ZoidLab MemoryMaker

Part of the ZoidLab platform · Foundry Package 04 · Live at memorymaker.zoidlab.ai

Design, govern, and deploy AI memory systems. — Where AI memory becomes manageable.

MemoryMaker is the control plane for how AI remembers, forgets, retrieves, ranks, expires, and protects information. Not a vector-database UI — an inspectable, governed, tenant-aware memory layer for Nyquest agents, workflows, websites, and enterprise apps.

Access — Nyquest Pro. MemoryMaker is a Pro workspace: sign in with your Nyquest account (a Pro or Teams plan is required; the whole app is gated). One .zoidlab.ai session is shared across every ZoidLab app.

The core workflow

Project → Memory Store → Ingest → Inspect → Recall Test → Rules → Governance → Export / Deploy

It answers three questions clearly: what does the AI remember, why did it remember that, and when should it forget it.

• Memory Explorer — inspect every stored memory: content, source, tags, sensitivity, confidence, expiration, access history; edit / archive / delete / forget.
• Ingestion — manual text, JSON, or TXT/MD/CSV upload; every entry runs the rules engine (secret redaction, sensitivity scoring, retention).
• Recall Test Lab — enter a query and see what's retrieved, the score and reason, the assembled context, what was excluded, and suggested tuning.
• Rules — remember / forget / expiration / redaction / sensitivity / access / retrieval rules.
• Governance — risk level, PII risk, retention, redaction, tenant isolation, right-to-be-forgotten, visible badges, and a risk scan.
• Analytics — growth, recall scores, by source / sensitivity / type, expiring-soon, high-risk.
• Export — the Nyquest Memory Package (memory.package.json) + YAML.

Stack

Aligned to the ZoidLab platform standard (Builder / Marketplace / Prompter / Foundry):

• Frontend — Next.js 15, React 19, TypeScript, TailwindCSS (dark)
• Backend — FastAPI (Python), SQLite (Postgres+pgvector-portable — all access behind database.py; JSONB stored as JSON text). Vector backend is a config-level abstraction (postgres_pgvector, chroma, qdrant, pinecone, weaviate, redis_vector, milvus, mock).
• Recall — deterministic hybrid recall (keyword overlap + tag/source/type + recency + confidence
  • expiration/archive exclusion). Set ENABLE_MOCK_EMBEDDINGS=false to wire real vector search.
• Auth — shared ZoidLab / Nyquest SSO cookie (zb_session), Pro-gated.
• Deploy — systemd + Cloudflare Tunnel (memorymaker-api :8500, memorymaker-web :3500).

Quick start (local)

cp .env.example .env          # set BUILDER_SESSION_SECRET

cd backend
python -m venv .venv && . .venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --port 8500   # seeds 7 projects + 10 stores on first boot

cd ../frontend
npm install && npm run dev     # http://localhost:3500

Or with Docker: docker compose up --build (frontend :3500, backend :8500, SQLite volume).

API summary

Projects GET/POST /api/projects, GET/PUT/DELETE /api/projects/{id}. Stores GET/POST /api/stores, GET/PUT/DELETE /api/stores/{id}, /clone. Memories GET/POST /api/stores/{id}/memories, GET/PUT /api/memories/{id}, /archive|/restore|/forget, DELETE /api/memories/{id}. Ingestion POST /api/stores/{id}/ingest/text|json|file, GET .../ingestion-jobs. Recall POST /api/stores/{id}/recall, GET .../recall-tests. Rules GET/POST /api/stores/{id}/rules, PUT/DELETE /api/rules/{id}, /enable|/disable. Governance GET/PUT /api/stores/{id}/governance, POST /api/stores/{id}/scan-risk. Analytics GET /api/stores/{id}/analytics. Export GET /api/stores/{id}/export/json|yaml.

Nyquest Memory Package

memory.package.json (schema_version 1.0) — storage, retrieval, chunking, retention, governance, rules, and example (or full) memories. Built by backend/exporter.py.

Data model

users, organizations, memory_projects, memory_stores, memory_entries, memory_rules, memory_ingestion_jobs, recall_tests, memory_access_logs, audit_logs — see backend/database.py.

Deploy notes (memorymaker.zoidlab.ai)

Two systemd services behind the shared Cloudflare Tunnel: memorymaker-web (Next, :3500) and memorymaker-api (FastAPI, :8500, localhost). The frontend proxies /api/* to the backend. Add the hostname to the tunnel ingress and route DNS with cloudflared tunnel route dns.