Manzil / GurmanGurman AI · Evidence-led recommendationsOpen the workspace

Gurman AI · Evidence-led recommendations

Find your nextcity plan.

Tell Gurman what you want to do — it surfaces real places from the Manzil catalogue and explains why they fit.

Real reviewsReal catalogue places
02

Trust first

Gurman does not make up what it does not know.

Manzil puts trust before impressive-sounding answers. A recommendation is tied to a real business, and missing information is not presented as fact.

01

Real records

Recommendations start in the catalogue

Gurman is designed to recommend only existing, publicly visible businesses from the Manzil catalogue.

02

Clear reason

Inspect every choice

The recommendation card shows why a place fits your request.

03

You stay in control

The decision remains yours

Gurman narrows and explains the options. You decide whether to contact or visit a place.

03

Gurman Intelligence

Not a chatbot. Manzil's decision-making layer.

Manzil is not built to simply display information — it is built to help people choose with confidence. Gurman Intelligence connects businesses, services, experiences, and user context to turn intent into a real-world plan.

How a decision is formed

01Request02Intent03Hybrid retrieval04Reasoning05Decision06Tool orchestrator07LLM presentation08Response
01The LLM never decides

The model has no direct access to databases or raw entities. It explains structured Decision and Explanation objects only.

04

Memory Model · Epic 05

Gurman remembers structured knowledge, not raw chats.

Instead of replaying raw chat histories into an LLM, the Memory Engine manages user context as strongly typed MemoryObjects. That means Gurman does not treat every request as a cold start.

The philosophy

Structured context, never raw chat

Every memory is stored with its provenance, confidence, retrieval priority, and lifecycle.
  • memoryId · a unique identity
  • source · explicit statement or observed behaviour
  • confidence · reliability of the inference

Six typed tiers · retrieval priority

01WorkspaceTimelineReal-time context for active collaborative steps, timelines, and bookings.
02MissionContextThe immediate task: a birthday tomorrow for 6 guests near Yunusabad with a 2.5M UZS budget.
03RelationshipContextSocial connections, group preference vectors, and shared budget limits.
04PreferenceContextLong-term tastes: cuisine, travel radius, atmosphere, and habits.
05BusinessContextAn AI profile of a business's operational reality, built from customer experience.
06MarketplaceContextCity and neighbourhood signals such as demand spikes, traffic, and weather.
05

Collaborative Workspace · Multiplayer

For group plans, Gurman serves consensus — not the person who spoke last.

Real-life plans do not belong to one user. The Workspace becomes the group's central object, while Gurman acts as an active mediator across different tastes, budgets, and schedules.

Group Consensus Engine

Gurman as moderator

Each person's preferences remain visible, but the decision is built around the group's shared constraints.

  1. 01Aggregate preferences
  2. 02Surface conflicts
  3. 03Propose a compromise
  4. 04Collect votes
  5. 05Mark the plan ready

Try it

Tell Gurman about your next plan.

Write naturally. Gurman helps move you from a request to real places and a more confident choice.

Open the workspace