Skip to content

About

Pirogramming GitHub Code RAG workshop starter

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

ย 

History

1 Commit

Folders and files

NameName
Last commit message
Last commit date
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 

Repository files navigation

ํ”ผ๋กœ๊ทธ๋ž˜๋ฐ GitHub Code RAG ์ฑ—๋ด‡ ์‹ค์Šต

๋ฒ„์ „: v1.0.0 ยท ๊ถŒ์žฅ Python: 3.13

GitHub ์ €์žฅ์†Œ์˜ ์ฝ”๋“œ๋ฅผ ๊ฒ€์ƒ‰ ๊ฐ€๋Šฅํ•œ ์ง€์‹์œผ๋กœ ๋งŒ๋“ค๊ณ , ๊ฒ€์ƒ‰๋œ ์ฝ”๋“œ๋งŒ ๊ทผ๊ฑฐ๋กœ ๋‹ต๋ณ€ํ•˜๋Š” Django ์ฑ—๋ด‡ ์‹ค์Šต์ž…๋‹ˆ๋‹ค. ๋ณต์žกํ•œ GitHub ์ˆ˜์ง‘ยท๋ณด์•ˆยทChroma ๊ต์ฒด ๋กœ์ง์€ ์™„์„ฑ๋œ ์ƒํƒœ๋กœ ์ œ๊ณต๋˜๋ฉฐ, ์‹ค์Šต์—์„œ๋Š” ์ฑ„ํŒ… ๋ฉ”์‹œ์ง€๊ฐ€ ๋‹ต๋ณ€๊ณผ ์ถœ์ฒ˜๋กœ ๋Œ์•„์˜ค๊ธฐ๊นŒ์ง€์˜ ์—ฌ์„ฏ ํ•จ์ˆ˜๋ฅผ ์™„์„ฑํ•ฉ๋‹ˆ๋‹ค.

๋ธŒ๋ผ์šฐ์ € ๋ฉ”์‹œ์ง€
  โ†’ Django POST API
  โ†’ ๋Œ€ํ™” ์„œ๋น„์Šค
  โ†’ Vector DB ๊ฒ€์ƒ‰
  โ†’ ๊ฒ€์ƒ‰ ๊ฒฐ๊ณผ๋ฅผ Context๋กœ ๋ณ€ํ™˜
  โ†’ ์ตœ๊ทผ ๋Œ€ํ™” + Context + ํ˜„์žฌ ์งˆ๋ฌธ์œผ๋กœ Prompt ์กฐ๋ฆฝ
  โ†’ Upstage LLM ํ˜ธ์ถœ
  โ†’ ๊ฒ€์ƒ‰ Metadata์—์„œ Citation ์ƒ์„ฑ
  โ†’ JSON ์‘๋‹ต๊ณผ Evidence ํ‘œ์‹œ

RAG๊ฐ€ ๋ฌด์—‡์ธ๊ฐ€์š”?

RAG(Retrieval-Augmented Generation)๋Š” LLM์ด ๋‹ต๋ณ€ํ•˜๊ธฐ ์ „์— ๊ด€๋ จ ์ž๋ฃŒ๋ฅผ ๊ฒ€์ƒ‰ํ•˜๊ณ , ๊ฒ€์ƒ‰ ๊ฒฐ๊ณผ๋ฅผ ์งˆ๋ฌธ๊ณผ ํ•จ๊ป˜ ์ „๋‹ฌํ•ด ๊ทผ๊ฑฐ ์žˆ๋Š” ๋‹ต๋ณ€์„ ์ƒ์„ฑํ•˜๊ฒŒ ํ•˜๋Š” ํŒŒ์ดํ”„๋ผ์ธ์ž…๋‹ˆ๋‹ค.

