LamChingFung-2425-Anti-FraudX

Feature: Model Switch API

Category: Secondary Version: 4.1 Last Updated: 2026-03-11

Overview

Allows runtime switching between Ollama and Gemini without restarting the backend server. The active provider is reflected immediately in all subsequent LLM calls.

API Endpoints

File: backend/api/model_switch_routes.py

Method Path Description
GET /api/model/current Get current provider name and active model info
POST /api/model/switch Switch provider ({"provider": "gemini"} or {"provider": "ollama"})
GET /api/model/validate Validate current provider config (API key, reachability)
GET /api/model/config Full config dump — model names, token limits, URLs

How Switching Works

  1. POST /api/model/switch updates the GEMINI_ENABLED runtime flag in LlmFactory.
  2. No server restart needed — the flag is read on every create_llm() call.
  3. Per-agent model overrides (e.g. GEMINI_MODEL_SCAMMER) remain unchanged.
  4. The frontend dashboard (frontend/index.html) shows the current provider status.

LlmFactory Methods Used

Method Purpose
get_current_provider() Returns 'gemini' or 'ollama'
get_provider_info() Returns dict: provider name, model IDs, token limits, base URL
validate_gemini_config() Checks GEMINI_API_KEY is set and models are valid

Relevant Files

backend/api/model_switch_routes.py   REST API endpoints
backend/llms/llm_factory.py          get_current_provider(), get_provider_info()
backend/.env                         GEMINI_ENABLED, model name overrides