Mokhtar
feat(production): complete phase 1-4 refactor with hybrid retrieval CAG architecture, E5 dense embeddings, grounding validator, precomputed cold-start index, and 24/24 passing system tests
d54db1a | # knowledge/models.py — Core data structures for knowledge management | |
| import hashlib | |
| from pydantic import BaseModel, Field | |
| class DocumentMetadata(BaseModel): | |
| doc_id: str | |
| filename: str | |
| path: str | |
| title: str = "Untitled Document" | |
| section: str = "" | |
| subsection: str = "" | |
| language: str = "ar" | |
| source_type: str = "markdown" | |
| class Document(BaseModel): | |
| metadata: DocumentMetadata | |
| content: str | |
| class Chunk(BaseModel): | |
| chunk_id: str | |
| doc_id: str | |
| filename: str | |
| path: str | |
| title: str | |
| section: str = "" | |
| subsection: str = "" | |
| text: str | |
| language: str = "ar" | |
| source_type: str = "markdown" | |
| token_count: int = 0 | |
| heading_context: str = "" | |
| chunk_index: int = 0 | |
| def formatted_for_context(self) -> str: | |
| """Format chunk with source metadata for LLM context payload.""" | |
| header_parts = [] | |
| if self.filename: | |
| header_parts.append(f"المستند: {self.filename}") | |
| if self.title and self.title != self.filename: | |
| header_parts.append(f"العنوان: {self.title}") | |
| if self.section: | |
| header_parts.append(f"القسم: {self.section}") | |
| if self.subsection: | |
| header_parts.append(f"الفرع: {self.subsection}") | |
| meta_header = " | ".join(header_parts) | |
| return f"--- [{meta_header}] ---\n{self.text.strip()}\n" | |