  • Chunk: ๊ธด ์ฝ”๋“œ๋‚˜ ๋ฌธ์„œ๋ฅผ ๊ฒ€์ƒ‰ํ•˜๊ธฐ ์ข‹์€ ํฌ๊ธฐ๋กœ ๋‚˜๋ˆˆ ์กฐ๊ฐ
  • Embedding: ์ฝ”๋“œ์™€ ์งˆ๋ฌธ์˜ ์˜๋ฏธ๋ฅผ ์ˆซ์ž ๋ฒกํ„ฐ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๊ณผ์ •
  • Chroma: ๋ฒกํ„ฐ, ์›๋ฌธ Chunk, ํŒŒ์ผ ๊ฒฝ๋กœ์™€ ์ค„ ๋ฒˆํ˜ธ๋ฅผ ํ•จ๊ป˜ ์ €์žฅํ•˜๋Š” Vector DB
  • Retrieval: ์งˆ๋ฌธ๊ณผ ์˜๋ฏธ๊ฐ€ ๊ฐ€๊นŒ์šด ์ฝ”๋“œ Chunk ๊ฒ€์ƒ‰
  • Augmentation: ๊ฒ€์ƒ‰ ๊ฒฐ๊ณผ๋ฅผ LLM์ด ์ฝ์„ Context๋กœ ์กฐ๋ฆฝ
  • Generation: Context๋ฅผ ๊ทผ๊ฑฐ๋กœ ๋‹ต๋ณ€ ์ƒ์„ฑ

์ด ํ”„๋กœ์ ํŠธ์˜ ๋ฐ์ดํ„ฐ ํ๋ฆ„์€ ๋‹ค์Œ๊ณผ ๊ฐ™์Šต๋‹ˆ๋‹ค.

GitHub URL โ†’ ์ฝ”๋“œ ์ˆ˜์ง‘ โ†’ Chunk โ†’ Embedding โ†’ Chroma
           โ†’ ๊ฒ€์ƒ‰ โ†’ Context โ†’ Prompt โ†’ LLM ๋‹ต๋ณ€ + Citation

RAG๋Š” ์ƒˆ๋กœ์šด ๋ชจ๋ธ ํ•˜๋‚˜๊ฐ€ ์•„๋‹ˆ๋ผ ๊ฒ€์ƒ‰, Context ๊ตฌ์„ฑ, Prompt ๊ตฌ์„ฑ, ์ƒ์„ฑ, ์ถœ์ฒ˜ ๋ฐ˜ํ™˜์„ ์—ฐ๊ฒฐํ•œ ์ฒ˜๋ฆฌ ๊ตฌ์กฐ์ž…๋‹ˆ๋‹ค.

ํ”„๋กœ์ ํŠธ ๊ตฌ์กฐ

pirogramming-portfolio-rag-workshop/
โ”œโ”€โ”€ config/                              Django ํ”„๋กœ์ ํŠธ ์„ค์ •
โ”œโ”€โ”€ portfolio/
โ”‚   โ”œโ”€โ”€ management/commands/
โ”‚   โ”‚   โ”œโ”€โ”€ ask_llm.py                   RAG ์—†๋Š” LLM ํ˜ธ์ถœ ๋น„๊ต
โ”‚   โ”‚   โ”œโ”€โ”€ workshop_check.py            API Key์™€ ์‹คํ–‰ ํ™˜๊ฒฝ ์ ๊ฒ€
โ”‚   โ”‚   โ””โ”€โ”€ mobile_serve.py              ์„ ํƒ ๊ธฐ๋Šฅ: Tailscale ์—ฐ๊ฒฐ
โ”‚   โ”œโ”€โ”€ services/
โ”‚   โ”‚   โ”œโ”€โ”€ conversation.py              ๋Œ€ํ™” ์ƒํƒœ์™€ ๊ฒ€์ƒ‰ ์—ฐ๊ฒฐ
โ”‚   โ”‚   โ”œโ”€โ”€ rag.py                       Context, Prompt, Citation, ๋‹ต๋ณ€
โ”‚   โ”‚   โ”œโ”€โ”€ vector_store.py              Chroma ์ €์žฅ๊ณผ ๊ฒ€์ƒ‰
โ”‚   โ”‚   โ”œโ”€โ”€ code_documents.py            ์ฝ”๋“œ Chunk์™€ ์ค„ Metadata
โ”‚   โ”‚   โ”œโ”€โ”€ repository_url.py            GitHub URL ๊ฒ€์ฆ
โ”‚   โ”‚   โ”œโ”€โ”€ source_policy.py              ์ˆ˜์ง‘ํ•  ํŒŒ์ผ ์„ ํƒ
โ”‚   โ”‚   โ”œโ”€โ”€ github_client.py             GitHub API ์š”์ฒญ๊ณผ ์žฌ์‹œ๋„
โ”‚   โ”‚   โ””โ”€โ”€ jd_analyzer.py               JD ์ฝ”๋“œ ๊ทผ๊ฑฐ ๋ถ„์„
โ”‚   โ”œโ”€โ”€ templates/portfolio/index.html   ์ฑ—๋ด‡ ํ™”๋ฉด
โ”‚   โ”œโ”€โ”€ static/portfolio/                CSS์™€ JavaScript
โ”‚   โ””โ”€โ”€ views.py                         HTTP ์š”์ฒญ๊ณผ JSON ์‘๋‹ต
โ”œโ”€โ”€ fixtures/                            GitHub ์žฅ์•  ์‹œ ์ƒ˜ํ”Œ ๋ฐ์ดํ„ฐ
โ”œโ”€โ”€ tests/                               ๋‹จ๊ณ„๋ณ„ ์ž๋™ ํ…Œ์ŠคํŠธ
โ”œโ”€โ”€ .env.example                         ํ™˜๊ฒฝ๋ณ€์ˆ˜ ์˜ˆ์‹œ
โ”œโ”€โ”€ manage.py
โ””โ”€โ”€ pyproject.toml                       Python๊ณผ ํŒจํ‚ค์ง€ ์„ค์ •

