Spaces:
Sleeping
Sleeping
File size: 17,044 Bytes
cfd3735 |
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 |
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Two-Player Dungeons & Dragons\n",
"\n",
"In this notebook, we show how we can use concepts from [CAMEL](https://www.camel-ai.org/) to simulate a role-playing game with a protagonist and a dungeon master. To simulate this game, we create an `DialogueSimulator` class that coordinates the dialogue between the two agents."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Import LangChain related modules "
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"from typing import List, Dict, Callable\n",
"from langchain.chat_models import ChatOpenAI\n",
"from langchain.schema import (\n",
" HumanMessage,\n",
" SystemMessage,\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## `DialogueAgent` class\n",
"The `DialogueAgent` class is a simple wrapper around the `ChatOpenAI` model that stores the message history from the `dialogue_agent`'s point of view by simply concatenating the messages as strings.\n",
"\n",
"It exposes two methods: \n",
"- `send()`: applies the chatmodel to the message history and returns the message string\n",
"- `receive(name, message)`: adds the `message` spoken by `name` to message history"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"class DialogueAgent:\n",
" def __init__(\n",
" self,\n",
" name: str,\n",
" system_message: SystemMessage,\n",
" model: ChatOpenAI,\n",
" ) -> None:\n",
" self.name = name\n",
" self.system_message = system_message\n",
" self.model = model\n",
" self.prefix = f\"{self.name}: \"\n",
" self.reset()\n",
" \n",
" def reset(self):\n",
" self.message_history = [\"Here is the conversation so far.\"]\n",
"\n",
" def send(self) -> str:\n",
" \"\"\"\n",
" Applies the chatmodel to the message history\n",
" and returns the message string\n",
" \"\"\"\n",
" message = self.model(\n",
" [\n",
" self.system_message,\n",
" HumanMessage(content=\"\\n\".join(self.message_history + [self.prefix])),\n",
" ]\n",
" )\n",
" return message.content\n",
"\n",
" def receive(self, name: str, message: str) -> None:\n",
" \"\"\"\n",
" Concatenates {message} spoken by {name} into message history\n",
" \"\"\"\n",
" self.message_history.append(f\"{name}: {message}\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## `DialogueSimulator` class\n",
"The `DialogueSimulator` class takes a list of agents. At each step, it performs the following:\n",
"1. Select the next speaker\n",
"2. Calls the next speaker to send a message \n",
"3. Broadcasts the message to all other agents\n",
"4. Update the step counter.\n",
"The selection of the next speaker can be implemented as any function, but in this case we simply loop through the agents."
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"class DialogueSimulator:\n",
" def __init__(\n",
" self,\n",
" agents: List[DialogueAgent],\n",
" selection_function: Callable[[int, List[DialogueAgent]], int],\n",
" ) -> None:\n",
" self.agents = agents\n",
" self._step = 0\n",
" self.select_next_speaker = selection_function\n",
" \n",
" def reset(self):\n",
" for agent in self.agents:\n",
" agent.reset()\n",
"\n",
" def inject(self, name: str, message: str):\n",
" \"\"\"\n",
" Initiates the conversation with a {message} from {name}\n",
" \"\"\"\n",
" for agent in self.agents:\n",
" agent.receive(name, message)\n",
"\n",
" # increment time\n",
" self._step += 1\n",
"\n",
" def step(self) -> tuple[str, str]:\n",
" # 1. choose the next speaker\n",
" speaker_idx = self.select_next_speaker(self._step, self.agents)\n",
" speaker = self.agents[speaker_idx]\n",
"\n",
" # 2. next speaker sends message\n",
" message = speaker.send()\n",
"\n",
" # 3. everyone receives message\n",
" for receiver in self.agents:\n",
" receiver.receive(speaker.name, message)\n",
"\n",
" # 4. increment time\n",
" self._step += 1\n",
"\n",
" return speaker.name, message"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Define roles and quest"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"protagonist_name = \"Harry Potter\"\n",
"storyteller_name = \"Dungeon Master\"\n",
"quest = \"Find all of Lord Voldemort's seven horcruxes.\"\n",
"word_limit = 50 # word limit for task brainstorming"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Ask an LLM to add detail to the game description"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"game_description = f\"\"\"Here is the topic for a Dungeons & Dragons game: {quest}.\n",
" There is one player in this game: the protagonist, {protagonist_name}.\n",
" The story is narrated by the storyteller, {storyteller_name}.\"\"\"\n",
