getting-started.md

Getting Started

Prerequisites

• Node 20 or newer
• pnpm 9 or newer (npm install -g pnpm)
• An Anthropic API key — get one at https://console.anthropic.com/

Install

git clone https://github.com/<you>/fieldwork
cd fieldwork
pnpm install

Configure

cp .env.example .env
# open .env and paste your ANTHROPIC_API_KEY

Run the web app

pnpm --filter @fieldwork/web dev

Open http://localhost:3000.

Run the CLI

pnpm --filter @fieldwork/cli exec fieldwork validate packages/scenarios/support-triage/manifest.yaml

Run the tests

pnpm test

What's playable today

• All six scenarios in the catalog (Support Ticket Classifier, Internal Doc Q&A, Data Pipeline Automation, Multi-Step Workflow Agent, Legacy System Migration, Production Incident Response) are authored, schema-valid, and loadable through the web app. Support Ticket Classifier and Internal Doc Q&A are verified end-to-end against a live API; the other four are authored but not yet exercised on live API.
• Full turn loop through inner Claude with JSON contract validation + retry
• Prompt caching and Haiku/Sonnet model tiering
• Stakeholder dialogue with trust meter (0-1 per stakeholder)
• Discoverable objectives (hidden until the trainee surfaces them)
• Per-turn objective state transitions: inner Claude proposes + deterministic manifest rubric rules (action_kind / payload_contains / payload_regex) override when they match
• Surprise engine with turn_count, objective_state, action_pattern, and random triggers
• Turn budget + cumulative USD cost display
• Collapsible action log viewer
• End-of-scenario debrief with structured per-turn critiques (summary callout, objective pills, "what you did" / "try instead" cards, focus)
• JSON file session persistence — survives server restart and rehydrates in the browser via ?session=<id> (turn counter, cost, trust, objectives, inbox, last-turn narrative all restore)
• fieldwork validate CLI wired to the scenario schema

See TODO.md for what's left to build.