The most requested feature for Gitlawb Zero just landed: context preservation in exec prompts.
PR #460 is titled “fix: preserve conversation context in exec prompts” — but it’s more than a fix. It’s a fundamental upgrade to how Zero works.
Before vs After
Before: You ask Zero a question, it answers. You ask a follow-up — it has no idea what you just said. Every interaction was stateless. This was by design — Zero’s philosophy is minimalism. But it made multi-turn conversations impossible.
After: Zero now preserves conversation context across exec prompts. Ask a question, get an answer, ask a follow-up, and Zero connects the dots. The session persists in memory, carrying the conversation thread forward.
Why This Matters
This is the feature that transforms Zero from “a CLI that answers questions” to “an AI assistant that holds conversations.” It’s the same transition that every coding agent goes through — and Zero just crossed that threshold.
For users, it means:
- No repeating context in every prompt
- Multi-step reasoning actually works
- Conversations feel natural instead of robotic
- Complex tasks can be broken into multiple prompts
How It Works
Zero’s session store now maintains conversation history between turns. Each exec prompt includes the relevant context from the previous exchange — not the full history (that would blow the context window), but enough to maintain coherence.
The implementation is intentionally lightweight. Zero doesn’t have a memory system like Hermes’s mem0 or a knowledge graph like pi.dev. But context preservation is the foundation that more advanced memory features will build on.
What’s Next
With context preservation in place, Zero can now support multi-turn workflows, gradual task decomposition, and actual conversations. The decentralized coding agent just became a lot more useful.