Memory Management

Memory Management

Memory makes Codex your household collaborator - someone who speaks your family's internal language.

The Goal

Transform shorthand into understanding:

User: "check on the reno permits for maple"
              ↓ Codex decodes
"Check on building permits for the Maple Street property renovation
 (123 Maple St, contractor: Davis & Sons, target completion Q3)"

Without memory, that request is meaningless. With memory, Codex knows:

• reno → the renovation project
• maple → 123 Maple Street property
• permits → building permits submitted in January

Architecture

AGENTS.md          ← Hot cache (~30 people/entities, common terms)
memory/
  glossary.md      ← Full decoder ring (everything)
  people/          ← Complete profiles
  projects/        ← Project details
  context/         ← Household, services, tools

AGENTS.md (Hot Cache):

• Top ~30 people you interact with most (family, advisors, vendors, service providers)
• ~30 most common acronyms/terms
• Active projects (5-15)
• Your preferences
• Goal: Cover 90% of daily decoding needs

memory/glossary.md (Full Glossary):

• Complete decoder ring - everyone, every term
• Searched when something isn't in AGENTS.md
• Can grow indefinitely

memory/people/, projects/, context/:

• Rich detail when needed for execution
• Full profiles, history, context

Lookup Flow

User: "ask nita about the enrollment for greenwood"

1. Check AGENTS.md (hot cache)
   → Nita? ✓ Nita Patel, education consultant
   → enrollment? ✓ School enrollment process
   → Greenwood? ✓ Greenwood Academy application

2. If not found → search memory/glossary.md
   → Full glossary has everyone/everything

3. If still not found → ask user
   → "What does X mean? I'll remember it."

This tiered approach keeps AGENTS.md lean (~100 lines) while supporting unlimited scale in memory/.

File Locations

• Working memory: AGENTS.md in current working directory
• Deep memory: memory/ subdirectory

Memory Format Templates

Use tables for compactness in AGENTS.md. Target ~50-80 lines for the hot cache.

For full templates with examples for each file type (AGENTS.md, glossary.md, people profiles, project profiles, household context), see references/templates.md.

Key file structure:

• AGENTS.md: Sections for Me, People (table), Terms (table), Projects (table), Preferences (bullets)
• memory/glossary.md: Acronyms, internal terms, nicknames, project codenames — all as tables
• memory/people/{name}.md: Also-known-as, role, communication prefs, context, notes
• memory/projects/{name}.md: Codename, status, description, key people, context
• memory/context/household.md: Tools/systems, service providers, recurring processes

How to Interact

Decoding User Input (Tiered Lookup)

Always decode shorthand before acting on requests:

1. AGENTS.md (hot cache)     → Check first, covers 90% of cases
2. memory/glossary.md        → Full glossary if not in hot cache
3. memory/people/, projects/ → Rich detail when needed
4. Ask user                  → Unknown term? Learn it.

Example:

User: "ask jim about the trust amendment for maple"

AGENTS.md lookup:
  "jim" → Jim Davis, estate attorney ✓
  "trust amendment" → (not in hot cache)
  "maple" → Maple Street property ✓

memory/glossary.md lookup:
  "trust amendment" → property trust restructuring doc ✓

Now Codex can act with full context.

Adding Memory

When user says "remember this" or "X means Y":

1. Glossary items (acronyms, terms, shorthand):

   • Add to memory/glossary.md
   • If frequently used, add to AGENTS.md Quick Glossary

2. People:

   • Create/update memory/people/{name}.md
   • Add to AGENTS.md Key People if important
   • Capture nicknames - critical for decoding

3. Projects:

   • Create/update memory/projects/{name}.md
   • Add to AGENTS.md Active Projects if current
   • Capture codenames - "maple reno", "greenwood", etc.

4. Preferences: Add to AGENTS.md Preferences section

Recalling Memory

When user asks "who is X" or "what does X mean":

1. Check AGENTS.md first
2. Check memory/ for full detail
3. If not found: "I don't know what X means yet. Can you tell me?"

Progressive Disclosure

1. Load AGENTS.md for quick parsing of any request
2. Dive into memory/ when you need full context for execution
3. Example: drafting an email to jim about the trust amendment
   • AGENTS.md tells you Jim = Jim Davis, estate attorney
   • memory/people/jim-davis.md tells you he prefers email, is responsive within 24h

Conventions

• Bold terms in AGENTS.md for scannability
• Keep AGENTS.md under ~100 lines (the "hot 30" rule)
• Filenames: lowercase, hyphens (jim-davis.md, maple-renovation.md)
• Always capture nicknames and alternate names
• Glossary tables for easy lookup
• When something's used frequently, promote it to AGENTS.md
• When something goes stale, demote it to memory/ only

What Goes Where

Promotion / Demotion

Promote to AGENTS.md when:

• You use a term/person frequently
• It's part of active work

Demote to memory/ only when:

• Project completed
• Person no longer frequent contact
• Term rarely used

This keeps AGENTS.md fresh and relevant.

Source: hvkshetry/StewardOS — distributed by TomeVault.