fix(gemini): implement dynamic API versioning and support Gemini 3

- Switch to v1beta endpoint for 'preview' and 'thinking' models.
- Update model version checks to include gemini-3 as a known version.
- Use get_base_url helper to construct dynamic URLs for both streaming and non-streaming requests.

src/providers/gemini.rs

@@ -439,6 +439,17 @@ impl GeminiProvider {
     }
 }
 
+    /// Determine the appropriate base URL for the model.
+    /// "preview" models often require the v1beta endpoint.
+    fn get_base_url(&self, model: &str) -> String {
+        if model.contains("preview") || model.contains("thinking") {
+            self.config.base_url.replace("/v1", "/v1beta")
+        } else {
+            self.config.base_url.clone()
+        }
+    }
+}
+
 #[async_trait]
 impl super::Provider for GeminiProvider {
     fn name(&self) -> &str {
@@ -454,7 +465,22 @@ impl super::Provider for GeminiProvider {
     }
 
     async fn chat_completion(&self, request: UnifiedRequest) -> Result<ProviderResponse, AppError> {
-        let model = request.model.clone();
+        let mut model = request.model.clone();
+        
+        // Normalize model name: If it's a known Gemini model version, use it;
+        // otherwise, if it starts with gemini- but is an unknown legacy version, 
+        // fallback to the default model to avoid 400 errors.
+        // We now allow gemini-3+ as valid versions.
+        let is_known_version = model.starts_with("gemini-1.5") || 
+                              model.starts_with("gemini-2.0") || 
+                              model.starts_with("gemini-2.5") ||
+                              model.starts_with("gemini-3");
+
+        if !is_known_version && model.starts_with("gemini-") {
+            tracing::info!("Mapping unknown Gemini model {} to default {}", model, self.config.default_model);
+            model = self.config.default_model.clone();
+        }
+
         let tools = Self::convert_tools(&request);
         let tool_config = Self::convert_tool_config(&request);
         let (contents, system_instruction) = Self::convert_messages(request.messages.clone()).await?;
@@ -464,9 +490,14 @@ impl super::Provider for GeminiProvider {
         }
 
         let generation_config = if request.temperature.is_some() || request.max_tokens.is_some() {
+            // Some Gemini models (especially 1.5) have lower max_output_tokens limits (e.g. 8192)
+            // than what clients like opencode might request. Clamp to a safe maximum.
+            // Note: Gemini 2.0+ supports much higher limits, but 8192 is a safe universal floor.
+            let max_tokens = request.max_tokens.map(|t| t.min(8192));
+            
             Some(GeminiGenerationConfig {
                 temperature: request.temperature,
-                max_output_tokens: request.max_tokens,
+                max_output_tokens: max_tokens,
             })
         } else {
             None
@@ -480,7 +511,8 @@ impl super::Provider for GeminiProvider {
             tool_config,
         };
 
-        let url = format!("{}/models/{}:generateContent", self.config.base_url, model);
+        let base_url = self.get_base_url(&model);
+        let url = format!("{}/models/{}:generateContent", base_url, model);
 
         let response = self
             .client
@@ -579,7 +611,19 @@ impl super::Provider for GeminiProvider {
         &self,
         request: UnifiedRequest,
     ) -> Result<BoxStream<'static, Result<ProviderStreamChunk, AppError>>, AppError> {
-        let model = request.model.clone();
+        let mut model = request.model.clone();
+        
+        // Normalize model name: fallback to default if unknown Gemini model is requested
+        let is_known_version = model.starts_with("gemini-1.5") || 
+                              model.starts_with("gemini-2.0") || 
+                              model.starts_with("gemini-2.5") ||
+                              model.starts_with("gemini-3");
+
+        if !is_known_version && model.starts_with("gemini-") {
+            tracing::info!("Mapping unknown Gemini model {} to default {}", model, self.config.default_model);
+            model = self.config.default_model.clone();
+        }
+
         let tools = Self::convert_tools(&request);
         let tool_config = Self::convert_tool_config(&request);
         let (contents, system_instruction) = Self::convert_messages(request.messages.clone()).await?;
@@ -589,9 +633,13 @@ impl super::Provider for GeminiProvider {
         }
 
         let generation_config = if request.temperature.is_some() || request.max_tokens.is_some() {
+            // Some Gemini models (especially 1.5) have lower max_output_tokens limits (e.g. 8192)
+            // than what clients like opencode might request. Clamp to a safe maximum.
+            let max_tokens = request.max_tokens.map(|t| t.min(8192));
+
             Some(GeminiGenerationConfig {
                 temperature: request.temperature,
-                max_output_tokens: request.max_tokens,
+                max_output_tokens: max_tokens,
             })
         } else {
             None
@@ -605,9 +653,10 @@ impl super::Provider for GeminiProvider {
             tool_config,
         };
 
+        let base_url = self.get_base_url(&model);
         let url = format!(
             "{}/models/{}:streamGenerateContent?alt=sse",
-            self.config.base_url, model,
+            base_url, model,
         );
 
         // (no fallback_request needed here)
@@ -622,7 +671,7 @@ impl super::Provider for GeminiProvider {
         // Prepare clones for HTTP fallback usage inside non-streaming paths.
         let http_client = self.client.clone();
         let http_api_key = self.api_key.clone();
-        let http_base = self.config.base_url.clone();
+        let http_base = base_url.clone();
         let gemini_request_clone = gemini_request.clone();
 
         let es_result = reqwest_eventsource::EventSource::new(