feat: migrate Gemini provider from OpenAI-compatible shim to native A…

Cargo.toml

@@ -21,6 +21,7 @@ rig = { version = "0.33", package = "rig-core", features = ["derive"] }
 
 # HTTP clients for LLM providers
 reqwest = { version = "0.13", features = ["json", "stream", "form", "query", "gzip"] }
+gemini-rust = "1.7"
 
 # Databases
 sqlx = { version = "0.8", features = ["runtime-tokio", "sqlite", "migrate", "chrono", "uuid"] }

src/config/providers.rs

@@ -25,7 +25,7 @@ pub(super) const MISTRAL_PROVIDER_BASE_URL: &str = "https://api.mistral.ai";
 pub(super) const NVIDIA_PROVIDER_BASE_URL: &str = "https://integrate.api.nvidia.com";
 pub(super) const FIREWORKS_PROVIDER_BASE_URL: &str = "https://api.fireworks.ai/inference";
 pub(crate) const GEMINI_PROVIDER_BASE_URL: &str =
-    "https://generativelanguage.googleapis.com/v1beta/openai";
+    "https://generativelanguage.googleapis.com/v1beta/";
 pub(super) const GITHUB_COPILOT_DEFAULT_BASE_URL: &str = "https://api.individual.githubcopilot.com";
 
 /// App attribution headers sent with every OpenRouter API request.

src/llm.rs

@@ -1,6 +1,7 @@
 //! LLM provider management and routing.
 
 pub mod anthropic;
+pub mod gemini;
 pub mod manager;
 pub mod model;
 pub mod pricing;