journeyby thundrstruck
← All journeys

Knowledge graph

Built from every journey that has a job context. The assistant uses it as background when drafting and editing maps.

⬇ Training data (JSONL)
2roles
4industries
4tasks
8stages
12stakeholders
88actions
9tools
13responsibilities
2maps generated
16AI edits applied

Unit Owner

Residential condominium property management · 2 journeys

ToolsBuilding management system (BMS) for central plant monitoring · Email and phone · Master keys and the unit-entry log · Preferred vendor list · Property management / work-order software (resident portal, ticketing, vendor di

Handling an occupant's report of an air conditioner blowing warm or uncooled air, from first report to confirmed fix and

StagesReport → Assess → Plan → Diagnose → Resolve → Follow-up

Works withUnit Owner (UO) ×2 · Sentient ×2 · Occupant (OC) ×2 · Service provider ×2 · Maintenance Technician (MT) ×2

ActionsDetect equipment fault · Report issue · Receive issue report · Generate case · Generate initial diagnosis · Send confirmation · Receive confirmation · Review case alert · Request additional details · Update case details in portal · Provide additional details · Provide decision support to UO · Record updated case details · Revise initial diagnosis · Update case · Send case alert · Receive case alert · Review case · +34 more

Building Manager

industry not set · 0 journeys

Training data

Every generated map (as you corrected it) and every AI edit you applied is logged as a training example — 18 so far. Once there are a few hundred, the JSONL can fine-tune the model on the M1 with MLX (see AI.md). Until then the graph above is passed in as context.