Read-only Replicate model discovery workflow. It searches public models, collections, and docs through recued-core.replicate, reads available hardware, and summarizes candidate models for a requested use case. It never runs predictions, creates models or deployments, uploads/downloads files, reads webhook secrets, starts trainings, or calls arbitrary Replicate API routes.

How it works 15 steps

Inspect the data fetches, transforms, gates, and output this recipe runs.

Process (15 steps)

■ normalized_query trim

Trim whitespace from setting query

■ search ?

■ hardware ?

■ models default

Apply default

■ collections default

Apply default

■ pages default

Apply default

■ model_count count

Count items in models

■ collection_count count

Count items in collections

■ page_count count

Count items in pages

■ summarize ?

■ model_rows slice

Take a subset of

■ collection_rows slice

Take a subset of

■ model_table to_table

Format results as a data table

■ collection_table to_table

Format results as a data table

■ card to_summary

Format results as a summary card

Settings 6 configurable

Configurable at install. Defaults shown — change them anytime in Recued.

focus setting = Summarize the best Replicate model candidates for this use case. Compare official status, model owner/name, generated descriptions, tags, relevance scores, usage signals, and practical next steps. Do not invent facts outside the Replicate data.

limit setting = 10

query setting =

replicate setting =

row limit setting = 10

max length setting = 900