Spaces:
Sleeping
Sleeping
File size: 53,308 Bytes
28677d8 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 | [
{
"id": "docs-ollama-models",
"certification": "Docs Study",
"title": "Ollama ๋ชจ๋ธ ์คํ ํ๋ฆ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-1", "docs-lab-2"],
"summary": "Ollama๋ ๋ก์ปฌ์์ LLM ๋ชจ๋ธ์ ๋ด๋ ค๋ฐ๊ณ ์คํํด CLI ๋๋ REST API๋ก ์ฌ์ฉํ ์ ์๊ฒ ํด์ฃผ๋ ๊ฒฝ๋ ๋ชจ๋ธ ์๋ฒ์
๋๋ค. `ollama run`์ผ๋ก ๋ํํ๊ณ , `ollama serve`๋ก API ์๋ฒ๋ฅผ ์ ์งํฉ๋๋ค.",
"example": "# ๋ชจ๋ธ ๋ค์ด๋ก๋ ๋ฐ ์คํ\nollama pull qwen2.5:14b\nollama run qwen2.5:14b\n# ์ค์น๋ ๋ชจ๋ธ ๋ชฉ๋ก\nollama list",
"common_mistake": "๋ชจ๋ธ ์ด๋ฆ์ ํ๊ทธ(:14b, :latest ๋ฑ)๋ฅผ ์๋ตํ๋ฉด `:latest`๋ก ์๋ ์ค์ ๋ฉ๋๋ค. ๋ชจ๋ธ์ด ๋ก์ปฌ์ ์์ ๋ `ollama run`์ด ์๋์ผ๋ก pullํฉ๋๋ค.",
"keywords": ["ollama", "pull", "run", "serve", "model tag", "GGUF", "local LLM"],
"source_id": "official-docs-ollama",
"details": [
"**`ollama run`**: ๋ชจ๋ธ์ด ์์ผ๋ฉด ์๋ ๋ค์ด๋ก๋ ํ ๋ํํ ์ธ์
์์. `--nowordwrap` ์ต์
์ผ๋ก ๊ธด ์ถ๋ ฅ ์ฒ๋ฆฌ ๊ฐ๋ฅ.",
"**`ollama serve`**: 11434 ํฌํธ์์ OpenAI ํธํ REST API ์ ๊ณต. `OLLAMA_HOST=0.0.0.0`์ผ๋ก ์ธ๋ถ ์ ๊ทผ ํ์ฉ ๊ฐ๋ฅ.",
"๋ชจ๋ธ ํ์ผ์ `~/.ollama/models`์ GGUF ํ์์ผ๋ก ์ ์ฅ๋ฉ๋๋ค. `Modelfile`๋ก ์ปค์คํ
์์คํ
ํ๋กฌํํธ์ ํ๋ผ๋ฏธํฐ๋ฅผ ์ค์ ํ ์ ์์ต๋๋ค."
]
},
{
"id": "docs-ollama-api",
"certification": "Docs Study",
"title": "Ollama REST API ํธ์ถ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-3"],
"summary": "Ollama๋ `localhost:11434`์์ REST API๋ฅผ ์ ๊ณตํฉ๋๋ค. `/api/generate`(์์ฑํ), `/api/chat`(๋ํํ), `/api/embeddings`(์๋ฒ ๋ฉ) ์๋ํฌ์ธํธ๋ฅผ OpenAI ํธํ ํ์์ผ๋ก๋ ์ง์ํฉ๋๋ค.",
"example": "# chat ์๋ํฌ์ธํธ\ncurl http://localhost:11434/api/chat -d '{\n \"model\": \"qwen2.5:7b\",\n \"messages\": [{\"role\": \"user\", \"content\": \"์๋
\"}],\n \"stream\": false\n}'",
"common_mistake": "Ollama ์๋ฒ(`ollama serve`)๊ฐ ์คํ ์ค์ด์ง ์๊ฑฐ๋ ๋ชจ๋ธ์ด ๋ก์ปฌ์ ์์ผ๋ฉด API ํธ์ถ์ด ์คํจํฉ๋๋ค. ๋จผ์ `ollama list`๋ก ๋ชจ๋ธ ์กด์ฌ๋ฅผ ํ์ธํ์ธ์.",
"keywords": ["ollama", "REST API", "generate", "chat", "embeddings", "localhost:11434", "stream"],
"source_id": "official-docs-ollama",
"details": [
"**OpenAI ํธํ API**: `http://localhost:11434/v1/chat/completions`๋ก OpenAI Python SDK๋ LangChain์ `ChatOpenAI(base_url=...)` ๊ทธ๋๋ก ์ฌ์ฉ ๊ฐ๋ฅํฉ๋๋ค.",
"**์คํธ๋ฆฌ๋ฐ**: `\"stream\": true` ์ค์ ์ ์๋ต์ JSON ์ค ๋จ์๋ก ์คํธ๋ฆฌ๋ฐ. `\"done\": true` ์ค์ด ์๋ฃ ์ ํธ์
๋๋ค.",
"**์๋ฒ ๋ฉ**: `POST /api/embeddings`์ `{\"model\":\"nomic-embed-text\", \"prompt\":\"ํ
์คํธ\"}`๋ฅผ ๋ณด๋ด๋ฉด ๋ฒกํฐ ๋ฐฐ์ด์ ๋ฐํํฉ๋๋ค."
]
},
{
"id": "docs-dotenv",
"certification": "Docs Study",
"title": "python-dotenv โ ํ๊ฒฝ๋ณ์ & .env ๊ด๋ฆฌ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-4"],
"summary": "python-dotenv๋ `.env` ํ์ผ์ ํค-๊ฐ ์์ `os.environ`์ ์๋์ผ๋ก ๋ก๋ํฉ๋๋ค. API ํค ๊ฐ์ ๋ฏผ๊ฐ ์ ๋ณด๋ฅผ ์ฝ๋์ ์ง์ ์ฐ์ง ์๊ณ ํ์ผ๋ก ๋ถ๋ฆฌํด ๊ด๋ฆฌํ๋ ํ์ค ๋ฐฉ๋ฒ์
๋๋ค.",
"example": "# .env ํ์ผ\nOPENAI_API_KEY=sk-...\nOLLAMA_BASE_URL=http://localhost:11434\n\n# Python ์ฝ๋\nfrom dotenv import load_dotenv\nimport os\n\nload_dotenv() # .env ํ์ผ ๋ก๋\napi_key = os.getenv('OPENAI_API_KEY')",
"common_mistake": "`load_dotenv()`๋ฅผ ํธ์ถํ๊ธฐ ์ ์ `os.getenv()`๋ฅผ ์ฐ๋ฉด ํญ์ `None`์ ๋ฐํํฉ๋๋ค. ๋ฐ๋์ ํ์ผ ์๋จ์์ ๊ฐ์ฅ ๋จผ์ `load_dotenv()`๋ฅผ ํธ์ถํ์ธ์.",
"keywords": ["dotenv", "load_dotenv", ".env", "os.getenv", "ํ๊ฒฝ๋ณ์", "API ํค"],
"source_id": "official-docs-python",
"details": [
"**์ฐ์ ์์**: `load_dotenv(override=False)`(๊ธฐ๋ณธ๊ฐ)๋ ์ด๋ฏธ ์ค์ ๋ ํ๊ฒฝ๋ณ์๋ฅผ ๋ฎ์ด์ฐ์ง ์์ต๋๋ค. CI/CD ํ๊ฒฝ์์ ์์คํ
ํ๊ฒฝ๋ณ์๊ฐ .env๋ณด๋ค ์ฐ์ ํฉ๋๋ค.",
"**.gitignore**: `.env`๋ ๋ฐ๋์ `.gitignore`์ ์ถ๊ฐํด์ผ ํฉ๋๋ค. `.env.example`์ ํค ์ด๋ฆ๋ง ์ ์ด ํ์์๊ฒ ์ด๋ค ๋ณ์๊ฐ ํ์ํ์ง ์๋ ค์ฃผ๋ ์ฉ๋๋ก ์ปค๋ฐํฉ๋๋ค.",
"**Streamlit ์ฐ๋**: ๋ก์ปฌ ๊ฐ๋ฐ ์ `load_dotenv()`๋ก ํค๋ฅผ ๋ก๋ํ๊ณ , ๋ฐฐํฌ ์์๋ `st.secrets`๋ก ์ ๊ทผํ๋๋ก ๋ถ๊ธฐํ๊ฑฐ๋ ์ฒ์๋ถํฐ `st.secrets`๋ง ์ฐ๋ ๋ฐฉ์ ์ค ํ๋๋ฅผ ์ ํํฉ๋๋ค."
]
},
{
"id": "docs-prompt-template",
"certification": "Docs Study",
"title": "ChatPromptTemplate โ ํ๋กฌํํธ ๊ตฌ์กฐํ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-5", "docs-lab-6"],
"summary": "ChatPromptTemplate์ ์์คํ
ํ๋กฌํํธ, ์ฌ์ฉ์ ๋ฉ์์ง, ๋ณ์ ํ๋ ์ด์คํ๋๋ฅผ ๊ตฌ์กฐํ๋ ๋ฐฉ์์ผ๋ก ์ ์ํฉ๋๋ค. ๋จ์ f-string ๋์ ์ฌ์ฌ์ฉ ๊ฐ๋ฅํ๊ณ LangChain ํ์ดํ๋ผ์ธ์ ์ฐ๊ฒฐ๋๋ ํ๋กฌํํธ๋ฅผ ๋ง๋ค ๋ ์๋๋ค.",
"example": "from langchain_core.prompts import ChatPromptTemplate\n\nprompt = ChatPromptTemplate.from_messages([\n (\"system\", \"๋น์ ์ ๋์์ด ๋๋ AI ์ด์์คํดํธ์
๋๋ค. ์ปจํ
์คํธ: {context}\"),\n (\"human\", \"{question}\")\n])\n\n# ์ฒด์ธ์ ์ฐ๊ฒฐ\nchain = prompt | llm | StrOutputParser()\nchain.invoke({\"context\": \"...\", \"question\": \"์์ฝํด์ค\"})",
"common_mistake": "`PromptTemplate`์ ๋จ์ ๋ฌธ์์ด ํฌ๋งท์ฉ(๋ณ์ ํ๋), `ChatPromptTemplate`์ ์ญํ ์ด ๊ตฌ๋ถ๋ ๋ฉ์์ง ๋ฐฐ์ด์ฉ์
๋๋ค. RAG์์ ์์คํ
ํ๋กฌํํธ์ ์ง๋ฌธ์ ๊ตฌ๋ถํ๋ ค๋ฉด ํญ์ ChatPromptTemplate์ ์๋๋ค.",
"keywords": ["ChatPromptTemplate", "from_messages", "system", "human", "placeholder", "PromptTemplate"],
"source_id": "official-docs-langchain",
"details": [
"**MessagesPlaceholder**: ๋ํ ํ์คํ ๋ฆฌ๋ฅผ ํต์งธ๋ก ์ฃผ์
ํ ๋ ์๋๋ค. `(\"placeholder\", \"{chat_history}\")`๋ก ์ ์ธํ๋ฉด ๋ฉ์์ง ๋ฆฌ์คํธ๊ฐ ๊ทธ ์๋ฆฌ์ ํผ์ณ์ง๋๋ค.",
"**partial**: `prompt.partial(context=\"๊ณ ์ ๊ฐ\")`์ผ๋ก ํน์ ๋ณ์๋ฅผ ๋ฏธ๋ฆฌ ์ฑ์ธ ์ ์์ต๋๋ค. context๋ ๊ณ ์ ํ๊ณ question๋ง ๋ฐํ์์ ๋ฐ์ ๋ ์ ์ฉํฉ๋๋ค.",
"**ํ์
์์ ์ฑ**: from_messages์ `(\"system\", \"...\")` ํํ ๋์ `SystemMessage(content=\"...\")`๋ฅผ ์ฌ์ฉํด๋ ๋์ผํฉ๋๋ค. ํํ ๋ฐฉ์์ด ๋ ๊ฐ๊ฒฐํฉ๋๋ค."
