fix: qwen3.5 reasoning model outputs to thinking field not content

- ollamaResponse now reads both content and thinking fields
- Falls back to thinking if content is empty
- Install script adds model warm-up prompt to preload into memory
- Warm-up uses OLLAMA_HOST for correct user context
- Makes first real receipt analysis faster

install.sh

@@ -213,9 +213,20 @@ case "$AI_CHOICE" in
             sleep 2
         done
         info "Pulling model $AI_MODEL (this may take a while)..."
-        # Pull explicitly via the API URL so it works regardless of which user
+        # Pull via the API URL so it works regardless of which user runs it
-        # runs the command (the Ollama service listens on 127.0.0.1:11434)
+        OLLAMA_HOST="http://127.0.0.1:11434"
-        sudo_if sh -c "OLLAMA_HOST=http://127.0.0.1:11434 ollama pull $AI_MODEL" 2>&1 | tail -3
+        export OLLAMA_HOST
+        if id -u ollama &>/dev/null; then
+            sudo -u ollama sh -c "OLLAMA_HOST=$OLLAMA_HOST ollama pull $AI_MODEL" 2>&1 | tail -3
+        else
+            ollama pull "$AI_MODEL" 2>&1 | tail -3
+        fi
+        # Warm-up: send a quick prompt to preload the model into memory
+        # so the first user request isn't delayed by model loading.
+        info "Warming up $AI_MODEL..."
+        curl -s --max-time 120 -X POST "$OLLAMA_HOST/api/chat" \
+            -d "{\"model\":\"$AI_MODEL\",\"messages\":[{\"role\":\"user\",\"content\":\"say ok\"}],\"stream\":false,\"options\":{\"num_predict\":10}}" > /dev/null 2>&1 && \
+            info "Model ready" || warn "Model warm-up timed out (will load on first use)"
         AI_BASE_URL="http://localhost:11434"
         ;;
     *)

internal/ai/ollama.go

@@ -37,6 +37,7 @@ type ollamaMessage struct {
 type ollamaResponse struct {
 	Message struct {
 		Content  string `json:"content"`
+		Thinking string `json:"thinking"`
 	} `json:"message"`
 	Error string `json:"error,omitempty"`
 }
@@ -106,7 +107,11 @@ func (p *ollamaProvider) ExtractReceipt(imagePath string) (*ReceiptData, error)
 		return &ReceiptData{}, fmt.Errorf("Ollama error: %s", apiResp.Error)
 	}
 
-	contentStr := stripMarkdownFences(apiResp.Message.Content)
+	contentStr := apiResp.Message.Content
+	if contentStr == "" {
+		contentStr = apiResp.Message.Thinking
+	}
+	contentStr = stripMarkdownFences(contentStr)
 	var receipt ReceiptData
 	if err := json.Unmarshal([]byte(contentStr), &receipt); err != nil {
 		return &ReceiptData{}, fmt.Errorf("parse JSON: %w (content: %s)", err, contentStr)