GREATMEMORY.

// Private-cloud AI memory framework

Cloud marketplace or private cloud. Same controlled memory plane.

GreatMemory packages retrieval, guardrails, audit, provider management, and a starter chat UI into one controlled workspace. Run it from a cloud marketplace listing or in your private cloud, with your data, database, network boundary, and LLM keys under your control. Multi-workspace support is on the roadmap for users who need access to multiple data estates.

Grounded GreatMemorystreaming

Live answer over governed workspace knowledge.

Legal workspace

matter search · citations required

single workspace now

guardrail checked
audit hashed
future multi-workspace
Private cloud · with local LLM3 clouds · AWS, Azure, or Google CloudBYO · LLM keys and database588.1 MB · measured peak RSS

// Deployment posture

A simple choice: marketplace install or private-cloud service.

01

Cloud marketplace

Deploy the same workspace stack from AWS, Azure, or Google Cloud marketplace packaging. Entitlement happens at deploy time; the customer keeps the cloud account and data plane.

02

Private cloud

Use Helm, Terraform, or compose in your environment. We can prescribe the required services for your team to buy and install, or procure and install the supporting services as part of the engagement.

03

Managed maintenance

Private-cloud customers can add full maintenance: upgrades, database checks, observability setup, runbook support, and custom quotes based on the operating model.

// Measured, not claimed

The line stays flat while the corpus grows.

Both charts are drawn from real test runs. The reproducible command, raw output, and full methodology are linked directly under the data.

Full methodology →

Resident memory across ingest

51.2 MB across 1,000 documents (31,000 chunks), real BGE-small embeddings, real SQLite.

Footprint at a glance

Idle, full-load peak, and the ceiling enforced on every push.

Reproduce the benchmark

From the greatmemory source tree:

cargo test -p gm-cli --release -- --ignored ram_regression --nocapture
Read exactly what is measured →

Evidence trail

  • The line chart is the verbatim output of crates/gm-cli/tests/ram_regression.rs run on 2026-06-12.
  • The idle figure is rss_bytes from GET /v1/stats after booting the published Docker image.
  • The same test runs in CI and fails if peak RSS crosses 1 GB.
Show raw test output
[ram_regression] docs=  50  ingested=  2.6 MB  rss= 535.9 MB  peak= 535.9 MB
[ram_regression] docs= 100  ingested=  5.1 MB  rss= 530.5 MB  peak= 535.9 MB
[ram_regression] docs= 150  ingested=  7.7 MB  rss= 529.3 MB  peak= 535.9 MB
[ram_regression] docs= 200  ingested= 10.2 MB  rss= 535.8 MB  peak= 535.9 MB
[ram_regression] docs= 250  ingested= 12.8 MB  rss= 512.6 MB  peak= 535.9 MB
[ram_regression] docs= 300  ingested= 15.4 MB  rss= 578.1 MB  peak= 578.1 MB
[ram_regression] docs= 350  ingested= 17.9 MB  rss= 545.5 MB  peak= 578.1 MB
[ram_regression] docs= 400  ingested= 20.5 MB  rss= 547.7 MB  peak= 578.1 MB
[ram_regression] docs= 450  ingested= 23.1 MB  rss= 520.9 MB  peak= 578.1 MB
[ram_regression] docs= 500  ingested= 25.6 MB  rss= 545.8 MB  peak= 578.1 MB
[ram_regression] docs= 550  ingested= 28.2 MB  rss= 550.6 MB  peak= 578.1 MB
[ram_regression] docs= 600  ingested= 30.7 MB  rss= 543.4 MB  peak= 578.1 MB
[ram_regression] docs= 650  ingested= 33.3 MB  rss= 556.5 MB  peak= 578.1 MB
[ram_regression] docs= 700  ingested= 35.9 MB  rss= 543.7 MB  peak= 578.1 MB
[ram_regression] docs= 750  ingested= 38.4 MB  rss= 530.4 MB  peak= 578.1 MB
[ram_regression] docs= 800  ingested= 41.0 MB  rss= 526.9 MB  peak= 578.1 MB
[ram_regression] docs= 850  ingested= 43.6 MB  rss= 494.9 MB  peak= 578.1 MB
[ram_regression] docs= 900  ingested= 46.1 MB  rss= 560.0 MB  peak= 578.1 MB
[ram_regression] docs= 950  ingested= 48.7 MB  rss= 588.1 MB  peak= 588.1 MB
[ram_regression] docs=1000  ingested= 51.2 MB  rss= 541.3 MB  peak= 588.1 MB
[ram_regression] SUMMARY: ingested 51.2 MB across 1000 docs -> 31000 chunks; peak RSS 588.1 MB (limit 1024 MB)
test ram_regression_50mb_ingest_stays_under_1gb ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2619.96s

// What ships

One workspace stack for search, chat, ingestion, and audit.

Hybrid retrieval

Vector similarity, BM25 full-text, extracted facts, recency, reranking, and graph expansion are fused into grounded search and chat responses.

Guardrails

A rule layer catches obvious PII patterns first; an optional customer-configured LLM layer can monitor, redact, or block before retrieval runs.

Audit and compliance trail

Security-relevant actions, chat outcomes, source IPs, user agents, redacted previews, content hashes, and exports are written into a tamper-evident audit chain.

Auth and roles

Local users, LDAP/AD login, admin/member roles, and API access sit in the control plane so every UI, custom app, and ETL caller shares the same policy path.

Provider management

Customers bring their own OpenAI, Anthropic, Google, Azure OpenAI, Bedrock, Vertex, or local provider keys; private and marketplace modes allow hosted LLM APIs.

Ingestion and ETL

REST sources, connector runs, delta ingest, purge, rollback, and third-party ETL tools all pass through the same auth, guardrail, and audit path.

Full-stack observability

App metrics, traces, host/container CPU and memory, and Postgres metrics flow through OpenTelemetry into bundled Grafana, cloud-native monitoring, or both.

MCP and API surfaces

Starter chat UI, customer-built UIs, REST integrations, streaming chat, ETL endpoints, and MCP tools run over the same workspace memory engine.

Pluggable storage

SQLite, Postgres, and MySQL backends support local evaluation, managed cloud databases, and enterprise database constraints.

Appliance-ready

For air-gapped appliance mode, local-only inference can hard-block external AI providers and keep telemetry inside the local Grafana stack.

The retrieval, temporal graph, trust, episodic, and reflection layers are grounded in current memory-systems research - see the research & roadmap →

// Engage

Tell us which operating model you want.

We will scope the install as a marketplace deployment or a private-cloud service. For private cloud, we can either specify the required services for your team to buy and install, or procure, install, and maintain them under a custom quote.

# same images, different packaging
GM_DEPLOY_MODE=marketplace   # cloud marketplace
GM_DEPLOY_MODE=private       # private cloud
GM_DEPLOY_MODE=appliance     # air-gapped local-only mode

# the workspace control plane owns
auth + LDAP/AD
guardrails + BYO providers
hash-chained audit
ingest + purge + ETL
Grafana + cloud-native telemetry