Frequently Asked Questions

Environment and Installation

What environment does LingXi run in?

LingXi is an engineering workflow product that supports both Codex and Claude Code. After installation, its runtime lives inside the target repository, primarily under:

• .lingxi/ (host-agnostic core)
• .codex/config.toml, .codex/hooks.json, .codex/agents/ (Codex adapter)
• .claude/settings.json, .claude/agents/, .claude/skills/ (Claude Code adapter)

Do I need local Node.js?

Yes.

LingXi's setup, bootstrap, automation registration, and lower-level runtime scripts currently depend on Node.js, so the target environment needs a working node.

What is the most important step after installation?

The most important step is making sure bootstrap and automation are actually in place.

If you used the official remote install script, bootstrap is usually already handled.
If you synced files manually, run:

node scripts/lx-bootstrap.mjs

That step also creates or merges the repo-local Codex hook config that enables automatic memory consumption for generic repository turns.

How do I uninstall LingXi?

From the project root:

yarn lx:uninstall
# or
npm run lx:uninstall

For CI or other non-interactive environments:

yarn lx:uninstall --yes

Workflow

What is LingXi's current core workflow?

The current foreground workflow is:

task → vet

task shapes a task document. vet challenges that task before implementation starts.

Does LingXi inject a lot of context into every conversation?

LingXi is designed to keep the foreground workflow narrow.

Explicit workflows run only when you call them. Background memory continues to distill and retrieve, and generic meaningful repository turns receive only the smallest useful memory brief through the hook adapter.

Why am I not seeing automatic memory injection in normal conversations?

Check these first:

1. whether bootstrap has already created the hook config for your host (Codex: .codex/config.toml + .codex/hooks.json; Claude Code: .claude/settings.json)
2. whether the repository is trusted and hooks are actually active in Codex or Claude Code
3. whether the current prompt is meaningful repository work rather than casual conversation
4. whether you are on native Windows, where the current Codex runtime still does not execute hooks natively

Memory System

Will memory become noisy over time?

LingXi controls noise through three layers:

1. taste adjudicate filters for value
2. governance decides create / merge / skip
3. retrieval returns only the smallest useful high-signal set

Its goal is to preserve the engineering judgment that is actually worth reusing.

How does memory affect task and vet?

task retrieves relevant memory while drafting and records memories that materially shaped the task in memory_refs.
vet retrieves relevant memory again and checks whether the task already reflects those important judgments.

Do I need to sync the memory index manually?

In the current mainline, INDEX.md is maintained automatically during memory writes, so no manual sync is needed.

Quality and Validation

How does LingXi avoid producing arbitrary output?

LingXi uses a hybrid model:

• LLMs handle semantic extraction, adjudication, governance, and ranking
• deterministic scripts handle schema, state safety, persistence, and indexing

So the goal is structured output, stable state, and auditable runtime behavior.

How do I verify that the current install surface is intact?

You can use these tests and checks from the main repository:

• npm test
• install-manifest-exists
• install-manifest-coverage
• install-manifest-version-sync

More

• Commands and Workflow Reference
• Memory System
• Quick Start