์ง์ ‘ ์ž‘์„ฑํ•˜๋Š” ์—ฌ์„ฏ ํ•จ์ˆ˜

Starter์—๋Š” ํ•จ์ˆ˜ ์„ ์–ธ, ํƒ€์ž… ํžŒํŠธ, import๊ฐ€ ์ด๋ฏธ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ฐ ๋Œ€์ƒ ํ•จ์ˆ˜์˜ raise WorkshopStepIncomplete(...)๋ฅผ ์™„์„ฑ ์ฝ”๋“œ๋กœ ๊ต์ฒดํ•ฉ๋‹ˆ๋‹ค.

Step ํŒŒ์ผ ํ•จ์ˆ˜ ํ•ต์‹ฌ ๊ฐœ๋…
1 portfolio/views.py messages ๋ธŒ๋ผ์šฐ์ € POST โ†’ Form ๊ฒ€์ฆ โ†’ ์„œ๋น„์Šค โ†’ JSON
2 portfolio/services/conversation.py _answer_code_question ์งˆ๋ฌธ โ†’ Top 6 ๊ฒ€์ƒ‰ โ†’ RAG ๋‹ต๋ณ€
3 portfolio/services/rag.py format_context ๊ฒ€์ƒ‰ Chunk โ†’ LLM Context
4 portfolio/services/rag.py _answer_prompt ์ตœ๊ทผ ๋Œ€ํ™” + Context + ์งˆ๋ฌธ
5 portfolio/services/rag.py _citations ๊ฒ€์ƒ‰ Metadata ๊ธฐ๋ฐ˜ ์ถœ์ฒ˜์™€ ์ค‘๋ณต ์ œ๊ฑฐ
6 portfolio/services/rag.py answer_with_context ๊ทผ๊ฑฐ ํ™•์ธ โ†’ LLM ํ˜ธ์ถœ โ†’ ๋‹ต๋ณ€ยท์ถœ์ฒ˜ ๋ฐ˜ํ™˜

๋‹ค์Œ ๋กœ์ง์€ ์™„์„ฑ๋œ ์ฝ”๋“œ๋กœ ์ œ๊ณต๋ฉ๋‹ˆ๋‹ค.

  • GitHub URL ๊ฒ€์ฆ: parse_github_repository_url
  • ํŒŒ์ผ ์ •์ฑ…๊ณผ ์˜ˆ์‚ฐ: select_source_candidates
  • ์ฝ”๋“œ Chunk์™€ ์ค„ ๋ฒˆํ˜ธ: chunk_source_file, build_code_documents
  • Chroma ๊ต์ฒด์™€ ๊ฒ€์ƒ‰: reset_and_index, search
  • JD ์ฝ”๋“œ ๊ทผ๊ฑฐ ๋ถ„์„: analyze_jd
  • GitHub ์žฌ์‹œ๋„, ๋น„๊ณต๊ฐœ ์ €์žฅ์†Œ ํ™•์ธ, ์›์ž์  index ๊ต์ฒด, ํ™”๋ฉด ๋ Œ๋”๋ง

