kofdai's picture
Deploy NullAI Knowledge System to Spaces
075a2b6 verified
"""
Knowledge Base API
知識タイル一覧表示と検証マーク機構
"""
from fastapi import APIRouter, Depends, HTTPException, Query, UploadFile, File
from fastapi.responses import FileResponse, StreamingResponse
from typing import List, Optional
from datetime import datetime
import os
import json
import io
from sqlalchemy.orm import Session
import logging
from uuid import uuid4
from backend.app.middleware.auth import get_current_user_optional, get_current_user, User
from backend.app.config import settings
from backend.app.database.session import get_db
from backend.app.services.knowledge_service import KnowledgeService, get_knowledge_service
from backend.app.schemas.knowledge import KnowledgeTile, KnowledgeListResponse, KnowledgeDetailResponse, EditRequest, VerificationMark
from backend.app.utils.iath_encoder import IathEncoder
logger = logging.getLogger(__name__)
router = APIRouter()
@router.get("/", response_model=KnowledgeListResponse)
async def list_knowledge_tiles(
domain_id: Optional[str] = Query(None, description="ドメインでフィルタ"),
verification_type: Optional[str] = Query(None, description="検証タイプでフィルタ"),
search: Optional[str] = Query(None, description="検索クエリ"),
page: int = Query(1, ge=1, description="ページ番号"),
page_size: int = Query(20, ge=1, le=100, description="ページサイズ"),
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
tiles_orm, total_count = service.list_tiles(
db=db, page=page, page_size=page_size,
domain_id=domain_id, verification_type=verification_type, search=search
)
tiles_pydantic = [KnowledgeTile.from_orm(t) for t in tiles_orm]
return KnowledgeListResponse(
tiles=tiles_pydantic,
total_count=total_count,
page=page,
page_size=page_size,
has_more=(page * page_size) < total_count
)
@router.get("/{tile_id}", response_model=KnowledgeDetailResponse)
async def get_knowledge_tile(
tile_id: str,
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
tile_orm = service.get_tile(db, tile_id=tile_id)
if not tile_orm:
raise HTTPException(status_code=404, detail="Knowledge tile not found")
tile_pydantic = KnowledgeTile.from_orm(tile_orm)
return KnowledgeDetailResponse(
tile=tile_pydantic,
full_content=tile_orm.content,
# TODO: Implement sources, related_tiles, and edit_history from DB
sources=[],
related_tiles=[],
edit_history=[]
)
@router.put("/{tile_id}", response_model=KnowledgeTile)
async def update_knowledge_tile(
tile_id: str,
request: EditRequest,
current_user: User = Depends(get_current_user),
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
updated_tile_orm = service.update_tile(
db=db, tile_id=tile_id, content=request.content, user=current_user
)
if not updated_tile_orm:
raise HTTPException(status_code=404, detail="Knowledge tile not found")
return KnowledgeTile.from_orm(updated_tile_orm)
@router.get("/coordinates")
async def get_coordinates_for_3d_visualization(
domain_id: Optional[str] = Query(None, description="特定ドメインのみ取得"),
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
"""
3D可視化用の座標データを取得
座標を持つタイルのみを返し、必要最小限の情報のみを含める
"""
# Fetch all tiles (large page size to get all)
tiles_orm, _ = service.list_tiles(db=db, page_size=10000, domain_id=domain_id)
# Filter tiles that have coordinates and extract minimal data
coordinates_data = []
for tile in tiles_orm:
if tile.coordinates: # Only include tiles with coordinates
coordinates_data.append({
"tile_id": tile.id,
"topic": tile.topic,
"domain_id": tile.domain_id,
"coordinates": tile.coordinates, # [x, y, z, c, g, v]
"confidence_score": tile.confidence_score,
"verification_type": tile.verification_type
})
return {
"tiles": coordinates_data,
"count": len(coordinates_data),
"domain_id": domain_id or "all"
}
@router.get("/export/json")
async def export_db_json(
domain_id: Optional[str] = Query(None, description="特定ドメインのみエクスポート"),
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
# Fetch all tiles for export
tiles_orm, _ = service.list_tiles(db=db, page_size=10000, domain_id=domain_id) # A large page size to get all
tiles_pydantic = [KnowledgeTile.from_orm(t).dict() for t in tiles_orm]
export_data = {
"metadata": {
"export_date": datetime.now().isoformat(),
"source": "NullAI Knowledge Base",
"domain_filter": domain_id or "all",
"tile_count": len(tiles_pydantic)
},
"tiles": tiles_pydantic
}
json_str = json.dumps(export_data, indent=2, ensure_ascii=False, default=str)
return StreamingResponse(
io.BytesIO(json_str.encode('utf-8')),
media_type="application/json",
headers={
