Read-only Perplexity embedding workflow. It creates embeddings for explicit caller-provided text samples and summarizes token usage, output shape, and readiness for semantic search or RAG. It never sends files, calls contextualized embeddings, creates Agent API responses, opens browser sessions, or uses arbitrary Perplexity routes.

How it works 6 steps

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

Process (6 steps)

■ embed ?

■ vectors default

Apply default

■ vector_count count

Count items in vectors

■ summarize ?

■ vector_table to_table

Format results as a data table

■ card to_summary

Format results as a summary card

Settings 7 configurable

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

focus setting = Summarize whether the selected Perplexity embedding setup is ready for semantic search or RAG. Cover model, dimensions, encoding, returned vector rows, token usage, cost metadata when present, and practical next steps.

texts setting =

dimensions setting = 1024

max length setting = 800

perplexity setting =

embedding model setting = pplx-embed-v1-0.6b

encoding format setting = base64_int8