Proposal: Consolidate all data paths under storage_root for better containerization

[tanbro]

Proposal: Consolidate All Data Paths Under storage_root

Problem

Currently, different data files are scattered across different locations, making containerized deployment complex:

Component	Current Default	Location
SQLite DB	~/.xagent/xagent.db	✅ Under storage_root
LanceDB	${pwd}/data/lancedb	❌ Separate path
Web uploads	src/xagent/web/uploads	❌ In source tree
Web static	src/xagent/web/static	❌ In source tree

This inconsistency creates problems:

• Multiple bind-mounts needed for containers
• Data mixed with source code (uploads, static)
• Unpredictable paths depending on working directory
• Complex backup/migration - data scattered everywhere

Current Examples

SQLite (Good):

storage_root = get_storage_root()  # ~/.xagent
return storage_root / "xagent.db"  # ~/.xagent/xagent.db ✅

LanceDB (Bad - cwd-relative):

# Returns "data/lancedb" - relative to cwd! ❌
return Path("data/lancedb")

Web uploads (Bad - in source tree):

web_dir = get_web_dir()  # src/xagent/web by default
return web_dir / "uploads"  # default: src/xagent/web/uploads ❌

Proposed Solution

Move all data paths under storage_root (default: ~/.xagent):

Component	New Default	Environment Variable
SQLite DB	~/.xagent/xagent.db	DATABASE_URL (unchanged)
LanceDB	~/.xagent/lancedb	LANCEDB_PATH (unchanged)
Web uploads	~/.xagent/uploads	XAGENT_UPLOADS_DIR (unchanged)
Web static	~/.xagent/static	XAGENT_WEB_STATIC_DIR (unchanged)

Benefits:

• ✅ Single bind-mount for containers: XAGENT_STORAGE_ROOT=/data
• ✅ Data separated from code - clean architecture
• ✅ Consistent - all data in one place
• ✅ Predictable - no cwd-relative paths

Related Issues

• Builds on: bug: uploads directory path is inconsistent across modules - uses relative paths instead of unified UPLOADS_DIR #243 (unified configuration module)
• Environment variables will still work as overrides