octomind 0.20.0

// Copyright 2025 Muvon Un Limited
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// Shell execution functionality for the Developer MCP provider

use super::super::{get_thread_working_directory, McpFunction, McpToolCall, McpToolResult};
use anyhow::{anyhow, Result};
use serde_json::{json, Value};
// Define the shell function for the MCP protocol with enhanced description
pub fn get_shell_function() -> McpFunction {
	McpFunction {
		name: "shell".to_string(),
		description: "Execute a command in the shell.

This will return the output and error concatenated into a single string, as
you would see from running on the command line. There will also be an indication
of if the command succeeded or failed.

Parameters:
- `command`: The shell command to execute (required)
- `background`: Run command in background and return PID instead of waiting for completion (default: false)

**Important**: Each shell command runs in its own process in current working directory. Things like directory changes or
sourcing files do not persist between tool calls. So you may need to repeat them each time by
stringing together commands, e.g. `cd example && ls` or `source env/bin/activate && pip install numpy`

**Important**: Use ripgrep - `rg` - when you need to locate a file or a code reference, other solutions
may show ignored or hidden files. For example *do not* use `find` or `ls -r`
- List files by name: `rg --files | rg <filename>`
- List files that contain a regex: `rg '<regex>' -l`

**Output Truncation:**
Output size is controlled by global mcp_response_tokens_threshold setting.
Avoid commands that produce large outputs (like `cat large_file` or `find /` without filters).
Use more specific commands to reduce output size if responses are truncated.

**Background Execution:**
When `background` is true, the command runs in the background and returns immediately with the process PID.
Background processes continue running until explicitly killed or the main application exits.
Use the returned PID with `kill <pid>` command to terminate background processes.
NO need to append `&` to your command when using `background: true` - this is handled automatically.

Examples:
- Foreground: `{\"command\": \"ls -la\"}`
- Background: `{\"command\": \"python -m http.server 8000\", \"background\": true}`
- Kill background: `{\"command\": \"kill 12345\"}` (where 12345 is the returned PID)
".to_string(),
		parameters: json!({
			"type": "object",
			"properties": {
				"command": {
					"type": "string",
					"description": "The shell command to execute (runs from current working directory)"
				},
				"background": {
					"type": "boolean",
					"default": false,
					"description": "Run command in background and return PID instead of waiting for completion (no need to append '&')"
				}
			},
			"required": ["command"]
		}),
	}
}

// Each entry: (triggering programs, required tool name, hint message).
// The hint is only shown when the recommended tool is actually enabled.
static SHELL_MISUSE_HINTS: &[(&[&str], &str, &str)] = &[
	(
		&["cat", "head", "tail", "less", "more"],
		"view",
		"⚠️ Prefer `view` for reading files (line-numbered, supports ranges). Use shell only when piping output.",
	),
	(
		&["grep", "egrep", "fgrep", "rg"],
		"ast_grep",
		"⚠️ Prefer `ast_grep` for code search or `view` with content= for text search (.gitignore-aware). Use shell grep only for unsupported raw flags.",
	),
	(
		&["find", "ls"],
		"view",
		"⚠️ Prefer `view` for directory listing (.gitignore-aware, pattern/content filtering). Use shell only for system paths outside the project.",
	),
	(
		&["sed", "awk"],
		"text_editor",
		"⚠️ Prefer `text_editor` str_replace/line_replace for file edits (atomic, tracked). Use sed/awk only for stream transforms in pipelines.",
	),
];

// Detect shell commands that should use a dedicated MCP tool instead.
// Returns a hint only when the recommended tool is actually enabled in the current session.
fn detect_shell_misuse(command: &str) -> Option<&'static str> {
	let cmd = command.trim();

	// Check if cmd is exactly `prog` or starts with `prog ` / `prog\t`
	let is_prog = |prog: &str| -> bool {
		cmd == prog || cmd.starts_with(&format!("{prog} ")) || cmd.starts_with(&format!("{prog}\t"))
	};

	for (progs, tool, hint) in SHELL_MISUSE_HINTS {
		if progs.iter().any(|p| is_prog(p)) {
			// Only warn if the recommended tool is actually available
			if crate::mcp::tool_map::get_server_for_tool(tool).is_some() {
				return Some(hint);
			}
		}
	}

	None
}

// Execute a shell command
pub async fn execute_shell_command(call: &McpToolCall) -> Result<McpToolResult> {
	use tokio::process::Command as TokioCommand;