]
},
{
"id": "docs-document-loaders",
"certification": "Docs Study",
"title": "LangChain Document Loaders โ ๋ฌธ์ ์ผ๊ด ๋ก๋",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-7"],
"summary": "Document Loader๋ PDF, ๋งํฌ๋ค์ด, ์น ํ์ด์ง, ๋๋ ํ ๋ฆฌ ๋ฑ ๋ค์ํ ์์ค์์ ๋ฌธ์๋ฅผ LangChain `Document` ๊ฐ์ฒด๋ก ๋ณํํฉ๋๋ค. RAG ํ์ดํ๋ผ์ธ์ ์ฒซ ๋จ๊ณ์
๋๋ค.",
"example": "from langchain_community.document_loaders import DirectoryLoader, PyPDFLoader\n\n# ๋๋ ํ ๋ฆฌ ๋ด ๋ชจ๋ .md ํ์ผ ๋ก๋\nloader = DirectoryLoader('./notes', glob='**/*.md')\ndocs = loader.load()\n\n# ๊ฐ ๋ฌธ์์ ์ถ์ฒ ํ์ธ\nfor doc in docs:\n print(doc.metadata['source'], len(doc.page_content))",
"common_mistake": "`loader.load()`๋ ๋ชจ๋ ๋ฌธ์๋ฅผ ๋ฉ๋ชจ๋ฆฌ์ ์ฌ๋ฆฝ๋๋ค. ๋์ฉ๋ ๋๋ ํ ๋ฆฌ์์๋ `loader.lazy_load()`๋ก ์ ๋๋ ์ดํฐ๋ฅผ ์ฐ๊ฑฐ๋ ๋ฐฐ์น๋ก ์ฒ๋ฆฌํด์ผ ํฉ๋๋ค.",
"keywords": ["DirectoryLoader", "PyPDFLoader", "WebBaseLoader", "glob", "metadata", "source", "lazy_load"],
"source_id": "official-docs-langchain",
"details": [
"**metadata['source']**: ๋ก๋๋ ๋ฌธ์๋ `metadata` ๋์
๋๋ฆฌ์ ํ์ผ ๊ฒฝ๋ก(`source`)๋ฅผ ํฌํจํฉ๋๋ค. RAG ๋ต๋ณ์ ์ถ์ฒ๋ฅผ ํ์ํ ๋ ์ด ๊ฐ์ ์๋๋ค.",
"**glob ํจํด**: `glob='**/*.md'`๋ ํ์ ํด๋๊น์ง ์ฌ๊ท ๊ฒ์ํฉ๋๋ค. `glob='*.pdf'`๋ ํ์ฌ ํด๋๋ง ๊ฒ์ํฉ๋๋ค.",
"**WebBaseLoader**: URL ๋ฆฌ์คํธ๋ฅผ ๋๊ธฐ๋ฉด ์น ํ์ด์ง๋ฅผ ํฌ๋กค๋งํด Document๋ก ๋ณํํฉ๋๋ค. BeautifulSoup ๊ธฐ๋ฐ์ด๋ฉฐ `bs_kwargs`๋ก ํ์ฑ ๋ฒ์๋ฅผ ์ง์ ํ ์ ์์ต๋๋ค."
]
},
{
"id": "docs-langchain-rag",
"certification": "Docs Study",
"title": "LangChain RAG ๊ธฐ๋ณธ ํ๋ฆ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-8", "docs-lab-9"],
"summary": "LangChain์ ๋ฌธ์ ๋ก๋ โ ๋ถํ โ ์๋ฒ ๋ฉ โ ๊ฒ์ โ ์์ฑ์ RAG ํ์ดํ๋ผ์ธ์ ์ปดํฌ๋ํธ ์กฐํฉ์ผ๋ก ๊ตฌํํฉ๋๋ค. Retriever๊ฐ ๊ด๋ จ ๋ฌธ์๋ฅผ ์ฐพ์ LLM ์ปจํ
์คํธ๋ก ์ฃผ์
ํฉ๋๋ค.",
"example": "from langchain_community.vectorstores import Chroma\nretriever = Chroma.from_documents(docs, embeddings).as_retriever(k=3)\nchain = (\n {\"context\": retriever, \"question\": RunnablePassthrough()}\n | prompt | llm | StrOutputParser()\n)",
"common_mistake": "Retriever๊ฐ ์ฐพ์ ๋ฌธ์๋ฅผ ๊ทธ๋๋ก ์ ๋ต์ผ๋ก ๋ฏฟ์ง ๋ง์ธ์. ๊ฒ์๋ ์ฒญํฌ๊ฐ ์ง๋ฌธ๊ณผ ๋ค๋ฅธ ๋งฅ๋ฝ์ผ ์ ์์ด ํ๋กฌํํธ์์ ๋ช
์์ ์ผ๋ก '์๋ ๋ฌธ์๋ฅผ ์ฐธ๊ณ ํด'๋ผ๊ณ ์ง์ํด์ผ ํฉ๋๋ค.",
"keywords": ["langchain", "RAG", "retriever", "vectorstore", "document loader", "text splitter"],
"source_id": "official-docs-langchain",
"details": [
"**ํ์ดํ๋ผ์ธ ๋จ๊ณ**: DocumentLoader โ TextSplitter(์ฒญํฌ ๋ถํ ) โ Embeddings โ VectorStore โ Retriever โ LLM Chain.",
"**TextSplitter**: `RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)`์ด ์ผ๋ฐ์ ์์์ . overlap์ด ์์ด์ผ ์ฒญํฌ ๊ฒฝ๊ณ์์ ๋ฌธ๋งฅ ์์ค์ ์ค์
๋๋ค.",
"`MultiQueryRetriever`๋ `ContextualCompressionRetriever`๋ก ๊ฒ์ ํ์ง์ ๋์ผ ์ ์์ต๋๋ค. ๋จ์ ์ ์ฌ๋ ๊ฒ์๋ณด๋ค ๊ด๋ จ ์ฒญํฌ ์ ๋ณ ์ ํ๋๊ฐ ์ฌ๋ผ๊ฐ๋๋ค."
]
},
{
"id": "docs-lcel",
"certification": "Docs Study",
"title": "LangChain LCEL โ ์ฒด์ธ ๊ตฌ์ฑ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-10", "docs-lab-11"],
"summary": "LCEL(LangChain Expression Language)์ `|` ํ์ดํ ์ฐ์ฐ์๋ก Runnable ์ปดํฌ๋ํธ๋ฅผ ์ง๋ ฌยท๋ณ๋ ฌ ์ฐ๊ฒฐํด ์ฒด์ธ์ ์ ์ธ์ ์ผ๋ก ์ ์ํฉ๋๋ค.",
"example": "from langchain_core.runnables import RunnablePassthrough\nchain = (\n RunnablePassthrough.assign(context=retriever)\n | prompt\n | llm\n | StrOutputParser()\n)\nresult = chain.invoke({\"question\": \"What is RAG?\"})",
"common_mistake": "์ฒด์ธ ๊ฐ ๋จ๊ณ์ ์
์ถ๋ ฅ ํ์
์ด ๋ง์์ผ ํฉ๋๋ค. LLM์ `str` ๋๋ `BaseMessage`๋ฅผ ๋ฐ๊ณ ๋ฐํํ๋ฏ๋ก, ์ ๋จ๊ณ๊ฐ ๋์
๋๋ฆฌ๋ฅผ ๋ฐํํ๋ฉด ํ๋กฌํํธ ํ
ํ๋ฆฟ์ด ์ค๊ฐ์์ ๋ฐ์ ์ฒ๋ฆฌํด์ผ ํฉ๋๋ค.",
"keywords": ["LCEL", "Runnable", "pipe", "chain", "invoke", "stream", "batch"],
"source_id": "official-docs-langchain",
"details": [
"**3๊ฐ์ง ์คํ ๋ฐฉ์**: `invoke`(๋จ์ผ ๋๊ธฐ), `stream`(์คํธ๋ฆฌ๋ฐ yield), `batch`(๋ฆฌ์คํธ ๋ณ๋ ฌ ์ฒ๋ฆฌ). ์คํธ๋ฆฌ๋ฐ UI์๋ `stream`์ ์๋๋ค.",
"**RunnablePassthrough**: ์
๋ ฅ์ ๊ทธ๋๋ก ๋ค์ ๋จ๊ณ๋ก ์ ๋ฌ. `.assign(key=fn)`์ผ๋ก ์ถ๊ฐ ํค๋ฅผ ๋ถ์ฌ ๋์
๋๋ฆฌ๋ฅผ ํ์ฅํฉ๋๋ค.",
"**RunnableParallel**: `{\"a\": fn1, \"b\": fn2}` ํํ๋ก ์ฌ๋ฌ ์ฒด์ธ์ ๋์์ ์คํํด ๊ฒฐ๊ณผ๋ฅผ ๋์
๋๋ฆฌ๋ก ํฉ์นฉ๋๋ค."
]
},
{
"id": "docs-vectorstore",
"certification": "Docs Study",
"title": "VectorStore์ Retriever โ ๋ฒกํฐ ๊ฒ์",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-11", "docs-lab-12"],
"summary": "๋ฌธ์๋ฅผ ์๋ฒ ๋ฉ ๋ฒกํฐ๋ก ๋ณํํด ์ ์ฅํ๊ณ , ์ฟผ๋ฆฌ์ ์ฝ์ฌ์ธ ์ ์ฌ๋๋ก ๊ด๋ จ ์ฒญํฌ๋ฅผ ๊ฒ์ํฉ๋๋ค. Chroma, FAISS, pgvector ๋ฑ ๋ค์ํ ๋ฐฑ์๋๋ฅผ LangChain์ด ์ถ์ํํฉ๋๋ค.",
"example": "from langchain_community.vectorstores import Chroma\nfrom langchain_ollama import OllamaEmbeddings\n\nembeddings = OllamaEmbeddings(model='nomic-embed-text')\ndb = Chroma.from_documents(docs, embeddings, persist_directory='./db')\nretriever = db.as_retriever(search_type='mmr', search_kwargs={'k': 4})",
"common_mistake": "์๋ฒ ๋ฉ ๋ชจ๋ธ์ ๋ณ๊ฒฝํ๋ฉด ๋ฒกํฐ ์ฐจ์์ด ๋ฌ๋ผ์ ธ ๊ธฐ์กด ์ปฌ๋ ์
๊ณผ ํธํ๋์ง ์์ต๋๋ค. ๋ชจ๋ธ์ ๋ฐ๊ฟ ๋๋ ์ปฌ๋ ์
์ ์๋ก ๋ง๋ค์ด์ผ ํฉ๋๋ค.",
"keywords": ["vectorstore", "Chroma", "FAISS", "embedding", "similarity", "MMR", "retriever"],
"source_id": "official-docs-langchain",
"details": [
"**๊ฒ์ ์ ํ**: `similarity`(์์ ์ ์ฌ๋ ์์ k๊ฐ), `mmr`(MMR: ๋ค์์ฑ ๊ณ ๋ ค), `similarity_score_threshold`(์๊ณ๊ฐ ์ด์๋ง).",
"**Chroma**: ๋ก์ปฌ ํ์ผ๋ก ์ ์ฅ ๊ฐ๋ฅ(`persist_directory`). ํ๋กํ ํ์
์ ์ ํฉ. ํ๋ก๋์
์ pgvector(PostgreSQL), Qdrant, Pinecone ๊ถ์ฅ.",
"์๋ฒ ๋ฉ์ ํ
์คํธ์ ์๋ฏธ์ ์ ์ฌ๋๋ฅผ ์ก์ง๋ง, ์ ํํ ํค์๋ ๋งค์นญ์ด ํ์ํ๋ฉด BM25 ๋ฑ ํค์๋ ๊ฒ์๊ณผ `EnsembleRetriever`๋ก ํ์ด๋ธ๋ฆฌ๋ ๊ตฌ์ฑ์ ๊ณ ๋ คํฉ๋๋ค."