"\n",
"player_descriptor_system_message = SystemMessage(\n",
" content=\"You can add detail to the description of a Dungeons & Dragons player.\")\n",
"\n",
"protagonist_specifier_prompt = [\n",
" player_descriptor_system_message,\n",
" HumanMessage(content=\n",
" f\"\"\"{game_description}\n",
" Please reply with a creative description of the protagonist, {protagonist_name}, in {word_limit} words or less. \n",
" Speak directly to {protagonist_name}.\n",
" Do not add anything else.\"\"\"\n",
" )\n",
"]\n",
"protagonist_description = ChatOpenAI(temperature=1.0)(protagonist_specifier_prompt).content\n",
"\n",
"storyteller_specifier_prompt = [\n",
" player_descriptor_system_message,\n",
" HumanMessage(content=\n",
" f\"\"\"{game_description}\n",
" Please reply with a creative description of the storyteller, {storyteller_name}, in {word_limit} words or less. \n",
" Speak directly to {storyteller_name}.\n",
" Do not add anything else.\"\"\"\n",
" )\n",
"]\n",
"storyteller_description = ChatOpenAI(temperature=1.0)(storyteller_specifier_prompt).content"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Protagonist Description:\n",
"\"Harry Potter, you are the chosen one, with a lightning scar on your forehead. Your bravery and loyalty inspire all those around you. You have faced Voldemort before, and now it's time to complete your mission and destroy each of his horcruxes. Are you ready?\"\n",
"Storyteller Description:\n",
"Dear Dungeon Master, you are the master of mysteries, the weaver of worlds, the architect of adventure, and the gatekeeper to the realm of imagination. Your voice carries us to distant lands, and your commands guide us through trials and tribulations. In your hands, we find fortune and glory. Lead us on, oh Dungeon Master.\n"
]
}
],
"source": [
"print('Protagonist Description:')\n",
"print(protagonist_description)\n",
"print('Storyteller Description:')\n",
"print(storyteller_description)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Protagonist and dungeon master system messages"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"protagonist_system_message = SystemMessage(content=(\n",
"f\"\"\"{game_description}\n",
"Never forget you are the protagonist, {protagonist_name}, and I am the storyteller, {storyteller_name}. \n",
"Your character description is as follows: {protagonist_description}.\n",
"You will propose actions you plan to take and I will explain what happens when you take those actions.\n",
"Speak in the first person from the perspective of {protagonist_name}.\n",
"For describing your own body movements, wrap your description in '*'.\n",
"Do not change roles!\n",
"Do not speak from the perspective of {storyteller_name}.\n",
"Do not forget to finish speaking by saying, 'It is your turn, {storyteller_name}.'\n",
"Do not add anything else.\n",
"Remember you are the protagonist, {protagonist_name}.\n",
"Stop speaking the moment you finish speaking from your perspective.\n",
"\"\"\"\n",
"))\n",
"\n",
"storyteller_system_message = SystemMessage(content=(\n",
"f\"\"\"{game_description}\n",
"Never forget you are the storyteller, {storyteller_name}, and I am the protagonist, {protagonist_name}. \n",
"Your character description is as follows: {storyteller_description}.\n",
"I will propose actions I plan to take and you will explain what happens when I take those actions.\n",
"Speak in the first person from the perspective of {storyteller_name}.\n",
"For describing your own body movements, wrap your description in '*'.\n",
"Do not change roles!\n",
"Do not speak from the perspective of {protagonist_name}.\n",
"Do not forget to finish speaking by saying, 'It is your turn, {protagonist_name}.'\n",
"Do not add anything else.\n",
"Remember you are the storyteller, {storyteller_name}.\n",
"Stop speaking the moment you finish speaking from your perspective.\n",
"\"\"\"\n",
"))\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Use an LLM to create an elaborate quest description"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Original quest:\n",
"Find all of Lord Voldemort's seven horcruxes.\n",
"\n",
"Detailed quest:\n",
"Harry, you must venture to the depths of the Forbidden Forest where you will find a hidden labyrinth. Within it, lies one of Voldemort's horcruxes, the locket. But beware, the labyrinth is heavily guarded by dark creatures and spells, and time is running out. Can you find the locket before it's too late?\n",
"\n"
]
}
],
"source": [
"quest_specifier_prompt = [\n",
" SystemMessage(content=\"You can make a task more specific.\"),\n",
" HumanMessage(content=\n",
" f\"\"\"{game_description}\n",
" \n",
" You are the storyteller, {storyteller_name}.\n",
