feat: Remove execute button feature, add copy buttons for free version

app.js

@@ -55,7 +55,7 @@
     btn.onclick = async () => {
       btn.disabled = true;
       btn.textContent = "Executing...";
-      
+
       try {
         const response = await fetch("/chat/execute", {
           method: "POST",
@@ -64,7 +64,7 @@
         });
 
         const result = await response.json();
-        
+
         if (result.success) {
           const outputText = result.stdout || result.stderr || "(no output)";
           appendMessage(output, "Command Output", `Exit Code: ${result.exit_code}\n\n${outputText}`, false);
@@ -80,7 +80,7 @@
         btn.textContent = "✗ Error";
         btn.style.background = "#ef4444";
       }
-      
+
       setTimeout(() => {
         btn.disabled = false;
         btn.textContent = "Execute";
@@ -94,102 +94,98 @@
     const regex = /<execute>(.*?)<\/execute>/gs;
     const commands = [];
     let match;
-    
+
     while ((match = regex.exec(text)) !== null) {
       commands.push(match[1].trim());
     }
-    
+
     return commands;
   }
 
+  // NEW: Add copy buttons to all code blocks in a container
+  function addCopyButtonsToCodeBlocks(container) {
+    // Find all <pre> elements that don't already have a copy button
+    container.querySelectorAll("pre").forEach((pre) => {
+      // Skip if already has a copy button wrapper
+      if (pre.parentElement && pre.parentElement.classList.contains("code-wrapper")) {
+        return;
+      }
+
+      // Get the code content
+      const codeEl = pre.querySelector("code");
+      const code = codeEl ? codeEl.textContent : pre.textContent;
+
+      if (!code || !code.trim()) return;
+
+      // Create wrapper
+      const wrapper = document.createElement("div");
+      wrapper.className = "code-wrapper";
+
+      // Create copy button
+      const copyBtn = createCopyButton(code);
+
+      // Insert wrapper before pre
+      pre.parentNode.insertBefore(wrapper, pre);
+      
+      // Move pre into wrapper and add button
+      wrapper.appendChild(copyBtn);
+      wrapper.appendChild(pre);
+    });
+  }
+
   function appendMessage(container, sender, message, useMarkdown = false) {
     const msgDiv = document.createElement("div");
     msgDiv.classList.add("message");
 
     // Check for execute tags
     const commands = extractExecuteTags(message);
-    
+
     if (commands.length > 0) {
       // Remove execute tags from display
       let displayText = message.replace(/<execute>.*?<\/execute>/gs, "");
-      
+
       msgDiv.innerHTML = `<strong>${sender}:</strong>${useMarkdown ? renderMarkdown(displayText) : `<pre>${escapeHtml(displayText)}</pre>`}`;
-      
+
       // Add execute buttons for each command
       commands.forEach(cmd => {
         const cmdWrapper = document.createElement("div");
         cmdWrapper.className = "command-wrapper";
-        
+
         const cmdPre = document.createElement("pre");
         cmdPre.className = "command-block";
         cmdPre.textContent = cmd;
-        
+
         const btnContainer = document.createElement("div");
         btnContainer.className = "command-buttons";
-        
+
         const executeBtn = createExecuteButton(cmd, container);
         const copyBtn = createCopyButton(cmd);
-        
+
         btnContainer.appendChild(executeBtn);
         btnContainer.appendChild(copyBtn);
-        
+
         cmdWrapper.appendChild(cmdPre);
         cmdWrapper.appendChild(btnContainer);
         msgDiv.appendChild(cmdWrapper);
       });
+    } else if (useMarkdown) {
+      // Render markdown first
+      msgDiv.innerHTML = `<strong>${sender}:</strong><div class="markdown-content">${renderMarkdown(String(message))}</div>`;
+      
+      // Add copy buttons to all code blocks
+      addCopyButtonsToCodeBlocks(msgDiv);
     } else {
-      // Regular message handling
-      const codeBlockMatch =
-        typeof message === "string"
-          ? message.match(/```(\w+)?\n([\s\S]*?)```/)
-          : null;
-
-      if (codeBlockMatch) {
-        const lang = codeBlockMatch[1] || "bash";
-        const code = codeBlockMatch[2];
-
-        const codeWrapper = document.createElement("div");
-        codeWrapper.className = "code-wrapper";
-
-        const pre = document.createElement("pre");
-        const codeEl = document.createElement("code");
-        codeEl.className = `language-${lang}`;
-        codeEl.textContent = code;
-        pre.appendChild(codeEl);
-
-        const copyBtn = createCopyButton(code);
-
-        codeWrapper.appendChild(copyBtn);
-        codeWrapper.appendChild(pre);
-
-        msgDiv.innerHTML = `<strong>${sender}:</strong>`;
-        msgDiv.appendChild(codeWrapper);
-      } else if (useMarkdown) {
-        msgDiv.innerHTML = `<strong>${sender}:</strong>${renderMarkdown(String(message))}`;
-
-        msgDiv.querySelectorAll("pre code").forEach((codeEl) => {
-          const code = codeEl.textContent;
-          const pre = codeEl.parentElement;
-          const wrapper = document.createElement("div");
-          wrapper.className = "code-wrapper";
-
-          const copyBtn = createCopyButton(code);
-
-          pre.parentNode.insertBefore(wrapper, pre);
-          wrapper.appendChild(copyBtn);
-          wrapper.appendChild(pre);
-        });
-      } else {
-        msgDiv.innerHTML = `<strong>${sender}:</strong> <pre>${escapeHtml(String(message))}</pre>`;
-      }
+      // Plain text message
+      msgDiv.innerHTML = `<strong>${sender}:</strong> <pre>${escapeHtml(String(message))}</pre>`;
     }
 
