Codex CLI 0.155.0 Brings Memory V2 and Better Error Handling

Maya

You know that feeling when you are three weeks into building a complex feature with an AI agent, and suddenly it just, um, completely forgets a core constraint you gave it on day one?

Ethan Park

Oh, totally. It gets buried under megabytes of subagent chatter, terminal outputs, and intermediate reasoning steps, right? The human instruction just gets washed out.

Maya

Exactly. That exact problem is what OpenAI is tackling with Codex CLI version 0.155.0. And hey, before we dive into the code, quick shoutout to Jellypod for sponsoring the episode and helping us bring you this daily breakdown.

Ethan Park

Yeah, huge thanks to Jellypod. So, Maya, this zero point one fifty five point zero release introduces Memory V2, which is a total redesign of how Codex handles long term state.

Maya

Right, and it starts with complete storage isolation. If you set memory version equals v2 in your config file, or pass the environment variable CODEX MEMORY VERSION set to v2, it spins up a completely separate database under your home directory in dot codex slash memory slash v2 slash.

Ethan Park

Exactly. It leaves your legacy V1 context completely untouched. But the really interesting part isn't just where it saves things. It is the algorithmic shift in how extraction actually works.

Maya

Wait, how so?

Ethan Park

So, in V1, context chunking treated almost all text equally. But in Memory V2, the extraction engine explicitly enforces a rule to prioritize human evidence in memory v2 extraction.

Maya

Ah! So explicit developer constraints, feedback, and directives get weighted higher than thousands of lines of agent reasoning or tool logs.

Ethan Park

Precisely. When the model condenses past sessions, human input takes precedence during chunking. So your architectural rules do not get overwritten by noisy subagent logs.

Maya

That is huge for long running projects. But wait, if someone wants to test this, do they have to wipe their existing memory setups or take a leap of faith?

Ethan Park

No, they built in a safe rollout mechanism. You can set memory dual write equals true in your settings.

Maya

Dual writing mode?

Ethan Park

Yeah! It writes every memory update to both V1 and V2 storage engines in parallel. That lets you inspect readiness metrics and verify accuracy before you cut over your workflow completely.

Maya

Okay, but, um, running two extraction pipelines in parallel... there has to be a cost trade off there, right?

Ethan Park

There definitely is. Dual writing literally doubles your background memory extraction API calls on every turn. So if you are on a tight token budget, that is something to monitor closely.

Maya

Makes sense. And what about summary only extraction?

Ethan Park

Well, V2 offers summary only extraction to cut down prompt size, but if critical diagnostic evidence from a tool run isn't explicitly anchored in human text, summary extraction might strip out those raw diagnostic logs.

Maya

Right, so if an error trace matters, you need to reference it in your prompt so the extraction engine knows it is human prioritized evidence.

Ethan Park

Spot on. Now, speaking of API calls and limits, they also fixed a super annoying diagnostic bug in this release.

Maya

Oh, what was that?

Ethan Park

In earlier builds, if your monthly billing quota ran out, Codex returned an HTTP 429 error that looked identical to a temporary rate limit.

Maya

Oh man! So developers were setting up retry loops or waiting five minutes thinking it was just network traffic, when in reality their credit card hit the monthly cap!

Ethan Park

Exactly! Now, 0.155.0 explicitly differentiates quota exhaustion from standard concurrency rate limits in error reports. No more chasing ghost retries.

Maya

That is a relief. And what about the TUI changes? I heard the Agent Command Center got some single key shortcut updates?

Ethan Park

Yes! In the agents overview, you can now press h to hide background tasks, a to archive them, and d for confirmed deletion of clean managed git worktrees.

Maya

Confirmed deletion of clean managed worktrees... so if you have background subagents spinning up isolated worktrees, you can clean them up fast without accidentally blowing away uncommitted code.

Ethan Park

Right. It checks that the worktree is clean before allowing the deletion. It really tightens up managing multiple background agent workspaces.

Maya

Really sleek release overall. Memory V2 isolation, human evidence priority, and cleaner terminal shortcuts. Good stuff.

Ethan Park

Absolutely. Alright, that is it for today's quick breakdown. Catch you all next time!