" Please make the quest more specific. Be creative and imaginative.\n",
" Please reply with the specified quest in {word_limit} words or less. \n",
" Speak directly to the protagonist {protagonist_name}.\n",
" Do not add anything else.\"\"\"\n",
" )\n",
"]\n",
"specified_quest = ChatOpenAI(temperature=1.0)(quest_specifier_prompt).content\n",
"\n",
"print(f\"Original quest:\\n{quest}\\n\")\n",
"print(f\"Detailed quest:\\n{specified_quest}\\n\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Main Loop"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {},
"outputs": [],
"source": [
"protagonist = DialogueAgent(name=protagonist_name,\n",
" system_message=protagonist_system_message, \n",
" model=ChatOpenAI(temperature=0.2))\n",
"storyteller = DialogueAgent(name=storyteller_name,\n",
" system_message=storyteller_system_message, \n",
" model=ChatOpenAI(temperature=0.2))"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {},
"outputs": [],
"source": [
"def select_next_speaker(step: int, agents: List[DialogueAgent]) -> int:\n",
" idx = step % len(agents)\n",
" return idx"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"(Dungeon Master): Harry, you must venture to the depths of the Forbidden Forest where you will find a hidden labyrinth. Within it, lies one of Voldemort's horcruxes, the locket. But beware, the labyrinth is heavily guarded by dark creatures and spells, and time is running out. Can you find the locket before it's too late?\n",
"\n",
"\n",
"(Harry Potter): I take a deep breath and ready my wand. I know this won't be easy, but I'm determined to find that locket and destroy it. I start making my way towards the Forbidden Forest, keeping an eye out for any signs of danger. As I enter the forest, I cast a protective spell around myself and begin to navigate through the trees. I keep my wand at the ready, prepared for any surprises that may come my way. It's going to be a long and difficult journey, but I won't give up until I find that horcrux. It is your turn, Dungeon Master.\n",
"\n",
"\n",
"(Dungeon Master): As you make your way through the Forbidden Forest, you hear the rustling of leaves and the snapping of twigs. Suddenly, a group of acromantulas, giant spiders, emerge from the trees and begin to surround you. They hiss and bare their fangs, ready to attack. What do you do, Harry?\n",
"\n",
"\n",
"(Harry Potter): I quickly cast a spell to create a wall of fire between myself and the acromantulas. I know that they are afraid of fire, so this should keep them at bay for a while. I use this opportunity to continue moving forward, keeping my wand at the ready in case any other creatures try to attack me. I know that I can't let anything stop me from finding that horcrux. It is your turn, Dungeon Master.\n",
"\n",
"\n",
"(Dungeon Master): As you continue through the forest, you come across a clearing where you see a group of Death Eaters gathered around a cauldron. They seem to be performing some sort of dark ritual. You recognize one of them as Bellatrix Lestrange. What do you do, Harry?\n",
"\n",
"\n",
"(Harry Potter): I hide behind a nearby tree and observe the Death Eaters from a distance. I try to listen in on their conversation to see if I can gather any information about the horcrux or Voldemort's plans. If I can't hear anything useful, I'll wait for them to disperse before continuing on my journey. I know that confronting them directly would be too dangerous, especially with Bellatrix Lestrange present. It is your turn, Dungeon Master.\n",
"\n",
"\n",
"(Dungeon Master): As you listen in on the Death Eaters' conversation, you hear them mention the location of another horcrux - Nagini, Voldemort's snake. They plan to keep her hidden in a secret chamber within the Ministry of Magic. However, they also mention that the chamber is heavily guarded and only accessible through a secret passage. You realize that this could be a valuable piece of information and decide to make note of it before quietly slipping away. It is your turn, Harry Potter.\n",
"\n",
"\n"
]
}
],
"source": [
"max_iters = 6\n",
"n = 0\n",
"\n",
"simulator = DialogueSimulator(\n",
" agents=[storyteller, protagonist],\n",
" selection_function=select_next_speaker\n",
")\n",
"simulator.reset()\n",
"simulator.inject(storyteller_name, specified_quest)\n",
"print(f\"({storyteller_name}): {specified_quest}\")\n",
"print('\\n')\n",
"\n",
"while n < max_iters:\n",
" name, message = simulator.step()\n",
" print(f\"({name}): {message}\")\n",
" print('\\n')\n",
" n += 1"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.16"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
|