Purpose
Initialize the Memory Hub for a project: create .memory/ skeleton, scan the project, and populate initial knowledge files and catalog.
Input
No arguments required. Run from the project root.
Required Flow
Step 1: Create skeleton
memory-hub initIf ALREADY_INITIALIZED error, stop and inform the user.
Step 2: Scan project and generate tech-stack knowledge
- Read the project root file listing
- Read package manager files (
package.json,pyproject.toml,Cargo.toml,go.mod,pom.xml, etc.) - Read entry files (
main.*,index.*,app.*,server.*, etc.) - Read config files (
.env.example,tsconfig.json,webpack.config.*, etc.)
Generate tech-stack.md content, write it to .memory/architect/tech-stack.md, then register it:
memory-hub index architect tech-stack.md \
--topic tech-stack --summary "技术栈、关键依赖、使用方式与限制"Step 3: Generate code conventions knowledge
Based on project structure from Step 2, generate conventions.md, write it to .memory/dev/conventions.md, then register it:
memory-hub index dev conventions.md \
--topic conventions --summary "目录命名规则、模块组织方式、代码约定"Step 4: Scan project modules and generate Catalog
Analyze directory structure, identify functional domains and key files, construct JSON:
Write the JSON to a temporary file, then run:
memory-hub catalog-update --file /tmp/modules.jsonStep 5: Quality gate
- List
unknowns— files/directories that don't clearly belong to any functional domain - Check
catalog.repairoutput (auto-triggered by catalog-update):
- ai_actions non-empty → execute self-healing, then run memory-hub catalog-repair again to confirm cleared - manual_actions non-empty → report to user
Step 6: Output summary
Report: files created, modules identified, unknowns requiring user confirmation.
Output
JSON envelope from each command. Final summary to user.
Error Handling
ALREADY_INITIALIZED→ inform user, do not proceed- Any command failure → stop and report the error