Spaces:
Running
on
Zero
Running
on
Zero
File size: 20,304 Bytes
886d8e9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 |
"""
The terminal interface is just a view. Just handles the very top layer.
If you were to build a frontend this would be a way to do it.
"""
try:
import readline
except ImportError:
pass
import os
import platform
import random
import re
import subprocess
import time
from ..core.utils.scan_code import scan_code
from ..core.utils.system_debug_info import system_info
from ..core.utils.truncate_output import truncate_output
from .components.code_block import CodeBlock
from .components.message_block import MessageBlock
from .magic_commands import handle_magic_command
from .utils.check_for_package import check_for_package
from .utils.display_markdown_message import display_markdown_message
from .utils.display_output import display_output
from .utils.find_image_path import find_image_path
from .utils.cli_input import cli_input
# Add examples to the readline history
examples = [
"How many files are on my desktop?",
"What time is it in Seattle?",
"Make me a simple Pomodoro app.",
"Open Chrome and go to YouTube.",
"Can you set my system to light mode?",
]
random.shuffle(examples)
try:
for example in examples:
readline.add_history(example)
except:
# If they don't have readline, that's fine
pass
def terminal_interface(interpreter, message):
# Auto run and offline (this.. this isn't right) don't display messages.
# Probably worth abstracting this to something like "debug_cli" at some point.
if not interpreter.auto_run and not interpreter.offline:
interpreter_intro_message = [
"**Open Interpreter** will require approval before running code."
]
if interpreter.safe_mode == "ask" or interpreter.safe_mode == "auto":
if not check_for_package("semgrep"):
interpreter_intro_message.append(
f"**Safe Mode**: {interpreter.safe_mode}\n\n>Note: **Safe Mode** requires `semgrep` (`pip install semgrep`)"
)
else:
interpreter_intro_message.append("Use `interpreter -y` to bypass this.")
interpreter_intro_message.append("Press `CTRL-C` to exit.")
display_markdown_message("\n\n".join(interpreter_intro_message) + "\n")
if message:
interactive = False
else:
interactive = True
active_block = None
voice_subprocess = None
while True:
if interactive:
### This is the primary input for Open Interpreter.
message = cli_input("> ").strip() if interpreter.multi_line else input("> ").strip()
try:
# This lets users hit the up arrow key for past messages
readline.add_history(message)
except:
# If the user doesn't have readline (may be the case on windows), that's fine
pass
if isinstance(message, str):
# This is for the terminal interface being used as a CLI — messages are strings.
# This won't fire if they're in the python package, display=True, and they passed in an array of messages (for example).
if message == "":
# Ignore empty messages when user presses enter without typing anything
continue
if message.startswith("%") and interactive:
handle_magic_command(interpreter, message)
continue
# Many users do this
if message.strip() == "interpreter --local":
print("Please exit this conversation, then run `interpreter --local`.")
continue
if message.strip() == "pip install --upgrade open-interpreter":
print(
"Please exit this conversation, then run `pip install --upgrade open-interpreter`."
)
continue
if interpreter.llm.supports_vision or interpreter.llm.vision_renderer != None:
# Is the input a path to an image? Like they just dragged it into the terminal?
image_path = find_image_path(message)
## If we found an image, add it to the message
if image_path:
# Add the text interpreter's message history
interpreter.messages.append(
{
"role": "user",
"type": "message",
"content": message,
}
)
# Pass in the image to interpreter in a moment
message = {
"role": "user",
"type": "image",
"format": "path",
"content": image_path,
}
try:
for chunk in interpreter.chat(message, display=False, stream=True):
yield chunk
# Is this for thine eyes?
if "recipient" in chunk and chunk["recipient"] != "user":
continue
if interpreter.verbose:
print("Chunk in `terminal_interface`:", chunk)
# Comply with PyAutoGUI fail-safe for OS mode
# so people can turn it off by moving their mouse to a corner
if interpreter.os:
if (
chunk.get("format") == "output"
and "failsafeexception" in chunk["content"].lower()
):
print("Fail-safe triggered (mouse in one of the four corners).")
