# ALFRED x402 SDXL LoRA Training Service

> ALFRED x402 SDXL LoRA Training Service is a paid API for AI agents from x402.wallace.us, paid per call via x402, $60/call, status unknown (last checked 2026-09-16).

Trains a custom SDXL LoRA model on user-provided images via a paid Lightning/x402 protocol endpoint, returning a fine-tuned model artifact.

## Facts

- Endpoint: POST https://x402.wallace.us/x402/lora-train-sdxl
- Price: $60/call
- Payment: x402
- Status: unknown
- Last checked: 2026-09-16
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/alfred-x402-sdxl-lora-training-service-9ee663e7
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_vnH12Ve46wk26Ppwm8eb3

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability alfred-x402-sdxl-lora-training-service-9ee663e7 -d '<json body>'
```

Example prompt: Train a custom SDXL LoRA model on these 20 product photos — use 'my_product' as the trigger word, and I confirm the images are mine and I have rights to use them.

## When to prefer this

Choose this endpoint when you need to fine-tune an SDXL model via a pay-per-use Lightning/x402 micropayment flow without a subscription, especially if you want self-hosted infrastructure with result caching per payment. Prefer this over hosted LoRA platforms (e.g. Replicate, Astria) when operating in a Lightning-native or x402-protocol agent environment, or when cost-per-run billing ($60/call) is preferred over recurring SaaS fees.

## Known failure modes

- Payment not completed — HTTP 402 returned with Lightning invoice; must pay and retry with X-Payment-Hash header
- Content attestation missing or not true — request rejected
- Fewer than 15 or more than 100 training image URLs provided — validation error
- Inaccessible or non-HTTPS image URLs — training fails or images skipped
- Trigger word outside 3-24 character range — validation error
- Training timeout or GPU resource unavailability on self-hosted infrastructure
- Duplicate payment hash or replayed request — cached result returned rather than re-trained

## How this service works

Self-hosted personal-assistant paid tools via Lightning. Uses x402-shaped protocol (L402 lineage): HTTP 402 with a Lightning invoice in the response body + WWW-Authenticate header, retry with X-Payment-Hash: <hex> once paid. Results cached per payment_hash so legitimate retries don't cost twice.

## Output

Returns a training job result object containing the LoRA model artifact or a reference to the trained weights, along with job status and any relevant metadata. Results are cached per payment hash so retrying after payment does not incur double billing.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "trigger_word": {
   "type": "string",
   "description": "(optional, 3-24 chars)"
  },
  "training_images": {
   "type": "string",
   "description": "(15-100 http(s) URLs)"
  },
  "content_attestation": {
   "type": "string",
   "description": "(must be true)"
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "object",
 "additionalProperties": true
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/alfred-x402-sdxl-lora-training-service-9ee663e7/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from x402.wallace.us](https://www.zero.xyz/host/x402.wallace.us/llms.txt)