haimaker-models
v0.3.0
Published
Offline JSON-backed haimaker.ai catalog — high-throughput pass-through + auto-routing.
Readme
haimaker-models
Offline JSON-backed haimaker.ai catalog — high-throughput pass-through + auto-routing.
Unofficial / community-maintained snapshot. Edit
data/*.jsonthen runnpm run build:data.
Install
npm install haimaker-modelsQuick start
import { listModels, getModel, getMetadata } from "haimaker-models";
const meta = getMetadata();
const models = listModels({ status: "active" });
const one = getModel(models[0]?.id ?? "");
const tagging = meta.requestTagging;
// tagging?.support === "Specialized (Workflow & Pipeline IDs)"Request tagging
getMetadata().requestTagging describes how Haimaker tracks calls. 0.2.0 support: Specialized (Workflow & Pipeline IDs).
workflow_id, execution job id, and step parameters (payload) — identifiers embedded in the workflow payload
Dataset pin: data-2026.09.24. Package version: 0.3.0.
Updating the catalog
- Edit
data/providers.jsonand/ordata/models.json - Bump
dataVersion/generatedAtindata/metadata.json - Run
npm run build:data - Run
npm testand publish
License
MIT. Not affiliated with Haimaker. Site: https://haimaker.ai
Live registry
loadLatestDataset() reads GET /v1/datasets/haimaker from https://ai-registry.app.x12i.com (AI_REGISTRY_URL, then AI_CATALOG_URL). A successful document is the dataset. Network or validation failure returns the bundled JSON. AI_REGISTRY_SOURCE=bundled skips the request. Sync helpers such as listModels stay on the bundled snapshot.
Monorepo
See docs/MONOREPO.md. Explore in the UI: npm run ui.
