resolve image test failures

Author: OriPekelman

lib/ruby_llm/models.json

@@ -4572,7 +4572,7 @@
     "max_tokens": 28800,
     "type": "chat",
     "family": "small",
-    "supports_vision": false,
+    "supports_vision": true,
     "supports_functions": true,
     "supports_json_mode": true,
     "input_price_per_million": 0.2,
@@ -4591,7 +4591,7 @@
     "max_tokens": 28800,
     "type": "chat",
     "family": "small",
-    "supports_vision": false,
+    "supports_vision": true,
     "supports_functions": true,
     "supports_json_mode": true,
     "input_price_per_million": 0.2,
@@ -4610,7 +4610,7 @@
     "max_tokens": 28800,
     "type": "chat",
     "family": "small",
-    "supports_vision": false,
+    "supports_vision": true,
     "supports_functions": true,
     "supports_json_mode": true,
     "input_price_per_million": 0.2,

lib/ruby_llm/providers/mistral/capabilities.rb

@@ -51,8 +51,14 @@ def output_price_for(model_id)
         # @param model_id [String] the model identifier
         # @return [Boolean] true if the model supports vision
         def supports_vision?(model_id)
-          # Determine vision support based on model ID pattern
-          model_id.match?(/pixtral/)
+          # Explicitly match the known vision-capable models
+          vision_models = [
+            'pixtral-12b-latest',
+            'pixtral-large-latest',
+            'mistral-medium-latest',
+            'mistral-small-latest'
+          ]
+          vision_models.any? { |id| model_id.include?(id) }
         end
 
         # Determines if the model supports function calling
@@ -96,10 +102,11 @@ def model_type(model_id)
         end
 
         # Determines if the model supports structured output
+        # based on the docs in https://docs.mistral.ai/capabilities/structured-output/structured_output_overview/ all mistral models support JSON mode
         # @param model_id [String] the model identifier
         # @return [Boolean] true if the model supports structured output
         def supports_structured_output?(model_id)
-          !model_id.match?(/embed|moderation/)
+          true
         end
 
         # Determines the model family for pricing and capability lookup

lib/ruby_llm/providers/mistral/chat.rb

@@ -61,25 +61,27 @@ def render_message(message)
           result = {}
           result[:role] = message.role
 
-          # Handle content formatting (text or multimodal array)
-          if message.content.is_a?(Array)
-            # Multimodal content: format each part
-            formatted_content = message.content.compact.map do |item|
-              if item.is_a?(Hash) && item[:type] == "image"
-                # Use the media helper to format image data correctly for Mistral
-                Mistral::Media.format_image_content(item)
-              elsif item.is_a?(Hash) && item[:type] == "text"
-                # Pass through text hashes
-                item
-              elsif item.is_a?(String)
-                 # Wrap plain strings in the text hash format
-                 { type: "text", text: item }
-              else
-                # Skip unknown item types in content array
-                RubyLLM.logger.warn "[MISTRAL WARN] Unknown item type in message content array: #{item.inspect}"
-                nil
+          # If the message content is a RubyLLM::Content with attachments, convert to multimodal array
+          if message.content.is_a?(RubyLLM::Content)
+            content = message.content
+            if content.attachments.any?
+              multimodal = []
+              multimodal << { type: "text", text: content.text } if content.text
+              content.attachments.each do |attachment|
+                case attachment
+                when RubyLLM::Attachments::Image
+                  multimodal << { type: "image", source: attachment.url? ? { url: attachment.source } : { type: 'base64', media_type: attachment.mime_type, data: attachment.encoded } }
+                # Add more attachment types if Mistral supports them
+                end
               end
-            end.compact # Remove any nils from formatting errors
+              # Always format multimodal for Mistral
+              result[:content] = Mistral::Media.format_content(multimodal)
+            else
+              result[:content] = content.text
+            end
+          elsif message.content.is_a?(Array)
+            # Multimodal content: format each part
+            formatted_content = Mistral::Media.format_content(message.content.compact)
             result[:content] = formatted_content unless formatted_content.empty?
           else
             # Simple text content