mirror of
https://github.com/tiennm99/claude-central-gateway.git
synced 2026-04-17 19:21:30 +00:00
- Add main proxy endpoint (api/v1/messages.js) - Add token validation via GATEWAY_TOKEN - Add model mapping via MODEL_MAP env var - Support full SSE streaming in Anthropic format - Add package.json with openai dependency - Add vercel.json for routing config - Update README with Quick Start guide - Consolidate memory to .claude/memory/ Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1.6 KiB
1.6 KiB
name, description, type
| name | description | type |
|---|---|---|
| Memory Saving Rules | Rules for saving and managing memories in this project | feedback |
How to Save Memories
Saving a memory is a two-step process:
Step 1 — Write the memory to its own file (e.g., user_role.md, feedback_testing.md) using this frontmatter format:
---
name: {{memory name}}
description: {{one-line description — used to decide relevance in future conversations}}
type: {{user, feedback, project, reference}}
---
{{memory content — for feedback/project types, structure as: rule/fact, then **Why:** and **How to apply:** lines}}
Step 2 — Add a pointer to that file in MEMORY.md. MEMORY.md is an index — each entry should be one line, under ~150 characters: - [Title](file.md) — one-line hook.
Memory Types
- user: Information about user's role, goals, preferences, and knowledge
- feedback: Guidance on how to approach work — what to avoid and what to keep doing
- project: Information about ongoing work, goals, initiatives, bugs, or incidents
- reference: Pointers to where information can be found in external systems
What NOT to Save
- Code patterns, conventions, architecture (read from code instead)
- Git history (use
git log/git blame) - Debugging solutions (the fix is in the code)
- Anything already documented in CLAUDE.md files
- Ephemeral task details
Why: Memories should capture unique context not derivable from other sources.
How to apply: Before saving, ask: "Is this information available elsewhere?" If yes, don't save as memory.