feat: migrate doc-service to use storage-service for file I/O (Phase 2)

- storage.py: replace aiofiles filesystem ops with httpx calls to
  storage-service PUT/GET/DELETE /objects/documents/{key}
- Document model: rename file_path → storage_key (plain object key, no path prefix)
- Migration 0008: ALTER COLUMN + data migration strips /data/documents/ prefix
- documents.py: update upload, delete, download endpoints; _extract_pdf_text
  now takes bytes (pdfplumber.open(BytesIO)) instead of a filesystem path
- file_watcher.py: store storage_key instead of file_path on ingestion
- doc-service config: add STORAGE_SERVICE_URL env var

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

features/doc-service/app/routers/documents.py

@@ -1,10 +1,10 @@
 import asyncio
+import io
 import json
 import math
 import uuid
 from datetime import datetime, timezone
 
-import aiofiles
 import pdfplumber
 from fastapi import APIRouter, BackgroundTasks, Depends, HTTPException, Query, UploadFile
 from fastapi.responses import StreamingResponse
@@ -29,7 +29,7 @@ from app.schemas.document import (
 from app.schemas.share import DocumentShareCreate, DocumentShareOut, SharedDocumentOut
 from app.services.ai_client import AIServiceError, classify_document
 from app.services.config_reader import load_doc_config
-from app.services.storage import delete_file, get_upload_path, save_upload
+from app.services.storage import delete_file, download_file, save_upload
 
 router = APIRouter()
 
@@ -118,10 +118,10 @@ def _doc_with_categories(
     )
 
 
-def _extract_pdf_text(file_path: str) -> str:
+def _extract_pdf_text(pdf_bytes: bytes) -> str:
     """Synchronous — must be called via asyncio.to_thread."""
     text_parts = []
-    with pdfplumber.open(file_path) as pdf:
+    with pdfplumber.open(io.BytesIO(pdf_bytes)) as pdf:
         for page in pdf.pages:
             page_text = page.extract_text()
             if page_text:
@@ -146,7 +146,8 @@ async def process_document(doc_id: str) -> None:
         await db.commit()
 
         try:
-            text = await asyncio.to_thread(_extract_pdf_text, doc.file_path)
+            pdf_bytes = await download_file(doc.storage_key)
+            text = await asyncio.to_thread(_extract_pdf_text, pdf_bytes)
             result = await classify_document(text)
 
             doc.raw_text = text[:500_000]  # cap stored text at 500k chars
@@ -187,13 +188,13 @@ async def upload_document(
         )
 
     doc_id = str(uuid.uuid4())
-    dest = await save_upload(file_data, user_id, doc_id)
+    storage_key = await save_upload(file_data, user_id, doc_id)
 
     doc = Document(
         id=doc_id,
         user_id=user_id,
         filename=file.filename or "upload.pdf",
-        file_path=str(dest),
+        storage_key=storage_key,
         file_size=len(file_data),
         status="pending",
     )
@@ -578,7 +579,7 @@ async def delete_document(
         if not can_delete_via_share and not can_delete_as_group_admin:
             raise HTTPException(status_code=403, detail="Not allowed to delete this document")
 
-    delete_file(doc.file_path)
+    await delete_file(doc.storage_key)
     await db.delete(doc)
     await db.commit()
 
@@ -609,13 +610,13 @@ async def download_file(
     if doc is None:
         raise HTTPException(status_code=404, detail="Document not found")
 
-    async def file_generator():
-        async with aiofiles.open(doc.file_path, "rb") as f:
-            while chunk := await f.read(64 * 1024):
-                yield chunk
+    try:
+        pdf_bytes = await download_file(doc.storage_key)
+    except FileNotFoundError:
+        raise HTTPException(status_code=404, detail="File not found in storage")
 
     return StreamingResponse(
-        file_generator(),
+        iter([pdf_bytes]),
         media_type="application/pdf",
         headers={"Content-Disposition": f'inline; filename="{doc.filename}"'},
     )