     container.appendChild(msgDiv);
     container.scrollTop = container.scrollHeight;
 
+    // Syntax highlighting
     if (window.Prism) {
       try {
-        Prism.highlightAllUnder(container);
+        Prism.highlightAllUnder(msgDiv);
       } catch (_) {}
     }
 
@@ -201,9 +197,9 @@
     const msgDiv = document.createElement("div");
     msgDiv.classList.add("message");
     msgDiv.innerHTML = `<strong>${sender}:</strong> <span class="streaming-content"></span>`;
-    
+
     const contentSpan = msgDiv.querySelector(".streaming-content");
-    
+
     container.appendChild(msgDiv);
     container.scrollTop = container.scrollHeight;
 
@@ -247,7 +243,7 @@
         if (done) break;
 
         buffer += decoder.decode(value, { stream: true });
-        
+
         // SSE format: "data: {json}\n\n"
         const lines = buffer.split("\n\n");
         buffer = lines.pop(); // Keep incomplete data in buffer
@@ -256,10 +252,10 @@
           if (!line.trim() || !line.startsWith("data: ")) continue;
 
           const jsonStr = line.substring(6); // Remove "data: " prefix
-          
+
           try {
             const data = JSON.parse(jsonStr);
-            
+
             if (data.type === "token") {
               fullText += data.text;
               streamUI.updateContent(fullText);

deepshell-backend/deepshell/__main__.py

@@ -10,6 +10,7 @@ import json
 import subprocess
 import shlex
 from typing import Any, List
+from datetime import datetime
 
 from .llm import get_global_client
 
@@ -37,12 +38,56 @@ app.add_middleware(
 
 app.mount("/static", StaticFiles(directory=str(STATIC_ROOT), html=False), name="static")
 
+# ============================================
+# In-Memory Session History
+# ============================================
+
+session_history = []  # Our "in-memory database"
+
+def add_to_history(role: str, content: str):
+    """Add a message to session history."""
+    session_history.append({
+        "role": role,
+        "content": content,
+        "timestamp": datetime.now().isoformat()
+    })
+
+def get_context(max_messages: int = 10) -> str:
+    """Build context from last N messages for LLM."""
+    if not session_history:
+        return ""
+    
+    # Get last N messages
+    recent = session_history[-max_messages:]
+    
+    # Format as conversation
+    context_lines = []
+    for msg in recent:
+        role = msg["role"].capitalize()
+        content = msg["content"]
+        context_lines.append(f"{role}: {content}")
+    
+    return "\n".join(context_lines)
+
+def clear_history():
+    """Clear session history."""
+    global session_history
+    session_history = []
+
+# ============================================
+# Pydantic Models
+# ============================================
+
 class ChatRequest(BaseModel):
     prompt: str
 
 class ExecuteRequest(BaseModel):
     command: str
 
+# ============================================
+# Static File Routes
+# ============================================
+
 @app.get("/")
 def root_page():
     if INDEX_PATH.exists():
@@ -67,6 +112,10 @@ def get_css():
 def favicon():
     return {"ok": True}
 
+# ============================================
+# Chat Endpoints
+# ============================================
+
 @app.get("/chat/ready")
 def chat_ready():
     try:
@@ -75,18 +124,48 @@ def chat_ready():
     except Exception as e:
         return {"status": "error", "detail": str(e)}
 