]
},
{
"id": "docs-streamlit-state",
"certification": "Docs Study",
"title": "Streamlit session_state โ ์ํ ๊ด๋ฆฌ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-16", "docs-lab-17"],
"summary": "Streamlit์ ์ํธ์์ฉ๋ง๋ค ์คํฌ๋ฆฝํธ ์ ์ฒด๋ฅผ rerunํฉ๋๋ค. ๋ฒํผ ํด๋ฆญยท์
๋ ฅ ๊ฒฐ๊ณผ๋ฅผ rerun ์ดํ์๋ ์ ์งํ๋ ค๋ฉด `st.session_state`์ ์ ์ฅํด์ผ ํฉ๋๋ค.",
"example": "if 'count' not in st.session_state:\n st.session_state.count = 0\nif st.button('์ฆ๊ฐ'):\n st.session_state.count += 1\nst.write(f'ํด๋ฆญ ํ์: {st.session_state.count}')",
"common_mistake": "์ผ๋ฐ ์ ์ญ ๋ณ์๋ ํจ์ ๋ด ์ง์ญ ๋ณ์๋ rerun๋ง๋ค ์ด๊ธฐํ๋ฉ๋๋ค. ์ํ ๋ณด์กด์ด ํ์ํ ๋ชจ๋ ๊ฐ์ ๋ฐ๋์ `st.session_state`๋ฅผ ์ฌ์ฉํ์ธ์.",
"keywords": ["session_state", "rerun", "widget", "state management", "callback"],
"source_id": "official-docs-streamlit",
"details": [
"**์ด๊ธฐํ ํจํด**: `st.session_state.setdefault('key', default_value)` ๋๋ `if 'key' not in st.session_state:` ๋ธ๋ก์ผ๋ก ์์ ํ๊ฒ ์ด๊ธฐํํฉ๋๋ค.",
"**Widget key**: `st.text_input('์ด๋ฆ', key='username')`์ผ๋ก ์์ ฏ ๊ฐ์ด ์๋์ผ๋ก `st.session_state.username`์ ์ ์ฅ๋ฉ๋๋ค.",
"**callback**: `on_change=my_func`์ผ๋ก ์์ ฏ ๋ณ๊ฒฝ ์ ํจ์๋ฅผ ์คํํฉ๋๋ค. callback ์์์ session_state๋ฅผ ์์ ํ๋ฉด rerun ์ ์ ์ฒ๋ฆฌ๋ฉ๋๋ค."
]
},
{
"id": "docs-streamlit-cache",
"certification": "Docs Study",
"title": "Streamlit Cache โ ์บ์ ๋ฐ์ฝ๋ ์ดํฐ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-18"],
"summary": "rerun๋ง๋ค ๋ฐ๋ณต ์คํ๋๋ ๋น์ฉ ๋์ ๊ณ์ฐยท๋ฆฌ์์ค ๋ก๋ฉ์ ์บ์ํด ์ฑ๋ฅ์ ๋์
๋๋ค. `@st.cache_data`๋ ์ง๋ ฌํ ๊ฐ๋ฅ ๋ฐ์ดํฐ, `@st.cache_resource`๋ DBยท๋ชจ๋ธ ๊ฐ์ ๊ณต์ ๊ฐ์ฒด์ ์๋๋ค.",
"example": "@st.cache_resource\ndef load_model():\n return SentenceTransformer('all-MiniLM-L6-v2')\n\n@st.cache_data(ttl=300)\ndef fetch_data(query: str):\n return db.execute(query).fetchall()",
"common_mistake": "`cache_data`์ `cache_resource`๋ฅผ ํผ๋ํ๋ฉด ๋ฌธ์ ๊ฐ ์๊น๋๋ค. LLM ๋ชจ๋ธ์ด๋ DB ์ฐ๊ฒฐ์ฒ๋ผ ๊ณต์ ํด์ผ ํ๋ ๊ฐ์ฒด๋ ๋ฐ๋์ `cache_resource`๋ฅผ ์จ์ผ ํฉ๋๋ค.",
"keywords": ["cache_data", "cache_resource", "TTL", "rerun", "performance"],
"source_id": "official-docs-streamlit",
"details": [
"**`@st.cache_data`**: ํจ์ ์ธ์๋ฅผ ํค๋ก ๊ฒฐ๊ณผ๋ฅผ ๋ณต์ฌํด ์บ์. DataFrame, ๋ฆฌ์คํธ ๋ฑ ์ง๋ ฌํ ๊ฐ๋ฅ ๋ฐํ๊ฐ์ ์ ํฉ. `ttl` ์ธ์๋ก ๋ง๋ฃ ์๊ฐ ์ค์ ๊ฐ๋ฅ.",
"**`@st.cache_resource`**: ์ฑ๊ธํค์ฒ๋ผ ํ ๋ฒ๋ง ์์ฑํด ๋ชจ๋ ์ธ์
์ด ๊ณต์ . ๋ชจ๋ธ ๋ก๋ฉ, DB ์ฐ๊ฒฐ, ๋ฒกํฐ ์คํ ์ด ์ด๊ธฐํ์ ์ ํฉ.",
"์บ์ ๋ฌดํจํ: `st.cache_data.clear()` / `st.cache_resource.clear()`๋ก ์๋ ์ด๊ธฐํ. ํจ์ ๋ณธ๋ฌธ์ด ๋ฐ๋๋ฉด ์๋์ผ๋ก ์บ์๊ฐ ๋ฌดํจํ๋ฉ๋๋ค."
]
},
{
"id": "docs-streamlit-secrets",
"certification": "Docs Study",
"title": "Streamlit Secrets โ ๋น๋ฐ ๊ด๋ฆฌ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-4"],
"summary": "API ํค, DB ๋น๋ฐ๋ฒํธ ๋ฑ ๋ฏผ๊ฐ ์ ๋ณด๋ฅผ ์ฝ๋์ ๋ฃ์ง ์๊ณ `.streamlit/secrets.toml` ํ์ผ ๋๋ Streamlit Cloud์ Secrets UI์ ์ ์ฅํฉ๋๋ค.",
"example": "# .streamlit/secrets.toml\n[database]\nhost = 'localhost'\npassword = 'secret'\n\n# ์ฝ๋์์ ์ ๊ทผ\nimport streamlit as st\ndb_pass = st.secrets['database']['password']",
"common_mistake": "`.streamlit/secrets.toml`์ `.gitignore`์ ์ถ๊ฐํ์ง ์์ผ๋ฉด ๋ฏผ๊ฐ ์ ๋ณด๊ฐ Git์ ์ปค๋ฐ๋ ์ ์์ต๋๋ค. ๋ฐ๋์ ์ ์ธํ์ธ์.",
"keywords": ["secrets", "secrets.toml", "API key", ".gitignore", "Streamlit Cloud"],
"source_id": "official-docs-streamlit",
"details": [
"๋ก์ปฌ ๊ฐ๋ฐ: `.streamlit/secrets.toml`์ TOML ํ์์ผ๋ก ์ ์ฅ. ์ฑ ์ฝ๋์์ `st.secrets`๋ก ๋์
๋๋ฆฌ์ฒ๋ผ ์ ๊ทผํฉ๋๋ค.",
"Streamlit Cloud: ์ฑ Settings โ Secrets์ ๊ฐ์ TOML ๋ด์ฉ์ ๋ถ์ฌ๋ฃ์ผ๋ฉด ๋ฐฐํฌ ์ ์๋ ์ฃผ์
๋ฉ๋๋ค.",
"`st.secrets`๋ `os.environ`๋ณด๋ค ๊ตฌ์กฐํ๋ ์ ๊ทผ์ ์ง์ํฉ๋๋ค. `st.secrets['key']`์ `st.secrets.key` ๋ ๋ฐฉ์ ๋ชจ๋ ๊ฐ๋ฅํฉ๋๋ค."
]
},
{
"id": "docs-hf-spaces",
"certification": "Docs Study",
"title": "Hugging Face Spaces โ ์ฑ ๋ฐฐํฌ",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-19"],
"summary": "Hugging Face Spaces๋ Streamlit, Gradio, Docker ์ฑ์ ๋ฌด๋ฃ๋ก ํธ์คํ
ํ๋ ํ๊ฒฝ์
๋๋ค. README.md ์๋จ์ YAML ๋ฉํ๋ฐ์ดํฐ๋ก SDK์ ํ์ผ์ ์ง์ ํฉ๋๋ค.",
"example": "# README.md ์๋จ (YAML frontmatter)\n---\ntitle: My App\nsdk: streamlit\napp_file: app.py\npython_version: '3.11'\n---",
"common_mistake": "API ํค ๋ฑ ๋ฏผ๊ฐ ์ ๋ณด๋ ์ฝ๋์ ํ๋์ฝ๋ฉํ์ง ๋ง์ธ์. Space Settings โ Secrets์ ๋ฑ๋กํ๊ณ ์ฝ๋์์๋ `os.environ['MY_KEY']`๋ก ์ฐธ์กฐํฉ๋๋ค.",
"keywords": ["huggingface", "spaces", "SDK", "secrets", "app_file", "README metadata"],
"source_id": "official-docs-huggingface",
"details": [
"**๋ฌด๋ฃ ํฐ์ด**: CPU ๊ธฐ๋ณธ ์ ๊ณต, ๊ณต๊ฐ Space๋ ๋ฌด๋ฃ. GPU๋ ๋น๊ณต๊ฐ Space๋ ์ ๋ฃ ํ๋ ๋๋ ZeroGPU(๋ฌด๋ฃ ๊ณต์ GPU) ํ์ฉ.",
"**Secrets**: `Settings โ Repository secrets`์ ํค-๊ฐ์ผ๋ก ๋ฑ๋ก. ๋น๋ ๋ฐ ๋ฐํ์์ ํ๊ฒฝ ๋ณ์๋ก ์ฃผ์
๋ฉ๋๋ค.",
"Space๋ฅผ Duplicate(๋ณต์ )ํ๋ฉด fork์ฒ๋ผ ๋
๋ฆฝ๋ Space๊ฐ ์์ฑ๋ฉ๋๋ค. ํ์ธ์ ๋ชจ๋ธยท์ฑ์ ๋ด Secrets์ ํจ๊ป ์ปค์คํ
ํ๋ ์ผ๋ฐ์ ์ธ ํจํด์
๋๋ค."
]
},
{
"id": "docs-hf-docker-space",
"certification": "Docs Study",
"title": "Hugging Face Docker Space",
"level": "์
๋ฌธ",
"related_practices": ["docs-lab-20"],
"summary": "Dockerfile๋ก ์ฑ ํ๊ฒฝ์ ์ง์ ์ ์ํ๋ Space ๋ฐฐํฌ ๋ฐฉ์์
๋๋ค. Streamlit/Gradio SDK ์ ์ฝ์ ๋์ด FastAPI, ์ปค์คํ
์๋ฒ ๋ฑ ์ด๋ค ์ฑ๋ ๋ฐฐํฌํ ์ ์์ต๋๋ค.",
"example": "# README.md\n---\ntitle: My API\nsdk: docker\napp_port: 7860\n---\n\n# Dockerfile\nFROM python:3.11-slim\nCOPY . .\nRUN pip install -r requirements.txt\nEXPOSE 7860\nCMD [\"uvicorn\", \"main:app\", \"--host\", \"0.0.0.0\", \"--port\", \"7860\"]",
"common_mistake": "Docker Space๋ ํฌํธ 7860์ ๊ธฐ๋ณธ์ผ๋ก ๋
ธ์ถํฉ๋๋ค(`app_port: 7860`). ์ฑ์ด ๋ค๋ฅธ ํฌํธ๋ฅผ ์ฌ์ฉํ๋ฉด README ๋ฉํ๋ฐ์ดํฐ์ `app_port`์ ์ผ์น์์ผ์ผ ํฉ๋๋ค.",
"keywords": ["docker", "Dockerfile", "Spaces", "app_port", "FastAPI", "custom server"],
"source_id": "official-docs-huggingface",
"details": [
"**๋น๋ ์บ์**: Space๋ ์ด์ ๋น๋ ๋ ์ด์ด๋ฅผ ์บ์ํฉ๋๋ค. `COPY requirements.txt .` ํ `RUN pip install`์ ๋ฐฐ์นํด ์์กด์ฑ ๋ ์ด์ด๋ฅผ ์ฑ ์ฝ๋๋ณด๋ค ์์ ๋๋ฉด ๋น๋๊ฐ ๋น ๋ฆ
๋๋ค.",
"**Secrets**: `docker build --build-arg`๊ฐ ์๋ Runtime Secrets๋ก ์ฃผ์
ํด์ผ ์ด๋ฏธ์ง์ ๋ฏผ๊ฐ ์ ๋ณด๊ฐ ํฌํจ๋์ง ์์ต๋๋ค. `os.environ`์ผ๋ก ์ ๊ทผํฉ๋๋ค.",
"**๋ชจ๋ธ ํ์ผ**: ๋ํ ๋ชจ๋ธ์ Space ์ ์ฅ์์ ์ง์ ๋ฃ์ง ๋ง๊ณ `huggingface_hub.snapshot_download()`๋ก ๋ฐํ์์ ๊ฐ์ ธ์ค๊ฑฐ๋ Dataset ์ ์ฅ์๋ฅผ ๋ง์ดํธํ๋ ๋ฐฉ์์ ์๋๋ค."
