Enable on-the-fly server construction

jimmy/config.py

@@ -1,6 +1,7 @@
 import os
 import tomllib
 import logging
+import urllib.parse
 from typing import Callable
 
 import httpx
@@ -14,7 +15,7 @@ class ServerConfig(BaseModel):
     base_url: AnyHttpUrl
     gpu: bool = False
     vram_gb: int = 4
-    throttle: bool = False
+    default_model: str = "llama3:latest"
 
     def __repr__(self):
         return "<ServerConfig name={0.name} base_url={0.base_url} gpu={0.gpu!s} vram_gb={0.vram_gb}>".format(self)
@@ -57,6 +58,40 @@ def get_server(name_or_base_url: str) -> ServerConfig | None:
     for server in servers:
         if server.name == name_or_base_url or server.base_url == name_or_base_url:
             return server
+
+    try:
+        parsed = urllib.parse.urlparse(name_or_base_url)
+    except ValueError:
+        pass
+    else:
+        if parsed.netloc and parsed.scheme in ["http", "https"]:
+            defaults = {
+                "name": ":temporary:",
+                "base_url": "{0.scheme}://{0.netloc}".format(parsed),
+                "gpu": False,
+                "vram_gb": 2,
+                "default_model": "orca-mini:3b"
+            }
+            if parsed.path and parsed.path.endswith(("/api", "/api/")):
+                defaults["base_url"] += parsed.path
+            parsed_qs = urllib.parse.parse_qs(parsed.query)
+            for key, values in parsed_qs.items():
+                if not values:
+                    continue
+                if key == "gpu":
+                    values = [
+                        values[0][0].lower() in ("t", "1", "y")
+                    ]
+                elif key == "vram_gb":
+                    try:
+                        values = [
+                            int(values[0])
+                        ]
+                    except ValueError:
+                        values = []
+                if values:
+                    defaults[key] = values[0]
+            return ServerConfig(**defaults)
     return None