USE_EMBEDDED_DB=true still connects to PostgreSQL if previously configured

[mikrub]

Bug Description

Setting USE_EMBEDDED_DB="true" does not switch Khoj to use SQLite when PostgreSQL was previously configured/connected.

Steps to Reproduce

1. Install Khoj with PostgreSQL backend configured
2. Run: USE_EMBEDDED_DB="true" khoj --anonymous-mode
3. Khoj still attempts to connect to PostgreSQL instead of using embedded SQLite

Expected Behavior

When USE_EMBEDDED_DB="true" is set, Khoj should use the embedded SQLite database regardless of any existing PostgreSQL configuration.

Actual Behavior

Khoj ignores the environment variable and continues to use PostgreSQL, leading to errors if the PostgreSQL schema has issues.

Environment

• Khoj version: 1.42.10
• Python: 3.12.12
• OS: macOS (Darwin 25.2.0)

Additional Context

This was discovered while trying to work around issue #1244 (migration mismatch). Setting USE_EMBEDDED_DB="true" was attempted as a workaround but the env var was not respected.

The documentation at https://docs.khoj.dev/get-started/setup/ suggests this should work:

USE_EMBEDDED_DB="true" khoj --anonymous-mode

[debanjum]

I don't see any issues in the code logic. How did you install Khoj locally? Via docker or via pip? You need to run at least pip install khoj[local] to have it install pglite to use with USE_EMBEDDED_DB="true".

Setting USE_EMBEDDED_DB flag to true will make Khoj use pglite (an embedded postgres db instead of a requiring a running postgres server)

[yang1002378395-cmyk]

分析

查看代码后发现 USE_EMBEDDED_DB 的实际行为：

# settings.py:127-175
USE_EMBEDDED_DB = is_env_var_true("USE_EMBEDDED_DB")

if USE_EMBEDDED_DB:
    import pgserver
    # 启动嵌入式 PostgreSQL（不是 SQLite）
    PGSERVER_INSTANCE = pgserver.get_server(PGSERVER_DATA_DIR)

关键点：

• USE_EMBEDDED_DB=true 启动的是嵌入式 PostgreSQL（pgserver）
• Khoj 不支持 SQLite 作为主数据库
• 所有数据库配置始终使用 PostgreSQL driver

这可能是文档误导问题。文档说 "embedded database" 让人以为是 SQLite，但实际上是嵌入式 PostgreSQL。

解决方案：

1. 如果你想用全新的数据库，需要清除之前的 PostgreSQL 数据或使用新的 POSTGRES_DB 名称
2. 或者设置 PGSERVER_DATA_DIR 到新目录让 pgserver 创建新实例

需要我改进文档说明吗？