]
},
{
"id": "docs-fastapi",
"certification": "Docs Study",
"title": "FastAPI โ LLM ์๋น์ค API ์๋ฒ",
"level": "์ค๊ธ",
"related_practices": ["docs-lab-21"],
"summary": "FastAPI๋ Python ํ์
ํํธ ๊ธฐ๋ฐ์ ๊ณ ์ฑ๋ฅ ๋น๋๊ธฐ API ํ๋ ์์ํฌ์
๋๋ค. RAG ์ฒด์ธ์ด๋ LLM ๊ธฐ๋ฅ์ REST API ์๋ํฌ์ธํธ๋ก ๋
ธ์ถํ ๋ ๊ฐ์ฅ ๋ง์ด ์ฐ์
๋๋ค.",
"example": "from fastapi import FastAPI\nfrom pydantic import BaseModel\n\napp = FastAPI()\n\nclass Query(BaseModel):\n question: str\n\n@app.post('/ask')\nasync def ask(q: Query):\n answer = chain.invoke(q.question)\n return {\"answer\": answer}\n\n# ์คํ\n# uvicorn main:app --reload",
"common_mistake": "LangChain ์ฒด์ธ ๊ฐ์ ๋ฌด๊ฑฐ์ด ๊ฐ์ฒด๋ฅผ ์์ฒญ๋ง๋ค ์์ฑํ์ง ๋ง์ธ์. ๋ชจ๋ ๋ ๋ฒจ์์ ํ ๋ฒ ์ด๊ธฐํํ๊ฑฐ๋ `lifespan` ์ด๋ฒคํธ๋ก ์ฑ ์์ ์ ๋ก๋ํ์ธ์.",
"keywords": ["fastapi", "uvicorn", "pydantic", "async", "REST API", "endpoint", "lifespan"],
"source_id": "official-docs-fastapi",
"details": [
"**Pydantic**: ์์ฒญ/์๋ต ์คํค๋ง๋ฅผ `BaseModel`๋ก ์ ์ํ๋ฉด ์๋ ๊ฒ์ฆ, OpenAPI ๋ฌธ์(`/docs`), JSON ์ง๋ ฌํ๊ฐ ๋ฌด๋ฃ๋ก ์ ๊ณต๋ฉ๋๋ค.",
"**์คํธ๋ฆฌ๋ฐ ์๋ต**: `StreamingResponse`์ async generator๋ก LLM ์ถ๋ ฅ์ Server-Sent Events(SSE)๋ก ์คํธ๋ฆฌ๋ฐํ ์ ์์ต๋๋ค.",
"**lifespan**: `@asynccontextmanager`๋ก ์ฑ ์์ ์ ๋ชจ๋ธยทDB ์ฐ๊ฒฐ์ ์ด๊ธฐํํ๊ณ ์ข
๋ฃ ์ ์ ๋ฆฌํ๋ ํจํด. FastAPI 0.95+ ๊ถ์ฅ ๋ฐฉ์์
๋๋ค."
]
},
{
"id": "docs-langsmith",
"certification": "Docs Study",
"title": "LangSmith โ LLM ์ถ์ & ํ๊ฐ",
"level": "์ค๊ธ",
"related_practices": ["docs-lab-13"],
"summary": "LangSmith๋ LangChain ํ์ดํ๋ผ์ธ์ ๋ชจ๋ LLM ํธ์ถ, ํ๋กฌํํธ, ์๋ต, ์ง์ฐ์๊ฐ์ ์๋ ์ถ์ ํฉ๋๋ค. ๋ฒ๊ทธ ๋๋ฒ๊น
, ํ๋กฌํํธ ๋น๊ต, ํ๊ฐ ๋ฐ์ดํฐ์
๊ด๋ฆฌ๋ฅผ ํ ๊ณณ์์ ํฉ๋๋ค.",
"example": "import os\nos.environ['LANGCHAIN_TRACING_V2'] = 'true'\nos.environ['LANGCHAIN_API_KEY'] = 'ls__...'\nos.environ['LANGCHAIN_PROJECT'] = 'my-rag-app'\n\n# ์ดํ LangChain ์ฝ๋๋ ์๋์ผ๋ก ์ถ์ ๋จ\nresult = chain.invoke({'question': '์ง๋ฌธ'})",
"common_mistake": "์ถ์ ํ์ฑํ๋ ํ๊ฒฝ๋ณ์ 3๊ฐ(`TRACING_V2=true`, `API_KEY`, `PROJECT`)๊ฐ ๋ชจ๋ ์์ด์ผ ํฉ๋๋ค. ํ๋๋ผ๋ ๋น ์ง๋ฉด ์กฐ์ฉํ ์ถ์ ์ด ์ ๋ฉ๋๋ค.",
"keywords": ["langsmith", "tracing", "LangChain", "debugging", "evaluation", "dataset", "prompt"],
"source_id": "official-docs-langchain",
"details": [
"**Run ํธ๋ ์ด์ค**: ์ฒด์ธ ์คํ๋ง๋ค tree ํํ๋ก ๊ฐ ๋จ๊ณ์ ์
์ถ๋ ฅยทํ ํฐ ์ยท์ง์ฐ์๊ฐ์ ๊ธฐ๋กํฉ๋๋ค. ์ด๋ ๋จ๊ณ์์ ์ค๋ฅ๊ฐ ๋ฌ๋์ง ์ฆ์ ํ์
ํ ์ ์์ต๋๋ค.",
"**๋ฐ์ดํฐ์
& ํ๊ฐ**: ํ๋ก๋์
์ค๋ฅ ์ผ์ด์ค๋ฅผ Dataset์ผ๋ก ์ ์ฅ โ Evaluator(LLM-as-judge ๋ฑ)๋ก ์๋ ํ๊ฐ โ CI์ ํตํฉํ๋ ํ๋ฆ์ ์ง์ํฉ๋๋ค.",
"**Playground**: LangSmith UI์์ ํน์ Run์ ํ๋กฌํํธ๋ฅผ ๋ฐ๋ก ์์ ํด ์ฌ์คํ ๊ฐ๋ฅ. ํ๋กฌํํธ ์์ง๋์ด๋ง ๋ฐ๋ณต์ ์ ์ฉํฉ๋๋ค."
]
},
{
"id": "docs-guardrails",
"certification": "Docs Study",
"title": "Guardrails AI โ LLM ์ถ๋ ฅ ๊ฒ์ฆ",
"level": "์ค๊ธ",
"related_practices": ["docs-lab-14"],
"summary": "Guardrails AI๋ LLM ์
๋ ฅ/์ถ๋ ฅ์ ๊ฒ์ฆ ๊ท์น(Guard)์ ์ ์ฉํฉ๋๋ค. ์ ํด ์ฝํ
์ธ ์ฐจ๋จ, JSON ์คํค๋ง ๊ฐ์ , PII ๋ง์คํน ๋ฑ ํ๋ก๋์
LLM ์ฑ์ ์์ ์ฅ์น ์ญํ ์ ํฉ๋๋ค.",
"example": "from guardrails import Guard\nfrom guardrails.hub import ToxicLanguage\n\nguard = Guard().use(ToxicLanguage(threshold=0.5, on_fail='exception'))\n\nresult = guard(\n llm_api=openai.chat.completions.create,\n prompt='์ฌ์ฉ์ ์
๋ ฅ: {{user_input}}',\n user_input=user_message\n)",
"common_mistake": "Guard๋ LLM ํธ์ถ์ ๋ํํ๋ฏ๋ก ์๋ต ์๊ฐ์ด ์ฆ๊ฐํฉ๋๋ค. ๋ชจ๋ ์์ฒญ์ heavy Guard๋ฅผ ๊ฑธ๋ฉด ์ง์ฐ์ด ์ปค์ง๋, ์
๋ ฅ ๊ฒ์ฆ๊ณผ ์ถ๋ ฅ ๊ฒ์ฆ์ ๋ถ๋ฆฌํด ํ์ํ ๊ฒ๋ง ์ ์ฉํ์ธ์.",
"keywords": ["guardrails", "guard", "validation", "toxic", "PII", "JSON schema", "on_fail", "safety"],
"source_id": "official-docs-guardrails",
"details": [
"**Guard ํ๋ธ**: `guardrails hub install hub://...`์ผ๋ก ์ปค๋ฎค๋ํฐ Guard๋ฅผ ์ค์นํฉ๋๋ค. ToxicLanguage, ValidJson, DetectPII, RegexMatch ๋ฑ ์์ญ ๊ฐ์ง๊ฐ ์์ต๋๋ค.",
"**on_fail ๋์**: `'exception'`(์์ธ ๋ฐ์), `'reask'`(LLM์ ์ฌ์์ฒญ), `'fix'`(์๋ ์์ ), `'noop'`(ํต๊ณผ) ์ค ์ ํํฉ๋๋ค.",
"**๊ตฌ์กฐํ ์ถ๋ ฅ**: `Guard.from_pydantic(OutputModel)`์ผ๋ก LLM์ด ํญ์ ์ง์ ํ ์คํค๋ง์ JSON์ ๋ฐํํ๋๋ก ๊ฐ์ ํ ์ ์์ต๋๋ค."
]
},
{
"id": "docs-gradio",
"certification": "Docs Study",
"title": "Gradio โ ML ๋ฐ๋ชจ UI",
"level": "์ค๊ธ",
"related_practices": ["docs-lab-22"],
"summary": "Gradio๋ ML ๋ชจ๋ธ ๋ฐ๋ชจ๋ฅผ ๋ช ์ค๋ก ๋ง๋๋ UI ๋ผ์ด๋ธ๋ฌ๋ฆฌ์
๋๋ค. `gr.Interface`๋ก ํจ์๋ฅผ ๊ฐ์ธ๋ฉด ์ฆ์ ์น UI๊ฐ ์๊ธฐ๊ณ , HuggingFace Spaces์ ๋ค์ดํฐ๋ธ ํตํฉ๋ฉ๋๋ค.",
"example": "import gradio as gr\n\ndef chat(message, history):\n response = chain.invoke(message)\n return response\n\ndemo = gr.ChatInterface(\n fn=chat,\n title='RAG ์ฑ๋ด',\n examples=['๋ฌธ์ ์์ฝํด์ค', '์ฃผ์ ํค์๋๋?']\n)\ndemo.launch()",
"common_mistake": "Streamlit๊ณผ ๋ฌ๋ฆฌ Gradio๋ ํจ์ ๋จ์๋ก UI๋ฅผ ์์ฑํฉ๋๋ค. ๋ณต์กํ ์ํ ๊ด๋ฆฌ๋ `gr.State()`๋ก, ์ฌ๋ฌ ์ปดํฌ๋ํธ๋ฅผ ์กฐํฉํ ๋๋ `gr.Blocks()` ์ปจํ
์คํธ๋ฅผ ์๋๋ค.",
"keywords": ["gradio", "Interface", "Blocks", "ChatInterface", "State", "HuggingFace", "demo"],
"source_id": "official-docs-gradio",
"details": [
"**Interface vs Blocks**: `gr.Interface`๋ ์
๋ ฅโํจ์โ์ถ๋ ฅ ๋จ์ ๊ตฌ์กฐ, `gr.Blocks`๋ ๋ ์ด์์๊ณผ ์ด๋ฒคํธ๋ฅผ ์์ ๋กญ๊ฒ ๊ตฌ์ฑ. LLM ์ฑ๋ด์๋ `gr.ChatInterface`๊ฐ ๊ฐ์ฅ ์ ํฉํฉ๋๋ค.",
"**์คํธ๋ฆฌ๋ฐ**: `gr.ChatInterface(fn=chat)`์์ `chat`์ generator ํจ์๋ก ๋ง๋ค๊ณ `yield`๋ก ํ ํฐ์ ๋ด๋ณด๋ด๋ฉด ์๋์ผ๋ก ์คํธ๋ฆฌ๋ฐ UI๊ฐ ๋ฉ๋๋ค.",
"**HF Spaces ํตํฉ**: `demo.launch()`๋ง ํ๋ฉด HF Spaces Gradio SDK๋ก ์ฆ์ ๋ฐฐํฌ๋ฉ๋๋ค. ๋ณ๋ ์๋ฒ ์ค์ ์์ด `sdk: gradio`๋ง README์ ์ ์ธํ๋ฉด ๋ฉ๋๋ค."
]
},
{
"id": "docs-chroma",
"certification": "Docs Study",
"title": "Chroma โ ์์ ๋ฒกํฐ DB",
"level": "์ค๊ธ",
"related_practices": ["docs-chroma-1", "docs-chroma-2"],
"summary": "Chroma๋ ์ฌ์์ํด๋ ๋ฐ์ดํฐ๊ฐ ์ ์ง๋๋ ์คํ์์ค ๋ฒกํฐ DB์
๋๋ค. FAISS๋ `save_local()`์ ์ง์ ํธ์ถํด์ผ ํ์ง๋ง, Chroma๋ `persist_directory`๋ง ์ง์ ํ๋ฉด ์๋์ผ๋ก ๋ก์ปฌ DB์ ์ ์ฅํฉ๋๋ค.",
"example": "from langchain_chroma import Chroma\n\n# ์์ฑ & ์๋ ์ ์ฅ\ndb = Chroma.from_documents(\n chunks,\n embeddings,\n persist_directory='./chroma_db'\n)\n\n# ๋ค์ ์คํ ์ โ ์๋ฒ ๋ฉ ์ฌ๊ณ์ฐ ์์ด ๋ก๋\ndb = Chroma(\n persist_directory='./chroma_db',\n embedding_function=embeddings\n)",
"common_mistake": "FAISS์ฒ๋ผ `save_local()`์ ๋ณ๋๋ก ํธ์ถํ ํ์๊ฐ ์์ต๋๋ค. `persist_directory`๋ฅผ ์ง์ ํ๋ฉด ์๋ ์ ์ฅ๋ฉ๋๋ค. ๋จ, `persist_directory`๋ฅผ ์๋ตํ๋ฉด ์ธ๋ฉ๋ชจ๋ฆฌ๋ก๋ง ๋์ํฉ๋๋ค.",
"keywords": ["Chroma", "persist_directory", "from_documents", "collection", "chromadb", "์์"],
"source_id": "official-docs-chromadb",
"details": [
"**FAISS vs Chroma**: FAISS๋ ์ธ๋ฉ๋ชจ๋ฆฌ ๊ธฐ๋ฐ์ผ๋ก `save_local()`๋ก ์๋ ์ ์ฅํด์ผ ํฉ๋๋ค. Chroma๋ `persist_directory`๋ง ์ง์ ํ๋ฉด SQLite ๊ธฐ๋ฐ ๋ก์ปฌ DB์ ์๋ ์ ์ฅ๋ฉ๋๋ค.",
"**์ปฌ๋ ์
**: Chroma๋ ๋ฐ์ดํฐ๋ฅผ ์ปฌ๋ ์
๋จ์๋ก ๊ด๋ฆฌํฉ๋๋ค. `collection_name` ํ๋ผ๋ฏธํฐ๋ก ์ฌ๋ฌ ๋ฌธ์ ์ธํธ๋ฅผ ๋ถ๋ฆฌํด ์ ์ฅํ๊ณ ๊ฐ๋ณ๋ก ๊ฒ์ํ ์ ์์ต๋๋ค.",
"**๋ฉํ๋ฐ์ดํฐ ํํฐ๋ง**: `db.similarity_search(query, filter={\"source\": \"file.pdf\"})`์ฒ๋ผ ๋ฉํ๋ฐ์ดํฐ ์กฐ๊ฑด์ผ๋ก ๊ฒ์ ๋ฒ์๋ฅผ ์ ํํ ์ ์์ต๋๋ค. FAISS๋ ์ง์ํ์ง ์๋ ๊ธฐ๋ฅ์
๋๋ค."
]
},
{
"id": "docs-sentence-transformers",
"certification": "Docs Study",
"title": "sentence-transformers โ ๋ก์ปฌ ์๋ฒ ๋ฉ",
"level": "์ค๊ธ",
"related_practices": ["docs-embed-1"],
"summary": "sentence-transformers๋ HuggingFace ์๋ฒ ๋ฉ ๋ชจ๋ธ์ ๋ก์ปฌ์์ ์คํํฉ๋๋ค. OpenAI API ์์ด ์์ ๋ก์ปฌ RAG๋ฅผ ๊ตฌ์ฑํ ์ ์๊ณ , ํ๊ตญ์ด ํนํ ๋ชจ๋ธ๋ ์ ํํ ์ ์์ต๋๋ค.",
"example": "from langchain_huggingface import HuggingFaceEmbeddings\n\nembeddings = HuggingFaceEmbeddings(\n model_name='sentence-transformers/all-MiniLM-L6-v2'\n)\n\n# ์๋ฒ ๋ฉ ๋ฒกํฐ ์์ฑ\nvector = embeddings.embed_query('ํ์ด์ฌ์ด ๋ญ์ผ?')\nprint(len(vector)) # 384์ฐจ์\n\n# Chroma์ ํจ๊ป\ndb = Chroma.from_documents(chunks, embeddings, persist_directory='./db')",
"common_mistake": "์ฒซ ์คํ ์ ๋ชจ๋ธ ํ์ผ์ ๋ค์ด๋ก๋ํ๋ฏ๋ก ๋๋ฆฝ๋๋ค. Streamlit ์ฑ์์๋ `@st.cache_resource`๋ก ๊ฐ์ธ ํ ๋ฒ๋ง ๋ก๋ํ์ธ์. ์๋ฒ ๋ฉ ๋ชจ๋ธ์ ์ค๊ฐ์ ๋ฐ๊พธ๋ฉด ๊ธฐ์กด VectorStore์ ํธํ๋์ง ์์ ์ ์ฒด ์ฌ์ธ๋ฑ์ฑ์ด ํ์ํฉ๋๋ค.",
"keywords": ["HuggingFaceEmbeddings", "sentence-transformers", "all-MiniLM-L6-v2", "embed_query", "๋ก์ปฌ ์๋ฒ ๋ฉ", "384์ฐจ์"],
"source_id": "official-docs-huggingface",
"details": [
"**๋ชจ๋ธ ์ ํ**: `all-MiniLM-L6-v2`๋ ๋น ๋ฅด๊ณ ๊ฒฝ๋(384์ฐจ์), `all-mpnet-base-v2`๋ ํ์ง์ด ๋ ๋์ง๋ง ๋๋ฆฝ๋๋ค. ํ๊ตญ์ด๊ฐ ๋ง๋ค๋ฉด `jhgan/ko-sroberta-multitask` ๊ฐ์ ํ๊ตญ์ด ํนํ ๋ชจ๋ธ์ ์๋๋ค.",
"**OpenAI vs ๋ก์ปฌ**: OpenAI `text-embedding-3-small`์ ํ์ง์ด ๋์ง๋ง API ๋น์ฉ์ด ๋ฐ์ํฉ๋๋ค. ๋ก์ปฌ ์๋ฒ ๋ฉ์ ๋ฌด๋ฃ์ด๋ GPU ์์ด๋ ๋์ฉ๋์์ ๋๋ฆฝ๋๋ค.",
"**์ฐจ์ ๊ณ ์ ์ฃผ์**: ํ ๋ฒ ์๋ฒ ๋ฉํ ๋ชจ๋ธ์ ๋ฐ๊พธ๋ฉด ๊ธฐ์กด ์ธ๋ฑ์ค์ ํธํ๋์ง ์์ต๋๋ค. ์ ๋ชจ๋ธ๋ก ์ ์ฒด ์ธ๋ฑ์ค๋ฅผ ์ฌ์์ฑํด์ผ ํฉ๋๋ค."
]
},
{
"id": "docs-ragas",
"certification": "Docs Study",
"title": "Ragas โ RAG ํ์ง ์๋ ํ๊ฐ",
"level": "์ค๊ธ",
"related_practices": ["docs-ragas-1", "docs-ragas-2"],
"summary": "Ragas๋ RAG ์์คํ
์ ํ์ง์ ์๋์ผ๋ก ์์นํํฉ๋๋ค. Faithfulness(ํ๊ฐ ์๋๊ฐ), Answer Relevancy(์ง๋ฌธ๊ณผ ๋ง๋ ๋ต์ธ๊ฐ), Context Precision(๊ด๋ จ ๋ฌธ์๋ฅผ ์ ์ฐพ์๋๊ฐ)์ LLM-as-judge๋ก ๊ณ์ฐํฉ๋๋ค.",
"example": "from ragas import evaluate\nfrom ragas.metrics import faithfulness, answer_relevancy, context_precision\nfrom datasets import Dataset\n\ndata = {\n 'question': ['LangChain์ด ๋ญ์ผ?'],\n 'answer': ['LangChain์ LLM ์ฑ ํ๋ ์์ํฌ์
๋๋ค.'],\n 'contexts': [['LangChain์ LLM ์ ํ๋ฆฌ์ผ์ด์
๊ฐ๋ฐ์ ์ํ ํ๋ ์์ํฌ์
๋๋ค.']],\n 'ground_truth': ['LangChain์ LLM ์ฑ ๊ฐ๋ฐ ํ๋ ์์ํฌ์
๋๋ค.']\n}\n\ndataset = Dataset.from_dict(data)\nresult = evaluate(dataset, metrics=[faithfulness, answer_relevancy, context_precision])\nprint(result)",
"common_mistake": "Ragas๋ ํ๊ฐ์ LLM์ ์ฌ์ฉํฉ๋๋ค. ๊ธฐ๋ณธ์ผ๋ก OpenAI GPT๋ฅผ ์๋๋ค. `evaluate(..., llm=your_llm, embeddings=your_embeddings)`๋ก ๋ค๋ฅธ ๋ชจ๋ธ์ ์ง์ ํ ์ ์์ต๋๋ค.",
"keywords": ["ragas", "evaluate", "faithfulness", "answer_relevancy", "context_precision", "LLM-as-judge", "Dataset"],
"source_id": "official-docs-ragas",
"details": [
"**3๊ฐ์ง ํต์ฌ ๋ฉํธ๋ฆญ**: Faithfulness(๋ต๋ณ์ด ๊ฒ์๋ ์ปจํ
์คํธ์ ๊ทผ๊ฑฐํ๋๊ฐ โ ํ๊ฐ ํ์ง), Answer Relevancy(๋ต๋ณ์ด ์ง๋ฌธ๊ณผ ๊ด๋ จ ์๋๊ฐ), Context Precision(๊ด๋ จ ๋ฌธ์๊ฐ ์์ k๊ฐ ์์ ์๋๊ฐ).",
"**MLflow ์ฐ๋**: `result.to_pandas()`๋ก DataFrame ๋ณํ ํ `mlflow.log_metrics()`๋ก ๊ธฐ๋กํ๋ฉด chunk_size, k ํ๋ผ๋ฏธํฐ๋ณ ํ์ง ๋ณํ๋ฅผ ์ถ์ ํ ์ ์์ต๋๋ค.",
"**ground_truth ์์ด๋ ๊ฐ๋ฅ**: `ground_truth`๊ฐ ์์ผ๋ฉด faithfulness, answer_relevancy๋ง ๊ณ์ฐํฉ๋๋ค. Answer Correctness ๋ฉํธ๋ฆญ์ ground_truth๊ฐ ํ์ํฉ๋๋ค."
