Agent Pipelines

Every assessment is a deterministic chain of agents picked by mode. The Intake and Data Scout agents always run first; the middle stages are mode-specific; the Synthesis agent always closes the run and writes the report to agent_runs.result.

Pipeline matrix

ModeIntakeData ScoutNEB AssessorPlacemaking RadarSynthesisETA
Quick Scan
quick_scan
~45s
NEB Context
neb_context
~90s
Placemaking Zones
placemaking_zones
~90s
Full Assessment
full_assessment
~2 min

Pipelines in detail

Quick Scan quick_scan

Composite score and evidence inventory. Skips NEB and placemaking scoring — useful for triage.

Intake  →  Data Scout  →  Synthesis

NEB Context neb_context

Adds Beautiful · Sustainable · Together scoring (JRC139118) and an evidence-backed rationale.

Intake  →  Data Scout  →  NEB Assessor  →  Synthesis

Placemaking Zones placemaking_zones

Identifies 3-5 placemaking zones with per-zone suitability scores and a neighbourhood-level placemaking score.

Intake  →  Data Scout  →  Placemaking Radar  →  Synthesis

Full Assessment full_assessment

Runs every agent. Produces NEB alignment, placemaking radar, narrative, recommendations, and peer benchmarks.

Intake  →  Data Scout  →  NEB Assessor  →  Placemaking Radar  →  Synthesis

Selecting a mode

  • Quick Scan — fastest read on whether a neighbourhood has enough evidence to be worth a deeper run.
  • NEB Context — for grant applications, ERDF / Horizon submissions, and NEB Lab reviews.
  • Placemaking Zones — for urban designers and BIDs scoping street-level interventions.
  • Full Assessment — the default for consulting deliverables and investor screens.

Pipelines are enforced server-side in src/lib/runs.functions.ts; clients cannot reorder or skip stages.