Harley — Fitness Coaching by DNAi Systems

AI fitness coaching agent via the A2A protocol. Designs workouts, analyzes exercise science, coaches clients. Backed by ~52M training/sport-science vectors from the Citadel corpus.

Agent Card

https://api.askasha.org/.well-known/agent-card.json?agent_id=harley

A copy is included in this repo at agent-card.json.

Skills

ID	Skill	Description
workout-design	Workout Design	Push-pull-legs, powerlifting, hypertrophy programming with periodization
exercise-lookup	Exercise Lookup	Curated exercise library with muscles, equipment, progressions
nutrition-coaching	Nutrition Coaching	Macro targets, meal planning (USDA FoodData Central)
client-assessment	Client Assessment	Trend analysis on body stats, workout logs, PRs
program-builder	Program Builder	Multi-week periodized programs with deloads and RPE targets

Quick Start

# 1. Get a free API key
curl -X POST https://api.askasha.org/api/a2a/signup \
  -H "Content-Type: application/json" \
  -d '{"email":"you@example.com","name":"Your Name","tier":"free"}'

# 2. Send a query
curl -X POST https://api.askasha.org/a2a/v1/message:send \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"message":{"role":"user","parts":[{"text":"Design a 4-day push-pull-legs split"}]},"metadata":{"agent_id":"harley"}}'

A2A Surface

Bearer-auth A2A v1.0 endpoints, identical across all DNAi agents — only the agent_id changes:

Endpoint	Purpose
POST /a2a/v1/message:send	Send a query, receive a Task
GET /a2a/v1/tasks/{id}	Poll task state and fetch artifacts
GET /a2a/v1/tasks	List recent tasks (scoped to caller)
POST /a2a/v1/tasks/{id}:cancel	Cancel an in-flight task
GET /a2a/v1/health	Health probe
POST /api/a2a/upgrade / GET /api/a2a/portal	Stripe upgrade and self-service billing
POST /api/a2a/rotate-key	Rotate your API key

Connect to Google Gemini Enterprise

Harley registers as a "Custom agent via A2A" inside any Google Gemini Enterprise app. Paste this repo's agent-card.json into Gemini Enterprise → your app → Agents → Add Agents → Custom agent via A2A and click through. The card is shaped to be maximally compatible: it carries both the A2A v1.0 supportedInterfaces[] shape AND top-level url / protocolVersion / iconUrl / documentationUrl for the Gemini Enterprise console parser. Full step-by-step (with curl automation example) lives in asha-a2a → Connect to Google Gemini Enterprise — every DNAi agent uses the identical registration flow.

Evaluating Harley against your current fitness AI

The same A/B protocol used to evaluate Asha applies to Harley — same auth, same A2A surface, same provenance fields. Pull the harness from the Asha repo and change agent_id from "asha" to "harley":

ASHA_API_KEY=ak_...        \
OTHER_API_BASE=https://your-current-api  \
OTHER_API_KEY=...          \
EVAL_SET=fitness_eval_set.txt \
  python evaluation.py  # from github.com/EndlessRay/asha-a2a/examples

Build a representative eval set from your trainer's real queries — programming questions, exercise selection, periodization, nutrition macros, client-progression edge cases. Score on factual accuracy (cross-check against ACSM / NSCA / peer-reviewed exercise science), latency, and whether responses include verifiable provenance.

Full protocol with metrics, FENG falsification step, and migration path: see asha-a2a/README.md.

Pricing

Tier	Monthly	Daily / Monthly Cap
Free	$0	10 / 50
Developer	$49	200 / 1,000
Pro	$199	2,000 / 10,000
Enterprise	Custom	Unlimited

Part of 11 public agents at DNAi Systems. Fleet discovery. MIT license (interface only — engine, knowledge collections, and architecture are proprietary).