]
},
{
"id": "docs-summarization",
"certification": "Docs Study",
"title": "LangChain ์์ฝ ์ฒด์ธ โ map_reduce & stuff",
"level": "์ค๊ธ",
"related_practices": ["docs-news-1", "docs-news-2"],
"summary": "LangChain์ `load_summarize_chain`์ ๊ธด ๋ฌธ์๋ฅผ LLM ์ปจํ
์คํธ ํ๊ณ ๋ด์์ ์์ฝํ๋ ์ฒด์ธ์ ๋ง๋ญ๋๋ค. ๋ฌธ์๊ฐ ์งง์ผ๋ฉด `stuff`, ๊ธธ๋ฉด `map_reduce` ๋ฐฉ์์ ์๋๋ค.",
"example": "from langchain.chains.summarize import load_summarize_chain\nfrom langchain_core.documents import Document\n\n# map_reduce: ๊ฐ ์ฒญํฌ ์์ฝ โ ์ต์ข
์์ฝ\nchain = load_summarize_chain(llm, chain_type='map_reduce')\n\ndocs = [Document(page_content=article) for article in articles]\nresult = chain.invoke(docs)\nprint(result['output_text'])",
"common_mistake": "`stuff` ๋ฐฉ์์ ๋ชจ๋ ๋ฌธ์๋ฅผ ํ๋์ ํ๋กฌํํธ์ ๋ฃ์ต๋๋ค. ๋ฌธ์๊ฐ LLM ์ปจํ
์คํธ ์ฐฝ๋ณด๋ค ํฌ๋ฉด ์ค๋ฅ๊ฐ ๋ฉ๋๋ค. ์ฌ๋ฌ ๋ฌธ์๋ ๊ธด ๋ฌธ์๋ ํญ์ `map_reduce`๋ฅผ ๋จผ์ ๊ณ ๋ คํ์ธ์.",
"keywords": ["load_summarize_chain", "map_reduce", "stuff", "refine", "output_text", "์์ฝ"],
"source_id": "official-docs-langchain",
"details": [
"**chain_type ๋น๊ต**: `stuff`(์ ์ฒด ํ ๋ฒ), `map_reduce`(์ฒญํฌ๋ณ ์์ฝ ํ ํฉ์นจ), `refine`(์ด์ ์์ฝ์ ๋ค์ ์ฒญํฌ์ ๋ฐ์). ํ์ง์ refine > map_reduce > stuff์ด์ง๋ง ์๋๋ ๋ฐ๋์
๋๋ค.",
"**map_prompt / combine_prompt**: `load_summarize_chain`์ `map_prompt`์ `combine_prompt`๋ฅผ ์ง์ ํด ์์ฝ ์ง์๋ฌธ์ ์ปค์คํฐ๋ง์ด์งํ ์ ์์ต๋๋ค.",
"**๋น๋๊ธฐ**: `chain.ainvoke(docs)`๋ก ๋น๋๊ธฐ ์์ฒญ์ด ๊ฐ๋ฅํฉ๋๋ค. ์ฌ๋ฌ ๋ด์ค ๊ธฐ์ฌ๋ฅผ ๋ณ๋ ฌ ์์ฝํ ๋ `asyncio.gather`์ ํจ๊ป ์ฐ๋ฉด ์๋๊ฐ ํฌ๊ฒ ๊ฐ์ ๋ฉ๋๋ค."
]
},
{
"id": "docs-sql-chain",
"certification": "Docs Study",
"title": "LangChain Text-to-SQL โ ์์ฐ์ด๋ก DB ์กฐํ",
"level": "์ค๊ธ",
"related_practices": ["docs-sql-1", "docs-sql-2"],
"summary": "LangChain์ `create_sql_query_chain`์ ์์ฐ์ด ์ง๋ฌธ์ SQL ์ฟผ๋ฆฌ๋ก ๋ณํํฉ๋๋ค. `SQLDatabase`๋ก DB ์คํค๋ง๋ฅผ ์๋์ผ๋ก ์ฝ์ด LLM์๊ฒ ์ปจํ
์คํธ๋ฅผ ์ ๊ณตํฉ๋๋ค.",
"example": "from langchain_community.utilities import SQLDatabase\nfrom langchain.chains import create_sql_query_chain\n\ndb = SQLDatabase.from_uri('sqlite:///./mydb.db')\nchain = create_sql_query_chain(llm, db)\n\nsql = chain.invoke({'question': '๊ฐ์ฅ ๋ง์ด ํ๋ฆฐ ์ํ 5๊ฐ๋?'})\nprint(sql) # SELECT product, SUM(quantity) ...\n\nresult = db.run(sql)\nprint(result)",
"common_mistake": "LLM์ด ์์ฑํ SQL์ ๋ฐ๋ก `db.run()`์ ๋๊ธฐ๋ฉด ์๋ชป๋ ์ฟผ๋ฆฌ๊ฐ ์คํ๋ฉ๋๋ค. Guardrails์ `ValidSQL`์ด๋ ๋ณ๋ ํ์ฑ์ผ๋ก ๊ฒ์ฆํ ๋ค ์คํํ์ธ์.",
"keywords": ["SQLDatabase", "create_sql_query_chain", "from_uri", "db.run", "Text-to-SQL", "sqlite"],
"source_id": "official-docs-langchain",
"details": [
"**์คํค๋ง ์๋ ์ฃผ์
**: `SQLDatabase`๋ ์ฐ๊ฒฐ๋ DB์ ํ
์ด๋ธ๋ช
๊ณผ ์ปฌ๋ผ ์ ๋ณด๋ฅผ ์๋์ผ๋ก ํ๋กฌํํธ์ ์ฃผ์
ํฉ๋๋ค. LLM์ด ์กด์ฌํ๋ ํ
์ด๋ธ/์ปฌ๋ผ๋ง ์ฐธ์กฐํ๋๋ก ์ ๋ํฉ๋๋ค.",
"**QuerySQLDatabaseTool**: SQL ์์ฑ๊ณผ ์คํ์ ํ๋์ ํด๋ก ๋ฌถ์ ๊ณ ์์ค ์ธํฐํ์ด์ค. Agent์ ๋ฑ๋กํ๋ฉด LLM์ด ํ์ํ ๋ DB๋ฅผ ์ง์ ์กฐํํ ์ ์์ต๋๋ค.",
"**๋ณด์ ์ฃผ์**: `db.run()`์ ์ค์ DB๋ฅผ ์์ ํ ์ ์์ต๋๋ค. ์ฝ๊ธฐ ์ ์ฉ ์ฌ์ฉ์ ๊ณ์ ์ผ๋ก ์ฐ๊ฒฐํ๊ฑฐ๋ SELECT๋ง ํ์ฉํ๋ wrapper๋ฅผ ๋๋ ๊ฒ์ด ์์ ํฉ๋๋ค."
]
},
{
"id": "docs-langgraph",
"certification": "Docs Study",
"title": "LangGraph โ ์ํ ๊ธฐ๋ฐ ์์ด์ ํธ ์ํฌํ๋ก",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-graph-1", "docs-graph-2"],
"summary": "LangGraph๋ LLM ์์ด์ ํธ๋ฅผ ๊ทธ๋ํ(๋
ธ๋ + ์ฃ์ง)๋ก ์ ์ํฉ๋๋ค. ๋จ์ ์ฒด์ธ์ AโBโC์ ๋ฌ๋ฆฌ ์กฐ๊ฑด ๋ถ๊ธฐ, ๋ฃจํ, ๋ณ๋ ฌ ์คํ์ด ๊ฐ๋ฅํด ์ค์ ์์ด์ ํธ ํ๋์ ํํํฉ๋๋ค.",
"example": "from langgraph.graph import StateGraph, END\nfrom typing import TypedDict\n\nclass State(TypedDict):\n question: str\n answer: str\n\ndef retrieve(state: State) -> State:\n return {\"answer\": str(retriever.invoke(state[\"question\"]))}\n\ndef generate(state: State) -> State:\n return {\"answer\": chain.invoke(state)}\n\ngraph = StateGraph(State)\ngraph.add_node(\"retrieve\", retrieve)\ngraph.add_node(\"generate\", generate)\ngraph.set_entry_point(\"retrieve\")\ngraph.add_edge(\"retrieve\", \"generate\")\ngraph.add_edge(\"generate\", END)\n\napp = graph.compile()\nresult = app.invoke({\"question\": \"LangChain์ด ๋ญ์ผ?\"})",
"common_mistake": "LangGraph๋ LangChain ์ฒด์ธ์ด ์๋๋๋ค. `graph.compile()` ํ `app.invoke()`๋ฅผ ์๋๋ค. State TypedDict๋ฅผ ์ ํํ ์ ์ํ์ง ์์ผ๋ฉด ๋
ธ๋์์ ํค ์ค๋ฅ๊ฐ ๋ฐ์ํฉ๋๋ค.",
"keywords": ["LangGraph", "StateGraph", "State", "node", "edge", "compile", "conditional_edges", "END"],
"source_id": "official-docs-langgraph",
"details": [
"**๋
ธ๋ vs ์ฒด์ธ**: ์ฒด์ธ์ AโBโC ์ ํ. ๊ทธ๋ํ๋ A์์ ์กฐ๊ฑด์ ๋ฐ๋ผ B ๋๋ C๋ก ๋ถ๊ธฐ, D์์ A๋ก ๋ฃจํ ๋ฑ ๋ณต์กํ ํ๋ฆ์ ํํํ ์ ์์ต๋๋ค.",
"**conditional_edges**: `graph.add_conditional_edges('judge', router_fn, {'retry': 'retrieve', 'done': END})`์ฒ๋ผ ๋
ธ๋ ๋ฐํ๊ฐ์ ๋ฐ๋ผ ๋ค์ ๋
ธ๋๋ฅผ ๋์ ์ผ๋ก ๊ฒฐ์ ํฉ๋๋ค. ReAct ์์ด์ ํธ์ ํต์ฌ์
๋๋ค.",
"**checkpointer**: `MemorySaver`๋ฅผ `graph.compile(checkpointer=...)`์ ๋๊ธฐ๋ฉด ์คํ ์ํ๋ฅผ ์ ์ฅํด ์ค๋จ ํ ์ฌ๊ฐ, ๋ฉํฐํด ๋ํ ์ ์ง๊ฐ ๊ฐ๋ฅํฉ๋๋ค."
]
},
{
"id": "docs-airflow",
"certification": "Docs Study",
"title": "Apache Airflow โ ์ํฌํ๋ก ์ค์ผ์คํธ๋ ์ด์
",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-lab-23"],
"summary": "Airflow๋ DAG(Directed Acyclic Graph)๋ก ๋ฐ์ดํฐ ํ์ดํ๋ผ์ธ์ ์ ์ํ๊ณ ์ค์ผ์ค๋งยท๋ชจ๋ํฐ๋งํฉ๋๋ค. ML ํ์ดํ๋ผ์ธ(๋ฐ์ดํฐ ์์ง โ ์ ์ฒ๋ฆฌ โ ํ์ต โ ๋ฐฐํฌ)์ด๋ ๋ฌธ์ ์ฌ์ธ๋ฑ์ฑ ์๋ํ์ ์๋๋ค.",
"example": "from airflow import DAG\nfrom airflow.operators.python import PythonOperator\nfrom datetime import datetime\n\ndef reindex():\n # ๋ฌธ์ ๋ก๋ โ ์๋ฒ ๋ฉ โ VectorStore ์
๋ฐ์ดํธ\n pass\n\nwith DAG('daily_reindex', schedule='@daily', start_date=datetime(2024,1,1)) as dag:\n task = PythonOperator(task_id='reindex', python_callable=reindex)",
"common_mistake": "DAG ํ์ผ์ Airflow ์ค์ผ์ค๋ฌ๊ฐ ์ฃผ๊ธฐ์ ์ผ๋ก ํ์ฑํฉ๋๋ค. ํ์ผ ์ต์์์ ๋ฌด๊ฑฐ์ด ์ฐ์ฐ(DB ์ฐ๊ฒฐ, ๋ชจ๋ธ ๋ก๋)์ ๋๋ฉด ํ์ฑ๋ง๋ค ์คํ๋ผ ์ค์ผ์ค๋ฌ๊ฐ ๋๋ ค์ง๋๋ค. ๋ฐ๋์ task ํจ์ ์์ ๋ฃ์ผ์ธ์.",
"keywords": ["airflow", "DAG", "task", "operator", "schedule", "XCom", "pipeline", "orchestration"],
"source_id": "official-docs-airflow",
"details": [
"**Operator ์ข
๋ฅ**: `PythonOperator`(ํ์ด์ฌ ํจ์), `BashOperator`(์
ธ ๋ช
๋ น), `DockerOperator`(์ปจํ
์ด๋ ์คํ) ๋ฑ. ML ํ์ดํ๋ผ์ธ์ ์ฃผ๋ก PythonOperator๋ฅผ ์๋๋ค.",
"**XCom**: task ๊ฐ ๋ฐ์ดํฐ ์ ๋ฌ ๋ฉ์ปค๋์ฆ. `ti.xcom_push(key, value)` / `ti.xcom_pull(task_ids, key)`๋ก ์๋ ๋ฐ์ดํฐ๋ฅผ ๊ณต์ ํฉ๋๋ค. ๋์ฉ๋์ S3/GCS ๊ฒฝ๋ก๋ฅผ ์ ๋ฌํ๋ ๋ฐฉ์์ ์๋๋ค.",
"**์์กด์ฑ**: `task_a >> task_b >> task_c` ๋๋ `task_a.set_downstream(task_b)` ๋ก ์คํ ์์๋ฅผ ์ ์ํฉ๋๋ค. ๋ณ๋ ฌ ์คํ์ `[task_b, task_c] >> task_d` ํํ๋ก ํํํฉ๋๋ค."
