Read-only Together AI operations workflow. It lists model catalog entries, deployments, and hardware inventory, then summarizes capacity, model-selection, and deployment-health signals. It never uploads models, creates batches, starts fine-tuning, mutates deployments/endpoints/secrets/storage, or calls arbitrary Together AI routes.

How it works 17 steps

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

Process (17 steps)

■ models_raw ?

■ deployments_raw ?

■ hardware_raw ?

■ models_all default

Apply default

■ deployments_all default

Apply default

■ hardware_all default

Apply default

■ models slice

Take a subset of

■ deployments slice

Take a subset of

■ hardware slice

Take a subset of

■ model_count count

Count items in models all

■ deployment_count count

Count items in deployments all

■ hardware_count count

Count items in hardware all

■ summarize ?

■ model_table to_table

Format results as a data table

■ deployment_table to_table

Format results as a data table

■ hardware_table to_table

Format results as a data table

■ card to_summary

Format results as a summary card

Settings 5 configurable

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

focus setting = Summarize Together AI model availability, deployment state, hardware availability, capacity risks, and useful follow-up actions. Use only the Together AI list responses.

row limit setting = 30

max length setting = 900

togetherai setting =

model filter setting =