fix(admin): LLM provider/model switching now reads DB overrides correctly

- get_llm_provider() now accepts provider + model args so DB overrides
  propagate through to all provider constructors (was always reading
  env vars, ignoring the admin UI selection)
- /test-llm replaced with lightweight ping (max_tokens=20) instead of
  running a full bill analysis; shows model name + reply, no truncation
- /api/settings/llm-models endpoint fetches available models live from
  each provider's API (OpenAI, Anthropic REST, Gemini, Ollama)
- Admin UI model picker dynamically populated from provider API;
  falls back to manual text input on error; Custom model name option kept
- Default Gemini model updated: gemini-1.5-pro → gemini-2.0-flash

Co-Authored-By: Jack Levy

backend/app/services/llm_service.py

@@ -185,10 +185,10 @@ class LLMProvider(ABC):
 
 
 class OpenAIProvider(LLMProvider):
-    def __init__(self):
+    def __init__(self, model: str | None = None):
         from openai import OpenAI
         self.client = OpenAI(api_key=settings.OPENAI_API_KEY)
-        self.model = settings.OPENAI_MODEL
+        self.model = model or settings.OPENAI_MODEL
 
     def generate_brief(self, doc_text: str, bill_metadata: dict) -> ReverseBrief:
         prompt = build_prompt(doc_text, bill_metadata, MAX_TOKENS_DEFAULT)
@@ -220,10 +220,10 @@ class OpenAIProvider(LLMProvider):
 
 
 class AnthropicProvider(LLMProvider):
-    def __init__(self):
+    def __init__(self, model: str | None = None):
         import anthropic
         self.client = anthropic.Anthropic(api_key=settings.ANTHROPIC_API_KEY)
-        self.model = settings.ANTHROPIC_MODEL
+        self.model = model or settings.ANTHROPIC_MODEL
 
     def generate_brief(self, doc_text: str, bill_metadata: dict) -> ReverseBrief:
         prompt = build_prompt(doc_text, bill_metadata, MAX_TOKENS_DEFAULT)
@@ -249,11 +249,11 @@ class AnthropicProvider(LLMProvider):
 
 
 class GeminiProvider(LLMProvider):
-    def __init__(self):
+    def __init__(self, model: str | None = None):
         import google.generativeai as genai
         genai.configure(api_key=settings.GEMINI_API_KEY)
         self._genai = genai
-        self.model_name = settings.GEMINI_MODEL
+        self.model_name = model or settings.GEMINI_MODEL
 
     def _make_model(self, system_prompt: str):
         return self._genai.GenerativeModel(
@@ -274,9 +274,9 @@ class GeminiProvider(LLMProvider):
 
 
 class OllamaProvider(LLMProvider):
-    def __init__(self):
+    def __init__(self, model: str | None = None):
         self.base_url = settings.OLLAMA_BASE_URL.rstrip("/")
-        self.model = settings.OLLAMA_MODEL
+        self.model = model or settings.OLLAMA_MODEL
 
     def _generate(self, system_prompt: str, user_prompt: str) -> str:
         import requests as req
@@ -327,15 +327,20 @@ class OllamaProvider(LLMProvider):
             return parse_brief_json(raw2, "ollama", self.model)
 
 
-def get_llm_provider() -> LLMProvider:
-    """Factory — returns the configured LLM provider."""
-    provider = settings.LLM_PROVIDER.lower()
+def get_llm_provider(provider: str | None = None, model: str | None = None) -> LLMProvider:
+    """Factory — returns the configured LLM provider.
+
+    Pass ``provider`` and/or ``model`` explicitly (e.g. from DB overrides) to bypass env defaults.
+    """
+    if provider is None:
+        provider = settings.LLM_PROVIDER
+    provider = provider.lower()
     if provider == "openai":
-        return OpenAIProvider()
+        return OpenAIProvider(model=model)
     elif provider == "anthropic":
-        return AnthropicProvider()
+        return AnthropicProvider(model=model)
     elif provider == "gemini":
-        return GeminiProvider()
+        return GeminiProvider(model=model)
     elif provider == "ollama":
-        return OllamaProvider()
+        return OllamaProvider(model=model)
     raise ValueError(f"Unknown LLM_PROVIDER: '{provider}'. Must be one of: openai, anthropic, gemini, ollama")