]
},
{
"id": "docs-mlflow",
"certification": "Docs Study",
"title": "MLflow โ ์คํ ์ถ์ & ๋ชจ๋ธ ๊ด๋ฆฌ",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-lab-15"],
"summary": "MLflow๋ ML ์คํ์ ํ๋ผ๋ฏธํฐ, ๋ฉํธ๋ฆญ, ์ํฐํฉํธ๋ฅผ ์ถ์ ํ๊ณ ๋ชจ๋ธ์ ๋ฒ์ ๊ด๋ฆฌํฉ๋๋ค. RAG์์ chunk_size, k, ์๋ฒ ๋ฉ ๋ชจ๋ธ ๋ฑ ํ๋ผ๋ฏธํฐ ์คํ ๊ฒฐ๊ณผ๋ฅผ ์ฒด๊ณ์ ์ผ๋ก ๋น๊ตํ ๋ ์๋๋ค.",
"example": "import mlflow\n\nwith mlflow.start_run(run_name='rag-exp-1'):\n mlflow.log_param('chunk_size', 500)\n mlflow.log_param('k', 4)\n mlflow.log_metric('answer_relevance', 0.87)\n mlflow.log_metric('faithfulness', 0.92)\n mlflow.log_artifact('eval_results.json')",
"common_mistake": "run์ `with mlflow.start_run()` ๋ธ๋ก์ผ๋ก ๊ฐ์ธ์ง ์์ผ๋ฉด ์ค์๋ก ์ด์ run์ ๋ก๊น
๋๋ ๊ฒฝ์ฐ๊ฐ ์์ต๋๋ค. ํญ์ context manager๋ฅผ ์ฌ์ฉํ์ธ์.",
"keywords": ["mlflow", "experiment", "run", "log_param", "log_metric", "artifact", "model registry"],
"source_id": "official-docs-mlflow",
"details": [
"**UI**: `mlflow ui` ๋ช
๋ น์ผ๋ก localhost:5000์์ ์คํ ๋น๊ต ๋์๋ณด๋๋ฅผ ์คํํฉ๋๋ค. ํ๋ผ๋ฏธํฐ๋ณ ๋ฉํธ๋ฆญ ์ถ์ด๋ฅผ ์๊ฐ์ ์ผ๋ก ๋น๊ตํ ์ ์์ต๋๋ค.",
"**Model Registry**: `mlflow.log_model()`๋ก ๋ชจ๋ธ์ ์ ์ฅํ๊ณ `Staging`โ`Production` ๋จ๊ณ๋ฅผ ๊ด๋ฆฌํฉ๋๋ค. ๋ฐฐํฌ ์ `mlflow.pyfunc.load_model('models:/name/Production')`์ผ๋ก ๋ก๋ํฉ๋๋ค.",
"**LangChain ํตํฉ**: `mlflow.langchain.autolog()`๋ฅผ ํ์ฑํํ๋ฉด ์ฒด์ธ ์คํ๋ง๋ค ํ๋ผ๋ฏธํฐ์ ๋ฉํธ๋ฆญ์ด ์๋ ๊ธฐ๋ก๋ฉ๋๋ค."
]
},
{
"id": "docs-mcp",
"certification": "Docs Study",
"title": "MCP โ Model Context Protocol",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-lab-24"],
"summary": "MCP(Model Context Protocol)๋ LLM ์ ํ๋ฆฌ์ผ์ด์
์ด ์ธ๋ถ ๋๊ตฌยท๋ฐ์ดํฐ์์ค์ ํ์คํ๋ ๋ฐฉ์์ผ๋ก ์ฐ๊ฒฐํ๋ Anthropic ์คํ ํ๋กํ ์ฝ์
๋๋ค. Claude Desktop์ด๋ Claude Code์ ๋ก์ปฌ MCP ์๋ฒ๋ฅผ ์ฐ๊ฒฐํ๋ฉด ํ์ผ ์์คํ
, DB, ์ธ๋ถ API๋ฅผ LLM์ด ์ง์ ์ฌ์ฉํ ์ ์์ต๋๋ค.",
"example": "# mcp_server.py (FastMCP ์ฌ์ฉ)\nfrom mcp.server.fastmcp import FastMCP\n\nmcp = FastMCP('my-server')\n\n@mcp.tool()\ndef search_docs(query: str) -> str:\n '''๋ฒกํฐ DB์์ ๊ด๋ จ ๋ฌธ์๋ฅผ ๊ฒ์ํฉ๋๋ค.'''\n results = retriever.invoke(query)\n return '\\n'.join(r.page_content for r in results)\n\nmcp.run(transport='stdio')",
"common_mistake": "MCP ์๋ฒ๋ stdio ๋๋ SSE transport ์ค ํ๋๋ฅผ ์ ํํฉ๋๋ค. Claude Desktop์ stdio(๋ก์ปฌ ํ๋ก์ธ์ค), Claude Code๋ SSE(HTTP ์๋ฒ) ๋ฐฉ์์ ์ฃผ๋ก ์๋๋ค. transport ๋ถ์ผ์น ์ ์ฐ๊ฒฐ์ด ์ ๋ฉ๋๋ค.",
"keywords": ["MCP", "Model Context Protocol", "tool", "resource", "FastMCP", "stdio", "Claude", "agent"],
"source_id": "official-docs-mcp",
"details": [
"**3๊ฐ์ง ๊ธฐ๋ณธ ์์**: `tool`(LLM์ด ํธ์ถํ๋ ํจ์), `resource`(LLM์ด ์ฝ๋ ๋ฐ์ดํฐ), `prompt`(์ฌ์ฌ์ฉ ๊ฐ๋ฅํ ํ๋กฌํํธ ํ
ํ๋ฆฟ). ๋๋ถ๋ถ์ ์๋ฒ๋ `tool`๋ง ๊ตฌํํฉ๋๋ค.",
"**Claude Desktop ์ฐ๊ฒฐ**: `~/Library/Application Support/Claude/claude_desktop_config.json`์ ์๋ฒ ๊ฒฝ๋ก์ ๋ช
๋ น์ด๋ฅผ ๋ฑ๋กํ๋ฉด ๋ํ ์ค ์๋์ผ๋ก ๋๊ตฌ๊ฐ ๋ณด์
๋๋ค.",
"**FastMCP**: Anthropic ๊ณต์ Python SDK์ ๊ณ ์์ค ๋ํผ. `@mcp.tool()` ๋ฐ์ฝ๋ ์ดํฐ๋ก ํจ์๋ฅผ ๋๊ตฌ๋ก ๋ฑ๋กํ๋ฉด ์คํค๋ง ์์ฑ, ์
์ถ๋ ฅ ๊ฒ์ฆ์ด ์๋์ผ๋ก ์ฒ๋ฆฌ๋ฉ๋๋ค."
]
},
{
"id": "docs-litellm",
"certification": "Docs Study",
"title": "LiteLLM โ ์ฌ๋ฌ LLM์ ํ๋์ ์ธํฐํ์ด์ค๋ก",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-litellm-1"],
"summary": "LiteLLM์ OpenAI, Anthropic, Ollama, Gemini ๋ฑ 100๊ฐ ์ด์์ LLM ํ๋ก๋ฐ์ด๋๋ฅผ ๋์ผํ API ํ์์ผ๋ก ํธ์ถํฉ๋๋ค. ๋ชจ๋ธ์ ๋ฐ๊ฟ ๋ ์ฝ๋ ์์ ์์ด ๋ชจ๋ธ๋ช
๋ง ๋ณ๊ฒฝํ๋ฉด ๋ฉ๋๋ค.",
"example": "from litellm import completion\n\n# Ollama ๋ก์ปฌ\nresponse = completion(\n model='ollama/llama3.2',\n messages=[{'role': 'user', 'content': '์๋
'}]\n)\n\n# OpenAI โ ์ฝ๋ ๋์ผ, ๋ชจ๋ธ๋ช
๋ง ๋ณ๊ฒฝ\nresponse = completion(\n model='gpt-4o-mini',\n messages=[{'role': 'user', 'content': '์๋
'}]\n)\n\nprint(response.choices[0].message.content)",
"common_mistake": "๋ชจ๋ธ๋ช
ํ์์ `'ํ๋ก๋ฐ์ด๋/๋ชจ๋ธ๋ช
'`์
๋๋ค. Ollama๋ `'ollama/llama3.2'`, Anthropic์ `'anthropic/claude-3-5-sonnet-20241022'`. OpenAI๋ง ์์ธ๋ก ์ ๋์ฌ ์์ด `'gpt-4o-mini'`์ฒ๋ผ ์๋๋ค.",
"keywords": ["LiteLLM", "completion", "model", "provider", "๋ฒค๋ ๋
๋ฆฝ", "OpenAI ํธํ", "ChatLiteLLM"],
"source_id": "official-docs-litellm",
"details": [
"**๋น์ฉ ์ถ์ **: `litellm.success_callback`์ผ๋ก ๊ฐ ํธ์ถ์ ํ ํฐ ์์ ์์ ๋น์ฉ์ ์๋ ๊ธฐ๋กํ ์ ์์ต๋๋ค. ์ฌ๋ฌ ๋ชจ๋ธ์ ์คํํ ๋ ๋น์ฉ ๋น๊ต์ ์ ์ฉํฉ๋๋ค.",
"**LangChain ํตํฉ**: `from langchain_community.chat_models import ChatLiteLLM`์ผ๋ก LangChain ์ฒด์ธ์์ LiteLLM์ ์๋๋ค. LCEL ํ์ดํ๋ผ์ธ์ ๋ฐ๋ก ์ฐ๊ฒฐ๋ฉ๋๋ค.",
"**Fallback**: `completion(..., fallbacks=['gpt-4o-mini', 'ollama/llama3.2'])`์ฒ๋ผ ์ฃผ ๋ชจ๋ธ ์คํจ ์ ์๋์ผ๋ก ๋ฐฑ์
๋ชจ๋ธ๋ก ์ ํํ๋ ๊ธฐ๋ฅ์ ์ง์ํฉ๋๋ค."
