Add timeout + lower temperature for AI calls

internal/api/ai.go

@@ -7,6 +7,7 @@ import (
 	"net/http"
 	"os"
 	"strings"
+	"time"
 
 	"markdownhub/internal/files"
 )
@@ -158,10 +159,8 @@ func (s *Server) handleAIChat(w http.ResponseWriter, r *http.Request) {
 
 	if req.Mode == "edit" {
 		systemPrompt = "You are a document editor. The user will give you a markdown document and an instruction. " +
-			"Apply the instruction to the document and return ONLY the complete updated document. " +
-			"Do not add explanations, comments, or wrap the output in code fences. " +
-			"Do not prefix with any markers. Preserve the document's existing style and formatting. " +
-			"Return the raw markdown content only."
+			"Apply the instruction and return the COMPLETE updated document. " +
+			"Do not add explanations or wrap in code fences. Return raw markdown only."
 		userMsg = "Document:\n\n" + req.Content + "\n\nInstruction: " + req.Message
 	} else {
 		systemPrompt = `You are a helpful writing assistant. The user has a markdown document open and is asking a question about it.
@@ -201,6 +200,7 @@ func callLLM(endpoint, apiKey, model, systemPrompt, userContent string) (string,
 			{"role": "system", "content": systemPrompt},
 			{"role": "user", "content": userContent},
 		},
+		"temperature": 0.3,
 	}
 	body, _ := json.Marshal(payload)
 
@@ -214,7 +214,8 @@ func callLLM(endpoint, apiKey, model, systemPrompt, userContent string) (string,
 		req.Header.Set("Authorization", "Bearer "+apiKey)
 	}
 
-	resp, err := http.DefaultClient.Do(req)
+	client := &http.Client{Timeout: 120 * time.Second}
+	resp, err := client.Do(req)
 	if err != nil {
 		return "", err
 	}