Using Templates

Templates provide proven workflow patterns that generate DAGs automatically. This guide covers how to use, customize, and choose the right template.

Using a Template

Via SDK

const chain = await client.summon({
  goal: 'Review literature on mRNA vaccine delivery',
  room: 'bio-lab',
  template: 'literature_review',
});

Via CLI

# In the interactive REPL
/template literature_review "Review literature on mRNA vaccine delivery"

Via MCP

Use the society_summon tool with the template parameter.

Choosing a Template

For Software Development

• software_feature — Full feature development with parallel frontend/backend tracks
• bug_fix — Root cause analysis → fix → verify pipeline

For Research

• research_swarm — Best for broad topic investigation with parallel sub-domains
• literature_review — Academic-style systematic review
• literature_review_continuous — Ongoing monitoring of a research area
• hypothesis_swarm — When you need to compare competing theories
• research_monitor — Watching for new publications and evidence

For Medical/Scientific

• second_opinion — Clinical case review with expert validation
• rare_disease_diagnosis — When multiple specialists need to weigh in
• clinical_trial_monitor — Multi-site trial safety surveillance
• drug_interaction_analysis — Polypharmacy analysis for patient safety
• epidemiological_investigation — Outbreak investigation

For Business

• strategic_analysis — Market, competition, and strategy planning

For Content

• content_creation — Multi-stage editorial pipeline

For Simple Tasks

• simple_task — Quick execute → review → finalize
• parallel_execution — Split work into N parallel tracks

Customizing Templates

Configurable Options

// Research swarm with 5 parallel domains
await client.summon({
  goal: 'Comprehensive AI safety review',
  template: 'research_swarm',
  options: { domains: 5 },
});

// Rare disease diagnosis with custom specialists
await client.summon({
  goal: 'Undiagnosed cardiac phenotype',
  template: 'rare_disease_diagnosis',
  options: {
    specialists: ['cardiology', 'genetics', 'pathology', 'radiology'],
  },
});

// Clinical trial with 8 sites
await client.summon({
  goal: 'Phase III oncology trial monitoring',
  template: 'clinical_trial_monitor',
  options: { sites: 8 },
});

Template DAG Patterns

Sequential Pipeline

step_1 → step_2 → step_3 → step_4

Used by: simple_task, bug_fix

Fan-Out / Fan-In

        ┌─ domain_1 ─┐
scope ──┼─ domain_2 ─┼── synthesize
        └─ domain_3 ─┘

Used by: research_swarm, hypothesis_swarm, parallel_execution

Multi-Track with Cross-Review

protocol ──┬─ site_1_data ─┬─ site_1_safety ─┬─ aggregate ─ dsmb ─ report
           ├─ site_2_data ─┤─ site_2_safety ─┤
           └─ site_3_data ─┘─ site_3_safety ─┘

Used by: clinical_trial_monitor

Specialist Parallel

intake ─┬─ specialist_A ─┬─ cross_correlate ─ consensus ─ report
lit_scan┼─ specialist_B ─┤
        ├─ specialist_C ─┤
        └─ specialist_D ─┘

Used by: rare_disease_diagnosis

Listing and Searching

import { listTemplates, searchTemplates, getTemplate } from 'society-core';

// List by category
const medical = listTemplates('medical');

// Search by keyword
const results = searchTemplates('pharmacology');

// Get specific template
const template = getTemplate('rare_disease_diagnosis');
console.log(template.name);        // "Rare Disease Multi-Specialist Diagnosis"
console.log(template.tags);        // ["rare-disease", "diagnosis", ...]
console.log(template.description); // Full description