]
},
{
"id": "docs-chainlit",
"certification": "Docs Study",
"title": "Chainlit โ ์ฑ๋ด UI ํนํ ํ๋ ์์ํฌ",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-chainlit-1"],
"summary": "Chainlit์ LLM ์ฑ๋ด UI์ ํนํ๋ ํ๋ ์์ํฌ์
๋๋ค. Streamlit๊ณผ ๋ฌ๋ฆฌ ์น์์ผ ๊ธฐ๋ฐ์ผ๋ก ๋์ํด ์คํธ๋ฆฌ๋ฐ ์๋ต, ํ์ผ ์
๋ก๋, ๋ฉํฐ์คํ
์งํ ํ์๊ฐ ์์ฐ์ค๋ฝ์ต๋๋ค.",
"example": "import chainlit as cl\n\n@cl.on_message\nasync def main(message: cl.Message):\n # ์คํธ๋ฆฌ๋ฐ ์๋ต\n msg = cl.Message(content='')\n async for chunk in chain.astream({'question': message.content}):\n await msg.stream_token(chunk)\n await msg.send()\n\n# ์คํ: chainlit run app.py",
"common_mistake": "Chainlit์ `async` ๊ธฐ๋ฐ์
๋๋ค. LangChain์ `chain.invoke()`(๋๊ธฐ)๋ฅผ ์ฐ๋ฉด ์ด๋ฒคํธ ๋ฃจํ๊ฐ ๋ธ๋กํน๋ฉ๋๋ค. ๋ฐ๋์ `chain.ainvoke()` ๋๋ `chain.astream()`์ ์ฌ์ฉํ์ธ์.",
"keywords": ["chainlit", "on_message", "Message", "stream_token", "async", "astream", "on_chat_start"],
"source_id": "official-docs-chainlit",
"details": [
"**๋ฐ์ฝ๋ ์ดํฐ ๊ธฐ๋ฐ**: `@cl.on_chat_start`(์ธ์
์์ ์ ์ฒด์ธ ์ด๊ธฐํ), `@cl.on_message`(๋ฉ์์ง ์์ ), `@cl.on_stop`(์คํธ๋ฆฌ๋ฐ ์ค๋จ)์ผ๋ก ์ด๋ฒคํธ๋ฅผ ์ฒ๋ฆฌํฉ๋๋ค.",
"**Step UI**: `async with cl.Step(name='๊ฒ์ ์ค...'):` ๋ธ๋ก์ผ๋ก ์์ด์ ํธ์ ๊ฐ ๋จ๊ณ(๊ฒ์, ์ถ๋ก ๋ฑ)๋ฅผ UI์ ์ค์๊ฐ์ผ๋ก ํผ์ณ ๋ณด์ฌ์ค๋๋ค. LangGraph ์์ด์ ํธ์ ํจ๊ป ์ฐ๋ฉด ๊ฐ๋ ฅํฉ๋๋ค.",
"**ํ์ผ ์
๋ก๋**: `@cl.on_message`์์ `message.elements`๋ก ์ฒจ๋ถ ํ์ผ์ ๋ฐ์ ์ ์์ต๋๋ค. PDF ์
๋ก๋ โ ์ฆ์ RAG ์ธ๋ฑ์ฑ ํ๋ฆ์ ๋ช ์ค๋ก ๊ตฌํํฉ๋๋ค."
]
},
{
"id": "docs-redis",
"certification": "Docs Study",
"title": "Redis โ ๋ํ ๋ฉ๋ชจ๋ฆฌ ์์ํ & LLM ์บ์ฑ",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-redis-1"],
"summary": "Redis๋ ์ธ๋ฉ๋ชจ๋ฆฌ ๋ฐ์ดํฐ ์ ์ฅ์๋ก LLM ์ฑ์์ ๋ํ ํ์คํ ๋ฆฌ ์์ํ์ LLM ์๋ต ์บ์ฑ์ ์๋๋ค. ์๋ฒ๋ฅผ ์ฌ์์ํด๋ ๋ํ๊ฐ ์ ์ง๋๊ณ , ๋์ผ ์ง๋ฌธ์ ๋ํ LLM ํธ์ถ ๋น์ฉ์ ์ ๊ฐํฉ๋๋ค.",
"example": "from langchain_community.chat_message_histories import RedisChatMessageHistory\nfrom langchain_core.runnables.history import RunnableWithMessageHistory\n\ndef get_session_history(session_id: str):\n return RedisChatMessageHistory(\n session_id=session_id,\n url='redis://localhost:6379'\n )\n\nchain_with_history = RunnableWithMessageHistory(\n chain,\n get_session_history,\n input_messages_key='question',\n history_messages_key='chat_history'\n)",
"common_mistake": "Redis ์๋ฒ๊ฐ ์คํ ์ค์ด์ง ์์ผ๋ฉด ์ฐ๊ฒฐ ์ค๋ฅ๊ฐ ๋ฉ๋๋ค. ๋ก์ปฌ ๊ฐ๋ฐ ์ `docker run -d -p 6379:6379 redis`๋ก ๋จผ์ Redis๋ฅผ ์คํํ์ธ์.",
"keywords": ["Redis", "RedisChatMessageHistory", "session_id", "url", "์์ํ", "์บ์ฑ", "TTL"],
"source_id": "official-docs-redis",
"details": [
"**์ธ๋ฉ๋ชจ๋ฆฌ ๊ต์ฒด**: ๊ธฐ์กด `ChatMessageHistory`(์ฌ์์ ์ ์๋ฉธ)๋ฅผ `RedisChatMessageHistory`๋ก ๊ต์ฒดํ๋ฉด `get_session_history` ํจ์ ํ ์ค๋ง ๋ฐ๊ฟ๋ ์์์ฑ์ ์ป์ต๋๋ค.",
"**TTL**: `RedisChatMessageHistory(session_id, url, ttl=3600)`์ผ๋ก 1์๊ฐ ํ ์๋ ๋ง๋ฃ๋ฉ๋๋ค. ๋ฌดํํ ์์ด๋ ํ์คํ ๋ฆฌ์ ๋ฉ๋ชจ๋ฆฌ ๋์๋ฅผ ๋ฐฉ์งํฉ๋๋ค.",
"**LLM ์๋ต ์บ์ฑ**: `from langchain.cache import RedisCache; langchain.llm_cache = RedisCache(redis_client)`๋ก ๋์ผ ์ง๋ฌธ์ ๋ํ LLM ์๋ต์ ์บ์ํด ๋น์ฉ๊ณผ ์๋ต ์๊ฐ์ ์ค์
๋๋ค."
]
},
{
"id": "docs-unstructured",
"certification": "Docs Study",
"title": "Unstructured โ PDFยทWordยท์ด๋ฏธ์ง ํ์ฑ",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-unstructured-1"],
"summary": "Unstructured๋ PDF, Word, Excel, PowerPoint, ์ด๋ฏธ์ง ๋ฑ ๋น์ ํ ๋ฌธ์๋ฅผ ํ
์คํธ๋ก ๋ณํํฉ๋๋ค. LangChain์ `UnstructuredFileLoader`์ ํตํฉ๋์ด ํ์ฅ์์ ์๊ด์์ด ๋์ผ ์ฝ๋๋ก RAG ํ์ดํ๋ผ์ธ์ ์ฐ๊ฒฐํฉ๋๋ค.",
"example": "from langchain_community.document_loaders import UnstructuredFileLoader\n\n# ๋จ์ผ ํ์ผ โ ํ์ฅ์ ์๋ ๊ฐ์ง\nloader = UnstructuredFileLoader('./report.pdf')\ndocs = loader.load()\n\n# ํผํฉ ํ์ ๋๋ ํ ๋ฆฌ\nfrom langchain_community.document_loaders import DirectoryLoader\nloader = DirectoryLoader(\n './docs',\n glob='**/*',\n loader_cls=UnstructuredFileLoader\n)\ndocs = loader.load()",
"common_mistake": "`pip install unstructured`๋ ๊ธฐ๋ณธ ์ค์น์
๋๋ค. PDF์๋ `pip install unstructured[pdf]`, ์ด๋ฏธ์ง OCR์๋ `pip install unstructured[local-inference]`์ฒ๋ผ ๋ณ๋ extra๋ฅผ ์ค์นํด์ผ ํฉ๋๋ค.",
"keywords": ["Unstructured", "UnstructuredFileLoader", "PDF", "Word", "OCR", "๋น์ ํ ๋ฌธ์", "elements"],
"source_id": "official-docs-unstructured",
"details": [
"**์๋ ํ์ผ ๊ฐ์ง**: ํ์ฅ์์ ๋ฐ๋ผ ์๋์ผ๋ก ์ ์ ํ ํ์๋ฅผ ์ ํํฉ๋๋ค. `.pdf`, `.docx`, `.pptx`, `.jpg`, `.eml` ๋ฑ์ ๋์ผ ์ฝ๋๋ก ์ฒ๋ฆฌํฉ๋๋ค.",
"**์์ ๋ถ๋ฅ**: `mode='elements'`๋ก ๋ก๋ํ๋ฉด Title, NarrativeText, Table, Image ๋ฑ ์์ ํ์
์ด ๋ฉํ๋ฐ์ดํฐ์ ์ ์ฅ๋ฉ๋๋ค. ํค๋๋ฅผ ์ฒญํฌ ๋ฉํ๋ฐ์ดํฐ๋ก ํ์ฉํ๊ฑฐ๋ ํ
์ด๋ธ๋ง ๋ฐ๋ก ์ถ์ถํ ๋ ์๋๋ค.",
"**API vs ๋ก์ปฌ**: Unstructured Cloud API๋ ๊ณ ํ์ง OCR๊ณผ ๋ ์ด์์ ๋ถ์์ ์ ๊ณตํฉ๋๋ค. ๋ก์ปฌ์ ๋ฌด๋ฃ์ด๋ ์ด๋ฏธ์ง๋ ๋ณต์กํ ๋ ์ด์์์ PDF ํ์ง์ด ๋ฎ์ ์ ์์ต๋๋ค."
]
},
{
"id": "docs-conversation-memory",
"certification": "Docs Study",
"title": "LangChain ๋ํ ๋ฉ๋ชจ๋ฆฌ โ ๋ฉํฐํด ์ฑ๋ด",
"level": "๊ณ ๊ธ",
"related_practices": ["docs-lab-6", "docs-brain-1"],
"summary": "RunnableWithMessageHistory๋ LangChain ์ฒด์ธ์ ๋ํ ํ์คํ ๋ฆฌ๋ฅผ ์๋์ผ๋ก ์ฃผ์
ํฉ๋๋ค. ๋งค ์์ฒญ๋ง๋ค ์ด์ ๋ํ๋ฅผ ์ง์ ๊ด๋ฆฌํ์ง ์์๋ session_id๋ง ๋๊ธฐ๋ฉด ํ์คํ ๋ฆฌ๊ฐ ์ ์ง๋ฉ๋๋ค.",
"example": "from langchain_community.chat_message_histories import ChatMessageHistory\nfrom langchain_core.runnables.history import RunnableWithMessageHistory\n\nstore = {} # session_id โ ChatMessageHistory\n\ndef get_session_history(session_id: str):\n if session_id not in store:\n store[session_id] = ChatMessageHistory()\n return store[session_id]\n\nchain_with_history = RunnableWithMessageHistory(\n chain,\n get_session_history,\n input_messages_key=\"question\",\n history_messages_key=\"chat_history\"\n)\n\nchain_with_history.invoke(\n {\"question\": \"์๋
\"},\n config={\"configurable\": {\"session_id\": \"user-1\"}}\n)",
"common_mistake": "`session_id`๋ฅผ ๋ชจ๋ ์ฌ์ฉ์์๊ฒ ๋์ผํ๊ฒ ๋๊ธฐ๋ฉด ์ ์ฒด ์ฌ์ฉ์๊ฐ ํ๋์ ๋ํ ํ์คํ ๋ฆฌ๋ฅผ ๊ณต์ ํฉ๋๋ค. ์ฌ์ฉ์๋ณ๋ก ๊ณ ์ ํ session_id๋ฅผ ์์ฑํด์ผ ํฉ๋๋ค.",
"keywords": ["RunnableWithMessageHistory", "ChatMessageHistory", "session_id", "chat_history", "๋ฉํฐํด"],
"source_id": "official-docs-langchain",
"details": [
"**ํ์คํ ๋ฆฌ ์ ์ฅ์**: ์ธ๋ฉ๋ชจ๋ฆฌ(`ChatMessageHistory`)๋ ์๋ฒ ์ฌ์์ ์ ์ฌ๋ผ์ง๋๋ค. ํ๋ก๋์
์์๋ `RedisChatMessageHistory`, `SQLChatMessageHistory` ๋ฑ ์์ ์ ์ฅ์๋ฅผ ์๋๋ค.",
"**ํ์คํ ๋ฆฌ ๊ธธ์ด ์ ํ**: ๋ํ๊ฐ ๊ธธ์ด์ง๋ฉด ํ ํฐ์ด ์ด๊ณผ๋ฉ๋๋ค. `trim_messages()` ๋๋ ์ต๊ทผ N๊ฐ ๋ฉ์์ง๋ง ์ ์งํ๋ ๋ก์ง์ ์ถ๊ฐํด์ผ ํฉ๋๋ค.",
"**ํ๋กฌํํธ ์ฐ๋**: `ChatPromptTemplate`์ `MessagesPlaceholder(variable_name=\"chat_history\")`๋ฅผ ์ถ๊ฐํด์ผ ํ์คํ ๋ฆฌ๊ฐ ํ๋กฌํํธ์ ์ฃผ์
๋ฉ๋๋ค. history_messages_key์ placeholder ์ด๋ฆ์ด ์ผ์นํด์ผ ํฉ๋๋ค."
]
}
]
|