์ค€๋น„๋ฌผ

  • Python 3.11 ์ด์ƒ 3.14 ๋ฏธ๋งŒ โ€” Python 3.13 ๊ถŒ์žฅ
  • Git
  • GitHub ๊ณ„์ •๊ณผ Fine-grained personal access token
  • Upstage API Key
  • macOS, Linux ๋˜๋Š” Windows PowerShell

Python 3.14์—์„œ๋Š” ํ˜„์žฌ ๊ณ ์ •๋œ tokenizers๊ฐ€ ์‚ฌ์šฉํ•˜๋Š” PyO3 ๋ฒ„์ „์ด Python 3.13๊นŒ์ง€๋งŒ ์ง€์›ํ•ด ์„ค์น˜ ๊ณผ์ •์—์„œ wheel build๊ฐ€ ์‹คํŒจํ•ฉ๋‹ˆ๋‹ค. Rust๋ฅผ ๋‹ค์‹œ ์„ค์น˜ํ•˜๊ธฐ๋ณด๋‹ค Python 3.13 ๊ฐ€์ƒํ™˜๊ฒฝ์„ ์‚ฌ์šฉํ•˜์„ธ์š”.

API Key ์ค€๋น„

GitHub Fine-grained Token

์•ฑ์ด ๋ถ„์„ ๋Œ€์ƒ ์ €์žฅ์†Œ๋ฅผ ์ฝ์„ ๋•Œ ์‚ฌ์šฉํ•˜๋Š” ์ธ์ฆ ๊ฐ’์ž…๋‹ˆ๋‹ค.

  1. GitHub Fine-grained token ์ƒ์„ฑ ํ™”๋ฉด์„ ์—ฝ๋‹ˆ๋‹ค.
  2. ์ˆ˜์—…์— ํ•„์š”ํ•œ ์งง์€ ๋งŒ๋ฃŒ ๊ธฐ๊ฐ„์„ ์„ค์ •ํ•ฉ๋‹ˆ๋‹ค.
  3. ๋น„๊ณต๊ฐœ ์ €์žฅ์†Œ๋ฅผ ์‚ฌ์šฉํ•  ๊ฒฝ์šฐ Only select repositories๋กœ ํ•„์š”ํ•œ ์ €์žฅ์†Œ๋งŒ ์„ ํƒํ•ฉ๋‹ˆ๋‹ค.
  4. Repository permissions์—์„œ Contents: Read-only๋ฅผ ์„ ํƒํ•ฉ๋‹ˆ๋‹ค.
  5. ์ƒ์„ฑ ์งํ›„ ๊ฐ’์„ ๋ณต์‚ฌํ•ด .env์˜ GITHUB_TOKEN์—๋งŒ ๋„ฃ์Šต๋‹ˆ๋‹ค.

์“ฐ๊ธฐ ๊ถŒํ•œ, Issues, Pull requests, Actions ๊ถŒํ•œ์€ ํ•„์š”ํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.

Upstage API Key

Chat ๋ชจ๋ธ๊ณผ Embedding ๋ชจ๋ธ์„ ํ˜ธ์ถœํ•  ๋•Œ ์‚ฌ์šฉํ•˜๋Š” ์ธ์ฆ ๊ฐ’์ž…๋‹ˆ๋‹ค.

  1. Upstage Console API Keys์— ๋กœ๊ทธ์ธํ•ฉ๋‹ˆ๋‹ค.
  2. Create new key๋ฅผ ๋ˆŒ๋Ÿฌ Key๋ฅผ ๋ฐœ๊ธ‰ํ•ฉ๋‹ˆ๋‹ค.
  3. ๋ฐœ๊ธ‰ ์งํ›„ ํ•œ ๋ฒˆ๋งŒ ๋ณต์‚ฌํ•ด .env์˜ UPSTAGE_API_KEY์— ๋„ฃ์Šต๋‹ˆ๋‹ค.

