] = [] _hosts_cache_time: float = 0 _HOSTS_CACHE_TTL = 60 # seconds def discover_tailscale_hosts() -> List[str]: """Discover online Tailscale peers, returning their IPv4 addresses.""" global _hosts_cache, _hosts_cache_time now = time.time() if _hosts_cache and (now - _hosts_cache_time) < _HOSTS_CACHE_TTL: return list(_hosts_cache) hosts = [] try: result = subprocess.run( ["tailscale", "status", "--json"], capture_output=True, text=True, timeout=5 ) if result.returncode != 0: return hosts data = json.loads(result.stdout) # Add self self_ips = data.get("Self", {}).get("TailscaleIPs", []) for ip in self_ips: if "." in ip: # IPv4 only hosts.append(ip) break # Add online peers (skip funnel-ingress-nodes and android devices) for peer in data.get("Peer", {}).values(): if not peer.get("Online"): continue hostname = peer.get("HostName", "") if hostname == "funnel-ingress-node": continue os_name = peer.get("OS", "") if os_name == "android": continue peer_ips = peer.get("TailscaleIPs", []) for ip in peer_ips: if "." in ip: # IPv4 only hosts.append(ip) break _hosts_cache = hosts _hosts_cache_time = now logger.info(f"Tailscale discovery found {len(hosts)} hosts: {hosts}") except FileNotFoundError: logger.debug("tailscale command not found") except Exception as e: logger.warning(f"Tailscale discovery failed: {e}") return hosts class ModelDiscovery: def __init__(self, default_host: str, openai_api_key: Optional[str] = None): self.default_host = default_host self.openai_api_key = openai_api_key self.openai_compat_path = "/v1/chat/completions" def _get_hosts(self) -> List[str]: """Get all hosts to scan, using env override, Tailscale, or default.""" def _append_host(out: List[str], host: str) -> None: host = (host or "").strip() if not host or host in out: return out.append(host) # Manual override takes priority extra = os.getenv("LLM_HOSTS", "").strip() if extra: hosts = [h.strip() for h in extra.split(",") if h.strip()] # Always include the default host too if self.default_host not in hosts: hosts.insert(0, self.default_host) _append_host(hosts, "host.docker.internal") return hosts # Try Tailscale discovery ts_hosts = discover_tailscale_hosts() if ts_hosts: # Ensure default_host is included if self.default_host not in ts_hosts: ts_hosts.insert(0, self.default_host) _append_host(ts_hosts, "host.docker.internal") return ts_hosts hosts = [self.default_host] # Docker desktop/Linux compose maps this to the host machine. That is # the common "I started Ollama normally on this computer" case. _append_host(hosts, "host.docker.internal") for env_name in ("OLLAMA_BASE_URL", "OLLAMA_URL"): raw = os.getenv(env_name, "").strip() if not raw: continue try: parsed = urlparse(raw if "://" in raw else "http://" + raw) _append_host(hosts, parsed.hostname or "") except Exception: pass return hosts def _check_port(self, host: str, port: int) -> Optional[Dict[str, Any]]: """Check a single host:port for models.""" base = f"http://{host}:{port}/v1" try: r = httpx.get(f"{base}/models", timeout=3) if not r.is_success: return None data = r.json() or {} ids = [m.get("id") for m in (data.get("data") or []) if m.get("id")] if ids: return { "host": host, "port": port, "url": f"http://{host}:{port}{self.openai_compat_path}", "models": ids, "models_display": [i.lstrip("/") for i in ids] } except Exception: pass return None def discover_models(self) -> Dict[str, List[Dict[str, Any]]]: """Discover available models from all reachable hosts.""" hosts = self._get_hosts() items = [] logger.info(f"Scanning {len(hosts)} hosts for models: {hosts}") # Build list of (host, port) to check. 8000-8020 catches vLLM, # llama.cpp, SGLang, and Cookbook serves; 11434 catches Ollama. ports = list(range(8000, 8021)) + [11434] targets = [(h, p) for h in hosts for p in ports] seen_models = set() # dedupe by (port, model_ids) to avoid same machine via different IPs with ThreadPoolExecutor(max_workers=50) as pool: futures = {pool.submit(self._check_port, h, p): (h, p) for h, p in targets} for future in as_completed(futures): result = future.result() if result: key = (result["port"], tuple(sorted(result["models"]))) if key not in seen_models: seen_models.add(key) items.append(result) # Sort by host then port for consistent ordering items.sort(key=lambda x: (x["host"], x["port"])) logger.info(f"Discovered {len(items)} model endpoints across {len(hosts)} hosts") return {"hosts": hosts, "items": items} def get_providers(self) -> Dict[str, Any]: """Get all available providers""" discovery = self.discover_models() items = discovery["items"] providers = [{"provider": "vllm", "hosts": discovery["hosts"], "items": items}] if self.openai_api_key: openai_models = [ "gpt-5.2-codex", "gpt-4o-mini", "gpt-image-1.5", "gpt-4o", "gpt-5.2", "gpt-5.2-pro", ] providers.append({ "provider": "openai", "items": [{ "url": "https://api.openai.com/v1/chat/completions", "models": openai_models }] }) return {"providers": providers}