fix: resolve upgrade DB issue, production startup failure, add debug/stop commands

[MoYoez]

• I have read the checklist above (put an x in the brackets)

Description

Before

• Database Problem

First time I tried using "npm install -g @innei/kagura" to install the latest version of kagura, found that I cannot interact with slack bot, so I pull the main branch and built a version and linked to the latest version.

The problem is here, looks like the program itself will not migrate.

• Start Problem

Looks like the program just exit and didnt run in the background,

After

Also I added a simple command prefix (--debug), to check whever the program work correctly.

Linked Issues

none

AI collaboration (required; say if you used none)

1. Workflow

I used AI to read the relevant startup, database, and CLI code paths, and run local verification.

I reviewed and adjusted the behavior through several iterations, especially around migration timing, debug output, daemon startup, and stop semantics.

2. Prompt (or equivalent task description)

• Investigate why database table changes were not applied after a new build/release

• Check what happened when using "Kagura" and application not response.

3. Reasoning and trade-offs

My local machine is Windows, and I cannot fully run the normal Slack workflow there, so I focused on inspecting the program behavior and adding better diagnostics.

I was not sure what happened after running the kagura command, so I added a --debug mode to print startup diagnostics such as config paths, environment presence, database path, and migration path. During that investigation, I found that the CLI could return after startup instead of keeping the Slack Socket Mode process alive, so I changed the normal startup path to launch Kagura as a background daemon.

To make local testing easier, and to avoid manually killing Node processes from Task Manager, I also added a kagura stop command.

While debugging startup, I also found that upgraded versions did not automatically apply database migrations before using newer tables or columns. I fixed that by running packaged Drizzle migrations during database initialization.

Additional context

btw I found that the when using "pnpm dev" after following the readme, looks like the .env didnt read correctly?

working platform is Windows 11

[safedep]

SafeDep Report Summary

No dependency changes detected. Nothing to scan.

View complete scan results →

_{This report is generated by SafeDep Github App}

[Loongphy]

Same here.

 kagura
◇ injecting env (4) from .config/kagura/.env // tip: ⌘ multiple files { path: ['.env.local', '.env'] }
◇ injecting env (0) from .config/kagura/.env // tip: ⌘ enable debugging { debug: true }
◇ injecting env (0) from .env // tip: ◈ secrets for agents [www.dotenvx.com]
ℹ Memory reconciler disabled by config; expired-only prune via startup hook                     bootstrap 4:27:01 PM
[bootstrap:claude:session 4:27:01 PM] ℹ Claude Code runtime config at executor startup: CLAUDE_MODEL=(unset) ANTHROPIC_MODEL=(unset) ANTHROPIC_DEFAULT_SONNET_MODEL=(unset) ANTHROPIC_DEFAULT_HAIKU_MODEL=(unset) ANTHROPIC_DEFAULT_OPUS_MODEL=(unset) ANTHROPIC_BASE_URL=(unset)
[bootstrap:codex:session 4:27:01 PM] ℹ Codex CLI provider configured: model=(config default) reasoning=(config default) sandbox=danger-full-access
[bootstrap:pi-agent:session 4:27:01 PM] ℹ Pi Agent provider configured: command=pi args=["-p","--mode","json"] model=(unknown)
[bootstrap:slack:commands 4:27:01 PM] ℹ Registered 7 slash commands: /usage, /workspace, /memory, /session, /version, /provider, /model
[bootstrap:claude:session 4:27:02 PM] ℹ Claude Code auth status at executor startup: loggedIn=true authMethod=oauth_token apiProvider=firstParty subscriptionType=(unknown) email=(unknown) orgName=(unknown)
ℹ Slack Socket Mode application started.                                                        bootstrap 4:27:02 PM
⚠ Memory reconciler tick failed: no such table: memories                      bootstrap:memory-reconciler 4:27:02 PM

 ERROR  Failed to recover interrupted agent executions: no such table: agent_executions          bootstrap 4:27:02 PM

cc @Innei can we merge it?