์ด ์‹ค์Šต์€ ํ•œ๊ตญ์–ด ์ฝ”๋“œ ์งˆ๋ฌธ๊ณผ ์„ค๋ช…์„ ํ•œ provider์—์„œ ์ฒ˜๋ฆฌํ•˜๊ณ , langchain-upstage๋กœ Chat๊ณผ Embedding ์—ฐ๊ฒฐ์„ ๋‹จ์ˆœํ•˜๊ฒŒ ์œ ์ง€ํ•˜๊ธฐ ์œ„ํ•ด Upstage๋ฅผ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค. ๊ณ„์ • credit๊ณผ rate limit์ด ์žˆ์œผ๋ฏ€๋กœ ์ˆ˜์—… ์ „ workshop_check๋กœ ์‹ค์ œ ํ˜ธ์ถœ ๊ฐ€๋Šฅ ์—ฌ๋ถ€๋ฅผ ํ™•์ธํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.

์„ค์น˜

์ €์žฅ์†Œ๋ฅผ ๋‚ด๋ ค๋ฐ›์Šต๋‹ˆ๋‹ค.

git clone https://github.com/inoaole/pirogramming-portfolio-rag-workshop.git
cd pirogramming-portfolio-rag-workshop

macOS/Linux

python3.13 -m venv .venv
source .venv/bin/activate
python -m pip install -e '.[dev]'
cp .env.example .env

python3.13 ๋ช…๋ น์ด ์—†๋‹ค๋ฉด ์„ค์น˜๋œ Python 3.11~3.13์˜ ๋ช…๋ น์„ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.

Windows PowerShell

py -3.13 -m venv .venv
.venv\Scripts\Activate.ps1
python -m pip install -e ".[dev]"
Copy-Item .env.example .env

PowerShell์ด ๊ฐ€์ƒํ™˜๊ฒฝ ์‹คํ–‰์„ ๋ง‰์œผ๋ฉด ์‚ฌ์šฉ์ž ๊ณ„์ • ๋ฒ”์œ„์—์„œ ํ•œ ๋ฒˆ ์‹คํ–‰ํ•ฉ๋‹ˆ๋‹ค.

Set-ExecutionPolicy -Scope CurrentUser RemoteSigned

.env ์„ค์ •

.env ํŒŒ์ผ์— ๋ณธ์ธ์ด ๋ฐœ๊ธ‰ํ•œ ๊ฐ’์„ ์ž…๋ ฅํ•ฉ๋‹ˆ๋‹ค.

GITHUB_TOKEN=๋ณธ์ธ์˜_fine_grained_token
USE_SAMPLE_GITHUB=false
UPSTAGE_API_KEY=๋ณธ์ธ์˜_upstage_api_key
UPSTAGE_CHAT_MODEL=solar-pro3
UPSTAGE_EMBEDDING_MODEL=solar-embedding-1-large
CHROMA_PATH=.workshop/chroma

Token๊ณผ API Key๋ฅผ README, ๋ฉ”์‹ ์ €, ๋ธŒ๋ผ์šฐ์ € ์ž…๋ ฅ์ฐฝ, ํ™”๋ฉด ์บก์ฒ˜ ๋˜๋Š” Git commit์— ์˜ฌ๋ฆฌ์ง€ ์•Š์Šต๋‹ˆ๋‹ค. .env๋Š” Git์—์„œ ์ œ์™ธ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.

API ์—†์ด ๋กœ์ปฌ ์„ค์น˜๋ถ€ํ„ฐ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.

python manage.py workshop_check --offline

์‹ค์ œ Token๊ณผ Key๊นŒ์ง€ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.

python manage.py workshop_check

Ready for workshop์ด ํ‘œ์‹œ๋˜๋ฉด ์ค€๋น„๊ฐ€ ๋๋‚ฌ์Šต๋‹ˆ๋‹ค.

์‹ค์Šต ์ง„ํ–‰

๊ฐ Step์—์„œ ์ง€์ •๋œ ํ•จ์ˆ˜๋งŒ ์ˆ˜์ •ํ•˜๊ณ  ํ•ด๋‹น ํ…Œ์ŠคํŠธ๋ฅผ ํ†ต๊ณผ์‹œํ‚จ ๋’ค ๋‹ค์Œ ๋‹จ๊ณ„๋กœ ์ด๋™ํ•ฉ๋‹ˆ๋‹ค. ์•„๋ž˜ ๋ช…๋ น์€ ํ™œ์„ฑํ™”๋œ ๊ฐ€์ƒํ™˜๊ฒฝ์˜ Python์„ ์‚ฌ์šฉํ•˜๋„๋ก python -m pytest ํ˜•ํƒœ๋กœ ์‹คํ–‰ํ•ฉ๋‹ˆ๋‹ค.