+@app.get("/chat/history")
+def get_history():
+    """Get current session history."""
+    return {
+        "history": session_history,
+        "count": len(session_history)
+    }
+
+@app.post("/chat/clear")
+def clear_chat():
+    """Clear session history."""
+    clear_history()
+    return {"status": "ok", "message": "History cleared"}
+
 @app.post("/chat/run-agent")
 def run_agent(req: ChatRequest):
     prompt = (req.prompt or "").strip()
     if not prompt:
         return {"error": "prompt is required"}
 
+    # Add user message to history
+    add_to_history("user", prompt)
+
+    # Build context
+    context = get_context(max_messages=10)
+    
+    # Create full prompt with context
+    if context:
+        full_prompt = f"Previous conversation:\n{context}\n\nUser: {prompt}"
+    else:
+        full_prompt = prompt
+
     client = get_global_client(os.getenv("PROVIDER", "groq"))
-    resp = client.chat(prompt, max_tokens=1024, temperature=0.2)
+    resp = client.chat(full_prompt, max_tokens=1024, temperature=0.2)
     try:
         text = resp.choices[0].message.content
     except Exception:
         text = str(resp)
+    
+    # Add assistant response to history
+    add_to_history("assistant", text)
+    
     return {"answer": text}
 
 # ============================================
@@ -111,24 +190,24 @@ DANGEROUS_PATTERNS = [
 def is_command_safe(command: str) -> tuple[bool, str]:
     """Check if command is safe to execute."""
     cmd_lower = command.lower().strip()
-    
+
     # Check for dangerous patterns
     for pattern in DANGEROUS_PATTERNS:
         if pattern.lower() in cmd_lower:
             return False, f"Dangerous pattern detected: {pattern}"
-    
+
     # Extract first word (command name)
     try:
         parts = shlex.split(command)
         if not parts:
             return False, "Empty command"
-        
+
         cmd_name = parts[0].split("/")[-1]  # Handle /usr/bin/docker -> docker
-        
+
         # Check if command is in whitelist
         if cmd_name not in ALLOWED_COMMANDS:
             return False, f"Command '{cmd_name}' is not in the allowed list"
-        
+
         return True, "OK"
     except Exception as e:
         return False, f"Failed to parse command: {str(e)}"
@@ -144,7 +223,7 @@ def execute_command_safe(command: str, timeout: int = 30) -> dict:
             timeout=timeout,
             cwd=os.getcwd(),
         )
-        
+
         return {
             "success": True,
             "stdout": result.stdout,
@@ -181,7 +260,7 @@ def execute_command(req: ExecuteRequest):
             {"success": False, "error": "Command is required"},
             status_code=400
         )
-    
+
     # Safety check
     is_safe, reason = is_command_safe(command)
     if not is_safe:
@@ -189,9 +268,14 @@ def execute_command(req: ExecuteRequest):
             {"success": False, "error": f"Command blocked: {reason}"},
             status_code=403
         )
-    
+
     # Execute
     result = execute_command_safe(command)
+    
+    # Add command execution to history
+    if result["success"]:
+        add_to_history("system", f"Executed: {command}\nOutput: {result['stdout'][:200]}")
+    
     return result
 
 # ============================================
@@ -207,7 +291,7 @@ def _extract_text_from_chunk(chunk: Any) -> str:
                 delta = chunk["choices"][0].get("delta", {})
                 return delta.get("content", "") or ""
             return chunk.get("content", "") or chunk.get("text", "") or ""
-        
+
         # Handle object responses (SDK objects)
         if hasattr(chunk, "choices") and chunk.choices:
             choice = chunk.choices[0]
@@ -215,7 +299,7 @@ def _extract_text_from_chunk(chunk: Any) -> str:
                 return choice.delta.content or ""
             if hasattr(choice, "message") and hasattr(choice.message, "content"):
                 return choice.message.content or ""
-        
+
         return ""
     except Exception:
         return ""
@@ -226,46 +310,67 @@ async def _sse_generator(prompt: str):
     """
     # Send start event
     yield f"data: {json.dumps({'type': 'start'})}\n\n"
+
+    # Add user message to history
+    add_to_history("user", prompt)
+
+    # Build context from history
+    context = get_context(max_messages=10)
     
+    # Create full prompt with context
+    if context:
+        full_prompt = f"Previous conversation:\n{context}\n\nUser: {prompt}"
+    else:
+        full_prompt = prompt
+
     try:
         client = get_global_client(os.getenv("PROVIDER", "groq"))
     except Exception as e:
         yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
         yield f"data: {json.dumps({'type': 'done'})}\n\n"
         return
-    
+
+    full_response = ""
+
     try:
         # Try streaming with stream=True
         try:
-            response = client.chat(prompt, max_tokens=1024, temperature=0.2, stream=True)
-            
+            response = client.chat(full_prompt, max_tokens=1024, temperature=0.2, stream=True)
+
             # Handle streaming response
             if hasattr(response, "__iter__"):
                 for chunk in response:
                     text = _extract_text_from_chunk(chunk)
                     if text:
+                        full_response += text
                         yield f"data: {json.dumps({'type': 'token', 'text': text})}\n\n"
                     await asyncio.sleep(0)  # Yield control to event loop
             else:
                 # Not iterable, treat as single response
                 text = _extract_text_from_chunk(response)
                 if text:
+                    full_response = text
                     yield f"data: {json.dumps({'type': 'token', 'text': text})}\n\n"
-        
+
         except TypeError:
             # stream=True not supported, fallback to non-streaming
-            response = client.chat(prompt, max_tokens=1024, temperature=0.2)
+            response = client.chat(full_prompt, max_tokens=1024, temperature=0.2)
             try:
                 text = response.choices[0].message.content
             except Exception:
                 text = str(response)
-            
+
             if text:
+                full_response = text
                 yield f"data: {json.dumps({'type': 'token', 'text': text})}\n\n"
-    
+
     except Exception as e:
         yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
-    
+
+    # Add assistant response to history
+    if full_response:
+        add_to_history("assistant", full_response)
+
     # Send done event
     yield f"data: {json.dumps({'type': 'done'})}\n\n"
 
@@ -274,7 +379,7 @@ async def run_agent_stream(req: ChatRequest):
     prompt = (req.prompt or "").strip()
     if not prompt:
         return JSONResponse({"error": "prompt is required"}, status_code=400)
-    
+
     return StreamingResponse(
         _sse_generator(prompt),
         media_type="text/event-stream",

deepshell-backend/deepshell/llm.py

@@ -6,24 +6,26 @@ from groq import Groq
 
 _singleton = {"client": None, "provider": None}
 
-DEEPSHELL_SYSTEM_PROMPT = """You are DeepShell, an AI assistant that helps users with Docker, Kubernetes, and system administration tasks.
+DEEPSHELL_SYSTEM_PROMPT = """You are DeepShell, an AI assistant specialized in Docker, Kubernetes, and system administration tasks.
 
-When you recommend a command for the user to execute, wrap it in XML tags like this:
-<execute>docker ps</execute>
+Your role:
+- Provide clear, accurate commands and scripts
+- Explain what each command does
+- Format code in markdown code blocks with proper language tags
+- Be concise but thorough
+- Prioritize security and best practices
 
-Rules:
-1. Only suggest commands that are safe and relevant
-2. Explain what the command does before or after the execute tag
-3. One command per execute tag
-4. Use full command syntax (no placeholders like <container-id>)
-5. For multi-step tasks, provide commands one at a time
+When providing commands or scripts:
+- Use proper markdown code blocks: ```bash or ```yaml or ```python
+- Include comments explaining complex parts
+- Provide full working examples (no placeholders)
+- For multi-step tasks, show all steps clearly
 
 Example responses:
-- "To see running containers, use: <execute>docker ps</execute>"
-- "Let me check the disk usage: <execute>df -h</execute>"
-- "Here's how to list all images: <execute>docker images</execute>"
+- "To see running containers, use:\n```bash\ndocker ps\n```"
+- "Here's an Ansible playbook for Jenkins:\n```yaml\n---\n- hosts: localhost\n  tasks:\n    - name: Install Java\n      apt:\n        name: openjdk-11-jdk\n        state: present\n```"
 
-Be helpful, concise, and always prioritize user safety."""
+Be helpful, professional, and always prioritize user safety."""
 
 def get_global_client(provider: str = "groq") -> Any:
     provider = (provider or "groq").lower()
@@ -41,26 +43,26 @@ def get_global_client(provider: str = "groq") -> Any:
 
     class Wrapper:
         def chat(
-            self, 
-            prompt: str, 
-            max_tokens: int = 1024, 
+            self,
+            prompt: str,
+            max_tokens: int = 1024,
             temperature: float = 0.2,
             stream: bool = False,
             system_prompt: Optional[str] = None
         ):
             model = os.getenv("GROQ_MODEL", "llama-3.3-70b-versatile")
-            
+
             messages = []
-            
+
             # Add system prompt if provided, otherwise use default
             if system_prompt is not None:
                 if system_prompt:  # Only add if not empty string
                     messages.append({"role": "system", "content": system_prompt})
             else:
                 messages.append({"role": "system", "content": DEEPSHELL_SYSTEM_PROMPT})
-            
+
             messages.append({"role": "user", "content": prompt})
-            
+
             response = client.chat.completions.create(
                 model=model,
                 messages=messages,