|
| 1 | +from __future__ import annotations |
| 2 | + |
| 3 | +from dataclasses import dataclass |
| 4 | + |
| 5 | +import structlog |
| 6 | + |
| 7 | +from ai_server.analyzer._embedding_step import ( |
| 8 | + EmbeddingStepError, |
| 9 | + chunk_embed_and_upsert, |
| 10 | +) |
| 11 | +from ai_server.analyzer.sources.github_repo import ( |
| 12 | + GitHubRepoSourceExtractor, |
| 13 | + RepositoryFetchError, |
| 14 | +) |
| 15 | +from ai_server.chain.document_analysis_chain import DocumentAnalyzer |
| 16 | +from ai_server.core.client import CoreClient, CoreTokenError |
| 17 | +from ai_server.rag.chunker import MarkdownChunker |
| 18 | +from ai_server.rag.embedder import EmbeddingProvider |
| 19 | +from ai_server.storage.base import ObjectStorage |
| 20 | + |
| 21 | +log = structlog.get_logger(__name__) |
| 22 | + |
| 23 | + |
| 24 | +class RepositoryAnalyzeError(Exception): |
| 25 | + def __init__(self, *, code: str, message: str, retriable: bool) -> None: |
| 26 | + super().__init__(message) |
| 27 | + self.code = code |
| 28 | + self.message = message |
| 29 | + self.retriable = retriable |
| 30 | + |
| 31 | + |
| 32 | +@dataclass(frozen=True) |
| 33 | +class RepositoryAnalysisResult: |
| 34 | + summary: str |
| 35 | + tech_stack: list[str] |
| 36 | + document_path: str |
| 37 | + embedding_chunk_count: int |
| 38 | + |
| 39 | + |
| 40 | +# Core에서 사용자별 GitHub token 수령 → 레포 fetch → LLM 분석 → 마크다운 저장 → 청킹·임베딩 |
| 41 | +class RepositoryAnalyzer: |
| 42 | + def __init__( |
| 43 | + self, |
| 44 | + *, |
| 45 | + extractor: GitHubRepoSourceExtractor, |
| 46 | + core_client: CoreClient, |
| 47 | + chain: DocumentAnalyzer, |
| 48 | + storage: ObjectStorage, |
| 49 | + chunker: MarkdownChunker, |
| 50 | + embedder: EmbeddingProvider, |
| 51 | + analyzed_key_template: str, |
| 52 | + ) -> None: |
| 53 | + self._extractor = extractor |
| 54 | + self._core_client = core_client |
| 55 | + self._chain = chain |
| 56 | + self._storage = storage |
| 57 | + self._chunker = chunker |
| 58 | + self._embedder = embedder |
| 59 | + self._analyzed_key_template = analyzed_key_template |
| 60 | + |
| 61 | + async def analyze( |
| 62 | + self, |
| 63 | + *, |
| 64 | + repository_id: int, |
| 65 | + repo_full_name: str, |
| 66 | + default_branch: str = "main", |
| 67 | + user_id: int | None, |
| 68 | + analyzed_document_id: int, |
| 69 | + ) -> RepositoryAnalysisResult: |
| 70 | + if user_id is None: |
| 71 | + raise RepositoryAnalyzeError( |
| 72 | + code="MISSING_USER_ID", |
| 73 | + message="envelope.context.userId 없이는 GitHub token을 가져올 수 없음", |
| 74 | + retriable=False, |
| 75 | + ) |
| 76 | + |
| 77 | + log.info( |
| 78 | + "repository.token.fetch", |
| 79 | + user_id=user_id, |
| 80 | + repository_id=repository_id, |
| 81 | + ) |
| 82 | + try: |
| 83 | + access_token = await self._core_client.fetch_github_token(user_id) |
| 84 | + except CoreTokenError as err: |
| 85 | + raise RepositoryAnalyzeError( |
| 86 | + code=err.code, message=err.message, retriable=err.retriable |
| 87 | + ) from err |
| 88 | + |
| 89 | + log.info( |
| 90 | + "repository.extract.start", |
| 91 | + repository_id=repository_id, |
| 92 | + repo_full_name=repo_full_name, |
| 93 | + default_branch=default_branch, |
| 94 | + ) |
| 95 | + try: |
| 96 | + source = await self._extractor.extract( |
| 97 | + repo_full_name, |
| 98 | + access_token=access_token, |
| 99 | + ) |
| 100 | + except RepositoryFetchError as err: |
| 101 | + raise RepositoryAnalyzeError( |
| 102 | + code=err.code, message=err.message, retriable=err.retriable |
| 103 | + ) from err |
| 104 | + |
| 105 | + if not source.text.strip(): |
| 106 | + raise RepositoryAnalyzeError( |
| 107 | + code="EMPTY_REPO_CONTENT", |
| 108 | + message="레포에서 추출된 텍스트가 비어 있음", |
| 109 | + retriable=False, |
| 110 | + ) |
| 111 | + |
| 112 | + log.info( |
| 113 | + "repository.llm.start", |
| 114 | + repository_id=repository_id, |
| 115 | + text_chars=len(source.text), |
| 116 | + ) |
| 117 | + analysis = await self._chain.analyze( |
| 118 | + text=source.text, |
| 119 | + source_type=source.source_type, |
| 120 | + ) |
| 121 | + |
| 122 | + out_key = self._analyzed_key_template.format(repository_id=repository_id) |
| 123 | + await self._storage.put_text(out_key, analysis.markdown) |
| 124 | + log.info( |
| 125 | + "repository.markdown.saved", |
| 126 | + repository_id=repository_id, |
| 127 | + key=out_key, |
| 128 | + md_chars=len(analysis.markdown), |
| 129 | + ) |
| 130 | + |
| 131 | + try: |
| 132 | + chunk_count = await chunk_embed_and_upsert( |
| 133 | + document_id=analyzed_document_id, |
| 134 | + markdown=analysis.markdown, |
| 135 | + chunker=self._chunker, |
| 136 | + embedder=self._embedder, |
| 137 | + core_client=self._core_client, |
| 138 | + log_prefix="repository", |
| 139 | + ) |
| 140 | + except EmbeddingStepError as err: |
| 141 | + raise RepositoryAnalyzeError( |
| 142 | + code=err.code, message=err.message, retriable=err.retriable |
| 143 | + ) from err |
| 144 | + |
| 145 | + return RepositoryAnalysisResult( |
| 146 | + summary=analysis.summary, |
| 147 | + tech_stack=list(analysis.tech_stack), |
| 148 | + document_path=out_key, |
| 149 | + embedding_chunk_count=chunk_count, |
| 150 | + ) |
0 commit comments