	// Extract command parameter
	let command = match call.parameters.get("command") {
		Some(Value::String(cmd)) => {
			if cmd.trim().is_empty() {
				return Ok(McpToolResult::error(
					call.tool_name.clone(),
					call.tool_id.clone(),
					"Command parameter cannot be empty".to_string(),
				));
			}
			cmd.clone()
		}
		Some(_) => {
			return Ok(McpToolResult::error(
				call.tool_name.clone(),
				call.tool_id.clone(),
				"Command parameter must be a string".to_string(),
			));
		}
		None => {
			return Ok(McpToolResult::error(
				call.tool_name.clone(),
				call.tool_id.clone(),
				"Missing required 'command' parameter".to_string(),
			));
		}
	};

	// Extract background parameter
	let background = call
		.parameters
		.get("background")
		.and_then(|v| v.as_bool())
		.unwrap_or(false);

	// NOTE: We do NOT add MCP tool commands to shell history
	// NOTE: We do NOT add MCP tool commands to shell history
	// Only direct user commands via `octomind shell` CLI should persist to history
	// (see src/commands/shell.rs for user-initiated shell history)

	// Get the working directory from thread-local storage
	let working_dir = get_thread_working_directory();

	// Use tokio::process::Command for better cancellation support
	let mut cmd = if cfg!(target_os = "windows") {
		let mut cmd = TokioCommand::new("cmd");
		cmd.args(["/C", &command]);
		cmd.current_dir(&working_dir);
		cmd
	} else {
		let mut cmd = TokioCommand::new("sh");
		cmd.args(["-c", &command]);
		cmd.current_dir(&working_dir);
		cmd
	};

	// Configure the command based on execution mode
	if background {
		// Background execution: detach process and return PID immediately
		cmd.stdout(std::process::Stdio::null())
			.stderr(std::process::Stdio::null())
			.stdin(std::process::Stdio::null())
			.kill_on_drop(false); // Don't kill when dropped - let it run independently
	} else {
		// Foreground execution: capture output and wait for completion
		cmd.stdout(std::process::Stdio::piped())
			.stderr(std::process::Stdio::piped())
			.stdin(std::process::Stdio::null())
			.kill_on_drop(true); // CRITICAL: Kill process when dropped
	}

	// Spawn the process
	let child = cmd
		.spawn()
		.map_err(|e| anyhow!("Failed to spawn command: {}", e))?;

	// Handle background vs foreground execution
	if background {
		// Background execution: return PID immediately
		let pid = child
			.id()
			.ok_or_else(|| anyhow!("Failed to get process ID"))?;

		// Detach the child process so it continues running independently
		// We do this by forgetting the child handle, which prevents kill_on_drop
		std::mem::forget(child);

		return Ok(McpToolResult {
			tool_name: "shell".to_string(),
			tool_id: call.tool_id.clone(),
			result: json!({
				"success": true,
				"background": true,
				"pid": pid,
				"command": command,
				"message": format!("Command started in background with PID {pid}"),
				"note": format!("Use 'kill {pid}' to terminate this background process if needed")
			}),
		});
	}

	// Foreground execution: wait for completion and return output
	let result = child.wait_with_output().await;
	match result.map_err(|e| anyhow!("Command execution failed: {}", e)) {
		Ok(output) => {
			let stdout = String::from_utf8_lossy(&output.stdout).to_string();
			let stderr = String::from_utf8_lossy(&output.stderr).to_string();

			// Format the output more clearly with error handling
			let combined = if stderr.is_empty() {
				stdout
			} else if stdout.is_empty() {
				stderr
			} else {
				format!("{stdout}\n\nError: {stderr}")
			};

			// Apply global truncation (handled by global MCP response truncation)
			let final_output = combined;

			// Add detailed execution results including status code
			let status_code = output.status.code().unwrap_or(-1);
			let success = output.status.success();

			// Push misuse hint into accumulator — injected as a user message after all tools finish
			if let Some(hint) = detect_shell_misuse(&command) {
				crate::mcp::hint_accumulator::push_hint(hint);
			}

			// MCP Protocol Compliance: Use error() for failed commands, success() for successful ones
			if success {
				// Command succeeded - use success format
				Ok(McpToolResult::success(
					"shell".to_string(),
					call.tool_id.clone(),
					final_output,
				))
			} else {
				// Command failed - use error format per MCP protocol
				Ok(McpToolResult::error(
					"shell".to_string(),
					call.tool_id.clone(),
					format!(
						"Command failed with exit code {status_code}\nCommand: {command}\n\nOutput:\n{final_output}"
					),
				))
			}
		}
		Err(e) => Ok(McpToolResult::error(
			"shell".to_string(),
			call.tool_id.clone(),
			format!("Error: {e}"),
		)),
	}
}