Skip to content

Streaming responses for Cloud Functions #10607

Description

@dblythy

New Feature / Enhancement Checklist

Current Limitation

Cloud Functions buffer everything and return a single JSON response. A function that proxies an LLM cannot stream tokens to the client, so the user stares at a spinner until the full completion is done. There is no workaround inside Parse itself.

Feature / Enhancement Description

A streaming variant of Cloud Functions, most likely SSE. Either a new define API or a stream handle on the request, e.g. the function writes chunks as they arrive and the client consumes them incrementally. SDK support can come later; SSE is consumable with plain fetch/EventSource from day one.

Example Use Case

  1. Define a streaming cloud function that calls an LLM with streaming enabled
  2. Client calls the function
  3. Tokens arrive at the client as the model produces them

Alternatives / Workarounds

  • Mount a separate Express route next to Parse Server (loses auth/session integration, Parse.Cloud context, etc.)
  • Polling or LiveQuery hacks writing partial output to an object

3rd Party References

  • Supabase Edge Functions support streamed responses
  • Vercel AI SDK / most LLM APIs are streaming-first

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions