LG Assistant - AI Knowledge & Workspace Search

What it is

LG Assistant is a single web-chat interface that Alberto talks to. It does two things:

1Answers from your knowledge base. It draws from resources you compile into a dedicated knowledge base: documents, guides, and reference materials you curate over time.
2Scans everything live. It reaches into your Gmail, Google Calendar, Google Drive, Calendly, and the Life Gateway CRM to pull real-time information on demand.

It is built on the same concept as Memorae, but made private and tailored specifically for Life Gateway.

How it works: two lanes

One chat box, two ways to answer. Each question routes to the right lane automatically.

Knowledge Base lane

Source: An "LG Knowledge Base" folder in your Google Drive.
Method: Indexed via semantic vector search.
Use: Recall over the resources you compile. You drag documents into the Drive folder; the server automatically syncs and indexes them.

Live Scan lane

Source: Gmail, Calendar, Calendly, CRM.
Method: Real-time API calls per question.
Use: Always-fresh answers, like "what is on my calendar Thursday" or "any emails from that client."

One chat box routes each question to the right lane automatically.

The Google Drive question, answered

Build a Knowledge Base folder in Drive, or just use the server?

Both. They are two ends of one pipe. The Drive folder is your human-friendly source of truth: just drag files in. The server is the brain that syncs, indexes, and answers. You never touch the server.

Can I just copy all of Drive into the server?

We automate the sync. Set it once, and it mirrors itself with no manual copying. Our recommendation: start with a curated folder rather than dumping the entire Drive. Curated means sharper answers and lower indexing cost. We can expand coverage later.

Architecture

01Web chat front-end

LibreChat, the interface Alberto types into.

02Knowledge Base index

Qdrant vector database, fed by an rclone sync of the Drive folder.

03Live tools

n8n workflows calling Gmail, Calendar, Calendly, and CRM APIs in real time.

04Answer synthesis

OpenAI API (GPT-5.4), on your own API account. Unlike the consumer ChatGPT app, API data is not used to train OpenAI's models.

05

Tenancy

Single Google account (yours), single-tenant, fully isolated stack.

Rollout plan

Phase	What you get	Status
Phase 1, KB MVP	Drive folder indexed plus web chat answering from your compiled resources.	Proposed
Phase 2, Live Scan	Add real-time Gmail, Calendar, and Calendly search.	Proposed
Phase 3, Full	Add CRM plus call and meeting transcripts, and optionally a WhatsApp interface like Memorae.	Proposed

Investment

The assistant runs on infrastructure we already operate. The only running cost is the AI itself, billed straight from your own OpenAI account — by usage, not by seat.

The cost is the build itself — a one-time $1,400. That covers the full build-out, including the testing and the several rounds of revision it takes to get the assistant answering the way Life Gateway needs.

After that, there is no per-seat charge. You load a small amount of OpenAI credit (see below) that the whole team draws from. Add as many people as you want at no extra licence cost.

One build. Usage-billed AI. No per-seat tax on your team.

$1,400

one-time build. Then only metered AI credit — a few dollars a month for the whole team, $0 per person.

What the AI costs to run

You load credit into your own OpenAI account and the whole team draws from it. There is no per-person fee — you pay only for the questions actually asked. To start, $20 of credit goes a long way.

Model	Cost per question	What $20 buys	How long for the team
GPT-5.4 Mini recommended	~$0.004	~4,900 questions	~3–8 months
GPT-5.4 Nano economy	~$0.001	~18,000 questions	~1–2.5 years

The ranges assume a typical workload of roughly 600 to 1,500 questions a month across all 9 people. Refill with another ~$20 a few times a year when it runs low — you are always topping up your own account, never paying us a subscription.

Put plainly: the AI for your whole team costs about the price of a couple of coffees a month. Compare that to $180 a month for 9 ChatGPT seats.

$20

of OpenAI credit lasts the whole team months — refill a few times a year.

How this compares to Memorae

Memorae is a great tool, but it bills per person, every year. There is no team plan — covering 9 people means 9 separate subscriptions. To get the Drive, Gmail, Calendar and CRM integrations Life Gateway needs, each person is on the top "Big Bang" tier.

	Memorae — Big Bang, 9 people	LG Assistant — this build
Year 1	~$2,430–$4,860	$1,400 one-time
Year 2 and beyond	~$2,430–$4,860 every year	~$0
People covered	Per person, capped — 9 separate bills	Unlimited, no per-seat charge
Where your data lives	Their cloud	Your server; AI runs on your own OpenAI account
Fit	Generic WhatsApp assistant	Tailored to Life Gateway and your CRM

Memorae's promotional rate (the lower figure above) runs on a countdown timer and resets to the standard rate after launch — so the real long-term cost is the higher number, billed every single year.

Our build pays for itself inside the first year against Big Bang, and roughly 1.5–2 years against the mid tier. After that it is essentially free while Memorae keeps billing per person, forever.

$0/seat

One build covers all 9 people — and anyone you add later.

What we need from you, Alberto

Four steps to get Phase 1 moving.

Create the folder

Make a folder in your Google Drive named "LG Knowledge Base" and drop in the documents you want the assistant to answer from.

Approve the APIs

Approve enabling the Gmail, Google Calendar, and Google Drive APIs on your account. We will send the consent screen.

Send the Calendly token

Generate your Calendly API token and send it over so we can wire up live booking search.

Confirm CRM scope

Tell us which Life Gateway CRM data you want included when we reach Phase 3.