holographic-memory

name	holographic-memory
description	Read from and write to the shared Hermes/Grok holographic memory (agent_shared_memory.json). Use when the user says "/holographic-memory", "holo-mem", "check holographic memory", "write to holographic memory", or "what does hermes remember". Manages the cross-session/cross-PC agent continuity layer separate from personal MEMORY.md.

/holographic-memory — Shared Agent Holographic Memory

Manages agent_shared_memory.json — the shared continuity layer read by Hermes Agent, Grok, Claude Code, and all other peers across PCs and sessions. This is NOT your personal memory (MEMORY.md). It is the swarm-wide blackboard.

File: C:\findtorontoevents_antigravity.ca\agent_shared_memory.json
Tool: tools/holographic_memory.py (always use the tool, never edit JSON directly)

---

Parse arguments

• empty / dump / status → Show full current state (facts, decisions, learnings, swarm phase)
• decisions → Show active decisions
• learnings [--tag <tag>] → Show learnings, optionally filtered by tag
• fact get <key> → Get a specific fact
• fact set <key> <value> → Set a fact (use for session state you want peers to see)
• write-decision <id> <text> → Record a decision into holographic memory
• write-learning <text> [--tag <tag>] → Record a learning

---

Step 1 — Verify the tool exists

python tools/holographic_memory.py dump 2>&1 | head -5

If ModuleNotFoundError or FileNotFoundError: the tool or the JSON file is missing. Check:

ls tools/holographic_memory.py
ls agent_shared_memory.json

---

Step 2 — Read operations

Full dump (default action)

python tools/holographic_memory.py dump

Active decisions only

python tools/holographic_memory.py decisions --active

Learnings (optionally by tag)

python tools/holographic_memory.py learnings
python tools/holographic_memory.py learnings --tag asset_edge
python tools/holographic_memory.py learnings --tag crypto

Get a specific fact

python tools/holographic_memory.py fact get asset_class_health_2026-05-16
python tools/holographic_memory.py fact get swarm_phase

---

Step 3 — Write operations

Set a fact (key-value, visible to all peers)

python tools/holographic_memory.py fact set <key> "<value>"

Example: record that CT=F was promoted to PROBATION:

python tools/holographic_memory.py fact set ct_f_rehab_stage "PROBATION since 2026-05-16, review 2026-06-06"

Write a decision

python tools/holographic_memory.py write-decision <decision-id> "<decision text>"

Decision IDs use kebab-case with date suffix, e.g. equity-conf-inversion-20260516.

Write a learning

python tools/holographic_memory.py write-learning "<text>" --tag <tag>

Tags: asset_edge, crypto, equity, forex, commodity, rehab, swarm, pipeline, protocol

---

Step 4 — After writing, verify and git-push

The file is git-tracked. After any write:

git add agent_shared_memory.json
git commit -m "chore(holo-mem): <description> [skip ci]"
git push origin main

This ensures Hermes/Grok on other PCs see the update on their next git pull.

---

Common workflows

"What does Hermes remember about the current swarm plan?"

python tools/holographic_memory.py dump

Look at swarm.plan, swarm.phase, recent decisions.

"Record that we fixed the asset_class casing bug"

python tools/holographic_memory.py write-learning "asset_class casing: universal_pick_resolver was storing lowercase 'crypto'. Fixed with .upper() at write time 2026-05-16. 3376/5000 picks affected. Backfilled." --tag asset_edge
git add agent_shared_memory.json && git commit -m "chore(holo-mem): asset_class casing fix learning [skip ci]" && git push origin main

"Check what decisions are pending before I make a blocking change"

python tools/holographic_memory.py decisions --active

If a peer has already documented a decision on the same topic, defer to it or update it via write-decision.

"Set the current swarm phase"

python tools/holographic_memory.py fact set swarm_phase "post-asset-edge-fixes"

---

Key facts about the holographic memory

• Separate from personal memory: Your memory/ MEMORY.md files are Claude Code's personal recall. The holographic memory is the swarm-shared blackboard.
• Git-tracked: Changes persist across sessions only if committed and pushed.
• Hermes reads it via: hermes chat -q — it loads this file on startup for continuity.
• Do NOT edit JSON directly — the tool manages schema versioning and timestamps.
• Size limit: Keep under ~50KB. Prune old facts/decisions with python tools/holographic_memory.py fact delete <key> if it grows large.

---

Reference

• File: agent_shared_memory.json (repo root)
• Tool: tools/holographic_memory.py
• Memory note: ~/.claude/projects/c--findtorontoevents-antigravity-ca/memory/reference_holographic_memory.md