Step 1 โ€” ๋ธŒ๋ผ์šฐ์ € ๋ฉ”์‹œ์ง€๋ฅผ ์„œ๋ฒ„๋กœ ์ „๋‹ฌํ•˜๊ธฐ

  • ์œ„์น˜: portfolio/views.py์˜ messages ํ•จ์ˆ˜ ์ „์ฒด
  • ์—ญํ• : ์ž…๋ ฅ ๊ฒ€์ฆ, session ์ „๋‹ฌ, ์„œ๋น„์Šค ํ˜ธ์ถœ, JSON ์‘๋‹ต
python -m pytest tests/test_step1_messages.py -q

Step 2 โ€” ์งˆ๋ฌธ์„ ๊ฒ€์ƒ‰๊ณผ RAG ๋‹ต๋ณ€์— ์—ฐ๊ฒฐํ•˜๊ธฐ

  • ์œ„์น˜: portfolio/services/conversation.py์˜ ConversationService._answer_code_question ๋ฉ”์„œ๋“œ ์ „์ฒด
  • ์ฃผ์˜: ํด๋ž˜์Šค ๋‚ด๋ถ€ ๋ฉ”์„œ๋“œ์ด๋ฏ€๋กœ def ์•ž์— 4์นธ์„ ๋“ค์—ฌ์”๋‹ˆ๋‹ค.
  • ์—ญํ• : Top 6 ๊ฒ€์ƒ‰, answer_with_context ํ˜ธ์ถœ, ConversationReply ๋ฐ˜ํ™˜
python -m pytest tests/test_step2_conversation.py -q

Step 3 โ€” ๊ฒ€์ƒ‰ ๊ฒฐ๊ณผ๋ฅผ Context๋กœ ๋งŒ๋“ค๊ธฐ

  • ์œ„์น˜: portfolio/services/rag.py์˜ format_context ํ•จ์ˆ˜ ์ „์ฒด
  • ์—ญํ• : repository, path, lines, URL, content๋ฅผ <portfolio_context>๋กœ ์กฐ๋ฆฝ
python -m pytest tests/test_step3_context.py -q

Step 4 โ€” ์ตœ๊ทผ ๋Œ€ํ™”์™€ ์งˆ๋ฌธ์„ Prompt๋กœ ์กฐ๋ฆฝํ•˜๊ธฐ

  • ์œ„์น˜: portfolio/services/rag.py์˜ _answer_prompt ํ•จ์ˆ˜ ์ „์ฒด
  • ์—ญํ• : ์ตœ๊ทผ ๋Œ€ํ™” 6๊ฐœ, ๊ฒ€์ƒ‰ Context, ํ˜„์žฌ ์งˆ๋ฌธ์„ ์ผ์ •ํ•œ ์ˆœ์„œ๋กœ ๊ฒฐํ•ฉ
python -m pytest tests/test_step4_prompt.py -q

Step 5 โ€” ๊ฒ€์ƒ‰ Metadata์—์„œ ์ถœ์ฒ˜ ๋งŒ๋“ค๊ธฐ

  • ์œ„์น˜: portfolio/services/rag.py์˜ _citations ํ•จ์ˆ˜ ์ „์ฒด
  • ์—ญํ• : ๋นˆ URL ์ œ์™ธ, ์ค‘๋ณต ์ œ๊ฑฐ, relevance ๋ณด์กด
python -m pytest tests/test_step5_citations.py -q

Step 6 โ€” ๊ทผ๊ฑฐ ๊ธฐ๋ฐ˜ ๋‹ต๋ณ€ ๋ฐ˜ํ™˜ํ•˜๊ธฐ

  • ์œ„์น˜: portfolio/services/rag.py์˜ answer_with_context ํ•จ์ˆ˜ ์ „์ฒด
  • ์—ญํ• : ๊ฒ€์ƒ‰ ๊ฒฐ๊ณผ๊ฐ€ ์—†์œผ๋ฉด ๋ชจ๋ธ์„ ํ˜ธ์ถœํ•˜์ง€ ์•Š๊ณ , ์žˆ์œผ๋ฉด Upstage ๋‹ต๋ณ€๊ณผ Citation์„ ๋ฐ˜ํ™˜
python -m pytest tests/test_step6_answer.py -q