break
if "end" in chunk and active_block:
active_block.refresh(cursor=False)
if chunk["type"] in [
"message",
"console",
]: # We don't stop on code's end — code + console output are actually one block.
active_block.end()
active_block = None
# Assistant message blocks
if chunk["type"] == "message":
if "start" in chunk:
active_block = MessageBlock()
render_cursor = True
if "content" in chunk:
active_block.message += chunk["content"]
if "end" in chunk and interpreter.os:
last_message = interpreter.messages[-1]["content"]
# Remove markdown lists and the line above markdown lists
lines = last_message.split("\n")
i = 0
while i < len(lines):
# Match markdown lists starting with hyphen, asterisk or number
if re.match(r"^\s*([-*]|\d+\.)\s", lines[i]):
del lines[i]
if i > 0:
del lines[i - 1]
i -= 1
else:
i += 1
message = "\n".join(lines)
# Replace newlines with spaces, escape double quotes and backslashes
sanitized_message = (
message.replace("\\", "\\\\")
.replace("\n", " ")
.replace('"', '\\"')
)
# Display notification in OS mode
if interpreter.os:
interpreter.computer.os.notify(sanitized_message)
# Speak message aloud
if platform.system() == "Darwin" and interpreter.speak_messages:
if voice_subprocess:
voice_subprocess.terminate()
voice_subprocess = subprocess.Popen(
[
"osascript",
"-e",
f'say "{sanitized_message}" using "Fred"',
]
)
else:
pass
# User isn't on a Mac, so we can't do this. You should tell them something about that when they first set this up.
# Or use a universal TTS library.
# Assistant code blocks
elif chunk["role"] == "assistant" and chunk["type"] == "code":
if "start" in chunk:
active_block = CodeBlock()
active_block.language = chunk["format"]
render_cursor = True
if "content" in chunk:
active_block.code += chunk["content"]
# Execution notice
if chunk["type"] == "confirmation":
if not interpreter.auto_run:
# OI is about to execute code. The user wants to approve this
# End the active code block so you can run input() below it
if active_block:
active_block.refresh(cursor=False)
active_block.end()
active_block = None
code_to_run = chunk["content"]
language = code_to_run["format"]
code = code_to_run["content"]
should_scan_code = False
if not interpreter.safe_mode == "off":
if interpreter.safe_mode == "auto":
should_scan_code = True
elif interpreter.safe_mode == "ask":
response = input(
" Would you like to scan this code? (y/n)\n\n "
)
print("") # <- Aesthetic choice
if response.strip().lower() == "y":
should_scan_code = True
if should_scan_code:
scan_code(code, language, interpreter)
response = input(
" Would you like to run this code? (y/n)\n\n "
)
print("") # <- Aesthetic choice
if response.strip().lower() == "y":
# Create a new, identical block where the code will actually be run
# Conveniently, the chunk includes everything we need to do this:
active_block = CodeBlock()
active_block.margin_top = False # <- Aesthetic choice
active_block.language = language
active_block.code = code
else:
# User declined to run code.
interpreter.messages.append(
{
"role": "user",
"type": "message",
"content": "I have declined to run this code.",
}
)
break
# Computer can display visual types to user,
# Which sometimes creates more computer output (e.g. HTML errors, eventually)
if (
chunk["role"] == "computer"
and "content" in chunk
and (
chunk["type"] == "image"
or ("format" in chunk and chunk["format"] == "html")
or ("format" in chunk and chunk["format"] == "javascript")
)
):
if interpreter.os and interpreter.verbose == False:
# We don't display things to the user in OS control mode, since we use vision to communicate the screen to the LLM so much.
# But if verbose is true, we do display it!
continue
# Display and give extra output back to the LLM
extra_computer_output = display_output(chunk)
# We're going to just add it to the messages directly, not changing `recipient` here.
# Mind you, the way we're doing this, this would make it appear to the user if they look at their conversation history,
# because we're not adding "recipient: assistant" to this block. But this is a good simple solution IMO.
# we just might want to change it in the future, once we're sure that a bunch of adjacent type:console blocks will be rendered normally to text-only LLMs
# and that if we made a new block here with "recipient: assistant" it wouldn't add new console outputs to that block (thus hiding them from the user)
if (
interpreter.messages[-1].get("format") != "output"
or interpreter.messages[-1]["role"] != "computer"
or interpreter.messages[-1]["type"] != "console"
):
# If the last message isn't a console output, make a new block
interpreter.messages.append(
{
"role": "computer",
"type": "console",
"format": "output",
"content": extra_computer_output,
}
)
else:
# If the last message is a console output, simply append the extra output to it
interpreter.messages[-1]["content"] += (
"\n" + extra_computer_output
)
interpreter.messages[-1]["content"] = interpreter.messages[-1][
"content"
].strip()
# Console
if chunk["type"] == "console":
render_cursor = False
if "format" in chunk and chunk["format"] == "output":
active_block.output += "\n" + chunk["content"]
active_block.output = (
active_block.output.strip()
) # ^ Aesthetic choice
# Truncate output
active_block.output = truncate_output(
active_block.output, interpreter.max_output
)
if "format" in chunk and chunk["format"] == "active_line":
active_block.active_line = chunk["content"]
# Display action notifications if we're in OS mode
if interpreter.os and active_block.active_line != None:
action = ""
code_lines = active_block.code.split("\n")
if active_block.active_line < len(code_lines):
action = code_lines[active_block.active_line].strip()
if action.startswith("computer"):
description = None
# Extract arguments from the action
start_index = action.find("(")
end_index = action.rfind(")")
if start_index != -1 and end_index != -1:
# (If we found both)
arguments = action[start_index + 1 : end_index]
else:
arguments = None
# NOTE: Do not put the text you're clicking on screen
# (unless we figure out how to do this AFTER taking the screenshot)
# otherwise it will try to click this notification!
if any(action.startswith(text) for text in [
"computer.screenshot",
"computer.display.screenshot",
"computer.display.view",
"computer.view"
]):
description = "Viewing screen..."
elif action == "computer.mouse.click()":
description = "Clicking..."
elif action.startswith("computer.mouse.click("):
if "icon=" in arguments:
text_or_icon = "icon"
else:
text_or_icon = "text"
description = f"Clicking {text_or_icon}..."
elif action.startswith("computer.mouse.move("):
if "icon=" in arguments:
text_or_icon = "icon"
else:
text_or_icon = "text"
if (
"click" in active_block.code
): # This could be better
description = f"Clicking {text_or_icon}..."
else:
description = f"Mousing over {text_or_icon}..."
elif action.startswith("computer.keyboard.write("):
description = f"Typing {arguments}."
elif action.startswith("computer.keyboard.hotkey("):
description = f"Pressing {arguments}."
elif action.startswith("computer.keyboard.press("):
description = f"Pressing {arguments}."
elif action == "computer.os.get_selected_text()":
description = f"Getting selected text."
if description:
interpreter.computer.os.notify(description)
if "start" in chunk:
# We need to make a code block if we pushed out an HTML block first, which would have closed our code block.
if not isinstance(active_block, CodeBlock):
if active_block:
active_block.end()
active_block = CodeBlock()
if active_block:
active_block.refresh(cursor=render_cursor)
# (Sometimes -- like if they CTRL-C quickly -- active_block is still None here)
if "active_block" in locals():
if active_block:
active_block.end()
active_block = None
time.sleep(0.1)
if not interactive:
# Don't loop
break
except KeyboardInterrupt:
# Exit gracefully
if "active_block" in locals() and active_block:
active_block.end()
active_block = None
if interactive:
# (this cancels LLM, returns to the interactive "> " input)
continue
else:
break
except:
if interpreter.debug:
system_info(interpreter)
raise
|