"Content-Disposition": f"attachment; filename=null_ai_knowledge_{datetime.now().strftime('%Y%m%d')}.json"
}
)
@router.get("/export/iath")
async def export_db_iath(
domain_id: Optional[str] = Query(None, description="特定ドメインのみエクスポート"),
precision: str = Query("standard", description="精度レベル: standard または high_precision"),
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
"""
.iath形式でエクスポート (dendritic-memory-editor互換)
JSON version 2形式で出力します
precision:
- standard: List Viewからの標準エクスポート
- high_precision: 3D Space Viewからの高精度エクスポート
"""
logger.info(f"Exporting to .iath format (domain={domain_id}, precision={precision})")
# Fetch all tiles for export
tiles_orm, _ = service.list_tiles(db=db, page_size=10000, domain_id=domain_id)
# Map domain_id to domain name
domain_name_map = {
"medical": "Medical",
"ai_fundamentals": "AI Fundamentals",
"logic_reasoning": "Logic & Reasoning",
"computer_science_theory": "Computer Science",
"engineering": "Engineering",
"philosophy": "Philosophy",
"law": "Law"
}
domain_name = domain_name_map.get(domain_id, "General") if domain_id else "General"
# Map domain_id to domain code
domain_code_map = {
"medical": 1,
"ai_fundamentals": 2,
"logic_reasoning": 3,
"computer_science_theory": 4,
"engineering": 5,
"philosophy": 6,
"law": 7
}
domain_code = domain_code_map.get(domain_id, 1) if domain_id else 1
# Convert tiles to .iath JSON format (version 2)
iath_tiles = []
for tile in tiles_orm:
# Parse coordinates
coordinates = tile.coordinates
if not coordinates or len(coordinates) < 3:
# Generate basic coordinates if missing
confidence = tile.confidence_score if tile.confidence_score else 0.5
coordinates = [
50.0 + (hash(tile.id) % 100) / 2, # x: pseudo-random positioning
50.0 + (hash(tile.id[::-1]) % 100) / 2, # y: pseudo-random positioning
10.0 + confidence * 20 # z: based on confidence
]
logger.warning(f"Tile {tile.id} missing coordinates, generated: {coordinates}")
# Extract x, y, z from coordinates
x = round(coordinates[0], 2) if len(coordinates) > 0 else 50.0
y = round(coordinates[1], 2) if len(coordinates) > 1 else 50.0
z = round(coordinates[2], 2) if len(coordinates) > 2 else 10.0
# Map verification types to dendritic-memory-editor compatible values
verification_map = {
"ai": "pending_review",
"community": "community",
"expert": "expert_verified",
"multi_expert": "expert_verified",
"none": "pending_review"
}
verification_status = verification_map.get(tile.verification_type, "pending_review")
# Map author marks to dendritic-memory-editor compatible values
author_mark_map = {
"ai": "community",
"community": "community",
"expert": "expert",
"multi_expert": "expert",
"none": "community"
}
author_mark = author_mark_map.get(tile.verification_type, "community")
# Generate author_id if not present (dendritic-memory-editor expects this field)
author_id = tile.author_id if hasattr(tile, 'author_id') and tile.author_id else "system"
# Build .iath compatible tile structure (version 2 format)
iath_tile = {
"id": tile.id,
"title": tile.topic,
"coordinates": {
"x": x,
"y": y,
"z": z
},
"content": tile.content,
"verification_status": verification_status,
"created_at": tile.created_at.strftime("%Y-%m-%d %H:%M:%S") if tile.created_at else datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
"updated_at": tile.updated_at.strftime("%Y-%m-%d %H:%M:%S") if tile.updated_at else datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
"author_id": author_id,
"author_mark": author_mark
}
iath_tiles.append(iath_tile)
# Create version 2 JSON format
export_data = {
"version": 2,
"header": {
"domain_code": domain_code,
"tile_count": len(iath_tiles),
"created_at": datetime.now().isoformat() + "Z",
"domain": domain_name
},
"tiles": iath_tiles
}
# Convert to JSON
json_str = json.dumps(export_data, ensure_ascii=False, indent=None)
logger.info(f"Successfully created .iath JSON with {len(iath_tiles)} tiles")
# Generate filename
domain_suffix = f"_{domain_name.replace(' ', '_')}" if domain_id else ""
filename = f"{domain_name.replace(' ', '_')}_tiles_{int(datetime.now().timestamp() * 1000)}.iath"
return StreamingResponse(
io.BytesIO(json_str.encode('utf-8')),
media_type="application/json",
headers={
"Content-Disposition": f"attachment; filename={filename}"
}
)
@router.patch("/{tile_id}/domain")
async def update_tile_domain(
tile_id: str,
domain_id: str = Query(..., description="新しいドメインID"),
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
"""
知識タイルのドメインを更新 (List Viewからのドメイン編集機能)
"""
tile_orm = service.get_tile(db, tile_id=tile_id)
if not tile_orm:
raise HTTPException(status_code=404, detail="Knowledge tile not found")
# Update domain
tile_orm.domain_id = domain_id
tile_orm.updated_at = datetime.utcnow()
db.commit()
db.refresh(tile_orm)
logger.info(f"Updated tile {tile_id} domain to {domain_id}")
return KnowledgeTile.from_orm(tile_orm)
@router.post("/import/iath")
async def import_iath_file(
file: UploadFile = File(...),
current_user: User = Depends(get_current_user),
db: Session = Depends(get_db),
service: KnowledgeService = Depends(get_knowledge_service)
):
"""
.iath ファイルをインポートしてデータベースに保存
- JSON version 2 形式の .iath ファイルをサポート
- 既存のタイルIDがある場合は更新、ない場合は新規作成
- エラーが発生したタイルは個別にログに記録
"""
try:
# ファイル内容を読み込む
content = await file.read()
iath_data = json.loads(content.decode('utf-8'))
# .iath フォーマットを検証
if not iath_data.get("version") or not iath_data.get("header") or not iath_data.get("tiles"):
raise HTTPException(
status_code=400,
detail="Invalid .iath file format. Required fields: version, header, tiles"
)
if not isinstance(iath_data["tiles"], list):
raise HTTPException(
status_code=400,
detail="Invalid .iath file format. 'tiles' must be an array"
)
# ヘッダー情報を取得
header = iath_data["header"]
domain_name = header.get("domain", "General")
# ドメイン名からドメインIDにマッピング
domain_name_to_id = {
"Medical": "medical",
"AI Fundamentals": "ai_fundamentals",
"Logic & Reasoning": "logic_reasoning",
"Computer Science": "computer_science_theory",
"Engineering": "engineering",
"Philosophy": "philosophy",
"Law": "law",
"General": "general"
}
domain_id = domain_name_to_id.get(domain_name, "general")
imported_count = 0
updated_count = 0
error_count = 0
errors = []
logger.info(f"Starting import of {len(iath_data['tiles'])} tiles from domain: {domain_name}")
# 各タイルをインポート
for tile_data in iath_data["tiles"]:
try:
tile_id = tile_data.get("id")
if not tile_id:
tile_id = str(uuid4())
# コンテンツを取得
content_text = tile_data.get("content", "")
topic = tile_data.get("title", "")
if not content_text and not topic:
error_msg = f"Tile {tile_id}: No content or title found"
logger.warning(error_msg)
errors.append(error_msg)
error_count += 1
continue
# 座標を取得
coords = tile_data.get("coordinates", {})
coordinates = [
coords.get("x", 50.0),
coords.get("y", 50.0),
coords.get("z", 50.0)
]
# verification_status を verification_type にマッピング
verification_map = {
"pending_review": "ai",
"community": "community",
"expert_verified": "expert",
"multi_expert": "multi_expert"
}
verification_status = tile_data.get("verification_status", "pending_review")
verification_type = verification_map.get(verification_status, "ai")
# author_mark を検証
author_mark = tile_data.get("author_mark", "community")
if author_mark == "expert":
verification_type = "expert"
# 既存のタイルをチェック
existing_tile = service.get_tile(db, tile_id=tile_id)
if existing_tile:
# 既存タイルを更新
existing_tile.domain_id = domain_id
existing_tile.topic = topic or existing_tile.topic
existing_tile.content = content_text
existing_tile.coordinates = coordinates
existing_tile.verification_type = verification_type
existing_tile.updated_at = datetime.utcnow()
db.commit()
db.refresh(existing_tile)
updated_count += 1
logger.info(f"Updated tile {tile_id}")
else:
# 新規タイルを作成
from backend.app.database.models import KnowledgeTileModel
new_tile = KnowledgeTileModel(
id=tile_id,
domain_id=domain_id,
topic=topic or "Imported",
content=content_text,
coordinates=coordinates,
verification_type=verification_type,
confidence_score=0.8,
author_id=current_user.id if current_user else None,
created_at=datetime.utcnow(),
updated_at=datetime.utcnow()
)
db.add(new_tile)
db.commit()
db.refresh(new_tile)
imported_count += 1
logger.info(f"Imported new tile {tile_id}")
except Exception as tile_error:
error_msg = f"Tile {tile_data.get('id', 'unknown')}: {str(tile_error)}"
logger.error(error_msg, exc_info=True)
errors.append(error_msg)
error_count += 1
# 次のタイルに続行
continue
return {
"success": True,
"domain": domain_name,
"total_tiles": len(iath_data["tiles"]),
"imported": imported_count,
"updated": updated_count,
"errors": error_count,
"error_details": errors if errors else None,
"message": f"Successfully processed {imported_count + updated_count} tiles from {domain_name}"
}
except json.JSONDecodeError as e:
logger.error(f"JSON decode error: {e}", exc_info=True)
raise HTTPException(
status_code=400,
detail=f"Invalid JSON format: {str(e)}"
)
except Exception as e:
logger.error(f"Import error: {e}", exc_info=True)
raise HTTPException(
status_code=500,
detail=f"Failed to import .iath file: {str(e)}"
)