์ฑ—๋ด‡ ์‹คํ–‰

์ฒซ ๋ฒˆ์งธ ํ„ฐ๋ฏธ๋„์—์„œ ์„œ๋ฒ„๋ฅผ ์‹คํ–‰ํ•ฉ๋‹ˆ๋‹ค.

python manage.py runserver 127.0.0.1:8000

๋ธŒ๋ผ์šฐ์ €์—์„œ http://127.0.0.1:8000์„ ์—ฝ๋‹ˆ๋‹ค.

  1. https://github.com/owner/repository ํ˜•์‹์˜ ์ •ํ™•ํ•œ ์ €์žฅ์†Œ URL์„ ๋ณด๋ƒ…๋‹ˆ๋‹ค.
  2. ๋น„๊ณต๊ฐœ ์ €์žฅ์†Œ๋ผ๋ฉด Upstage ์ „์†ก ์•ˆ๋‚ด๋ฅผ ์ฝ๊ณ  ๋™์˜ ๋˜๋Š” ์ทจ์†Œํ•ฉ๋‹ˆ๋‹ค.
  3. ์ธ๋ฑ์‹ฑ์ด ๋๋‚˜๋ฉด ์ฝ”๋“œ ๊ตฌ์กฐ๋‚˜ ํ•จ์ˆ˜ ํ˜ธ์ถœ ํ๋ฆ„์„ ์งˆ๋ฌธํ•ฉ๋‹ˆ๋‹ค.
  4. Evidence์—์„œ ํŒŒ์ผ ๊ฒฝ๋กœ, ์ค„ ๋ฒ”์œ„, relevance์™€ GitHub ๋งํฌ๋ฅผ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.
  5. JD ์ ํ•ฉ๋„ ๋ถ„์„ ๊ธฐ๋Šฅ์€ ์™„์„ฑ๋œ ํ™•์žฅ ๊ธฐ๋Šฅ์œผ๋กœ ์‹คํ–‰ํ•ด ์ฝ”๋“œ ๊ทผ๊ฑฐ๋ฅผ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.

์ „์ฒด ๊ฒ€์ฆ

์—ฌ์„ฏ Step์„ ๋ชจ๋‘ ์™„๋ฃŒํ•œ ๋’ค ์‹คํ–‰ํ•ฉ๋‹ˆ๋‹ค.

python -m pytest \
  tests/test_step1_messages.py \
  tests/test_step2_conversation.py \
  tests/test_step3_context.py \
  tests/test_step4_prompt.py \
  tests/test_step5_citations.py \
  tests/test_step6_answer.py -q
ruff check .
python manage.py check

Git clone์—์„œ ์ž‘์—…ํ–ˆ๋‹ค๋ฉด ๋‹ค์Œ๋„ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.

git diff --check
git status --short

๋ฌธ์ œ ํ•ด๊ฒฐ

Building wheel for tokenizers ์˜ค๋ฅ˜

์˜ค๋ฅ˜์— configured Python interpreter version (3.14) is newer than PyO3's maximum supported version (3.13)๊ฐ€ ๋ณด์ด๋ฉด Python 3.14 ๊ฐ€์ƒํ™˜๊ฒฝ์ž…๋‹ˆ๋‹ค. .venv๋ฅผ ์‚ญ์ œํ•˜๊ณ  Python 3.13์œผ๋กœ ๋‹ค์‹œ ๋งŒ๋“ญ๋‹ˆ๋‹ค.

WorkshopStepIncomplete ์˜ค๋ฅ˜

์˜ค๋ฅ˜์— ํ‘œ์‹œ๋œ Step ํ•จ์ˆ˜๊ฐ€ ์•„์ง ๋น„์–ด ์žˆ๋‹ค๋Š” ๋œป์ž…๋‹ˆ๋‹ค. ํ•ด๋‹น ํ•จ์ˆ˜์˜ body๋งŒ ์™„์„ฑํ•˜๊ณ  ๊ทธ Step ํ…Œ์ŠคํŠธ๋ฅผ ๋‹ค์‹œ ์‹คํ–‰ํ•ฉ๋‹ˆ๋‹ค. import, helper, test๋ฅผ ๊ณ ์ณ ์šฐํšŒํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.

