@mistflow-ai/runtime
v0.2.0
Published
Runtime helpers for Mistflow-built apps. Imported by scaffolded apps for hybrid search, batched embedding, and other AI-app utilities.
Downloads
350
Readme
@mistflow-ai/runtime
Runtime helpers for Mistflow-built apps. Imported by scaffolded apps.
What's in here
| Subpath | Helper | Purpose |
|---|---|---|
| @mistflow-ai/runtime/search | hybridSearch(), fuseHybridSearchResults(), checkHybridSearchIndexes() | Hybrid pgvector + tsvector retrieval with RRF fusion |
| @mistflow-ai/runtime/embed | embedAtScale() | Batched, retried, budget-aware embedding via AI Gateway |
| @mistflow-ai/runtime/nango | verifyNangoWebhook(), redaction helpers | Signed webhook verification + Connected Service helpers |
Cron + queue handlers in scaffolded apps use Mistflow's platform-managed cron path (see
lib/cron.tsin the customer scaffold) — not direct QStash signing. A previous spike of@mistflow-ai/runtime/qstashwas removed because it duplicated and conflicted with the platform cron service.
Architecture position
- Scaffolders (
@mistflow-ai/mcp) — emit code into customer apps - Runtime (
@mistflow-ai/runtime, this package) — provides the algorithms customer apps run - Templates (inside scaffolders as string emitters) — per-integration code customers will adapt
Library code lives here. Integration-specific code stays as templates in
the scaffolder. The split matches Vercel's @vercel/blob + scaffolders
pattern and Cloudflare's @cloudflare/sandbox + wrangler-templates pattern.
License
Elastic License 2.0 — source-available, anti-compete.