GitHub 401ยท403

Fine-grained Token์˜ ๋งŒ๋ฃŒ ์—ฌ๋ถ€, repository ์„ ํƒ, Contents: Read-only ๊ถŒํ•œ์„ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.

Upstage ์ธ์ฆ ์˜ค๋ฅ˜

.env์˜ UPSTAGE_API_KEY, UPSTAGE_CHAT_MODEL, UPSTAGE_EMBEDDING_MODEL ๊ฐ’์„ ํ™•์ธํ•˜๊ณ  python manage.py workshop_check๋ฅผ ๋‹ค์‹œ ์‹คํ–‰ํ•ฉ๋‹ˆ๋‹ค.

8000๋ฒˆ ํฌํŠธ ์ถฉ๋Œ

๊ธฐ์กด Django ์„œ๋ฒ„๋ฅผ ์‹คํ–‰ํ•œ ํ„ฐ๋ฏธ๋„์—์„œ Ctrl+C๋ฅผ ๋ˆ„๋ฆ…๋‹ˆ๋‹ค.

macOS/Linux:

lsof -nP -iTCP:8000 -sTCP:LISTEN

Windows PowerShell:

Get-NetTCPConnection -LocalPort 8000 -State Listen -ErrorAction SilentlyContinue

๋ณด์•ˆ๊ณผ ํ•œ๊ณ„

  • GitHub Token๊ณผ Upstage API Key๋Š” .env์—๋งŒ ์ €์žฅํ•˜๊ณ  ๊ณต์œ ํ•˜๊ฑฐ๋‚˜ commitํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
  • ์•ฑ์€ GitHub ์ €์žฅ์†Œ์— ์“ฐ์ง€ ์•Š์œผ๋ฉฐ Contents: Read-only ๊ถŒํ•œ๋งŒ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.
  • ๋น„๊ณต๊ฐœ ์ €์žฅ์†Œ๋Š” ๋™์˜ ํ›„ ์„ ํƒ๋œ ์ฝ”๋“œ Chunk๊ฐ€ Upstage Embedding ์š”์ฒญ์œผ๋กœ ์ „์†ก๋˜๊ณ , ๊ฒ€์ƒ‰๋œ Chunk๊ฐ€ Chat ์š”์ฒญ์œผ๋กœ ์ „์†ก๋ฉ๋‹ˆ๋‹ค.
  • Chroma ์›๋ฌธ๊ณผ ๋ฒกํ„ฐ๋Š” ๋กœ์ปฌ .workshop/chroma/์— ์ €์žฅ๋ฉ๋‹ˆ๋‹ค.
  • ํŒŒ์ผ ํ˜•์‹, ํฌ๊ธฐ, ์ „์ฒด ์šฉ๋Ÿ‰, Chunk ๊ฐœ์ˆ˜ ์ œํ•œ ๋•Œ๋ฌธ์— ๋ชจ๋“  ํŒŒ์ผ์ด ์ธ๋ฑ์‹ฑ๋œ๋‹ค๊ณ  ๋ณด์žฅํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
  • RAG ๋‹ต๋ณ€์€ ๊ฒ€์ƒ‰ ๊ฒฐ๊ณผ๋ฅผ ์š”์•ฝํ•œ ๊ฒƒ์ด๋ฏ€๋กœ ์ค‘์š”ํ•œ ํŒ๋‹จ์€ Evidence์˜ ์›๋ณธ ํŒŒ์ผ๊ณผ ์ค„ ๋ฒˆํ˜ธ๋กœ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.
  • JD ๊ฒฐ๊ณผ๋Š” ๊ณต๊ฐœ ์ฝ”๋“œ์—์„œ ํ™•์ธ๋˜๋Š” ๊ธฐ์ˆ  ๊ทผ๊ฑฐ ๋น„์œจ์ด๋ฉฐ ๊ฒฝ๋ ฅ, ํ˜‘์—… ๋Šฅ๋ ฅ ๋˜๋Š” ์‹ค์ œ ์ฑ„์šฉ ๊ฒฐ๊ณผ๋ฅผ ์˜ˆ์ธกํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.

About

Pirogramming GitHub Code RAG workshop starter

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages