Skip to main content
This document lists the currently supported URL parameters for the NanoGPT web app UI. These are intended for browser shortcuts, OpenSearch integration, and sharing prefilled prompts.
Note: Only documented parameters are guaranteed to keep working. Undocumented parameters may change without notice.

​
Conversation URLs

Base URL
  • /conversation/new
Aliases
  • /conversation -> redirects to /conversation/new
  • /ask -> redirects to /conversation/new

​
Parameters

  • ask - Prefills the prompt and auto-submits it when the page loads.
    • Use encodeURIComponent() for the value.
    • This is the closest equivalent to a ?q= style parameter.
  • model - Selects a model by model id.
    • If the model is a known image/video/audio model, the mode is switched automatically.
  • mode - Sets the UI mode (only if model does not override it).
    • Allowed values: chat, image, video, audio, voice.
  • online - Enables web search when set to true.
    • Any other value is ignored.
  • source - Attribution/referral tag.
    • source=translate also disables web search for that session.
  • transcriptId - Internal handoff key used by /youtube and /scrape to inject large prompts.
    • Requires a pre-stored value in session storage or the shared content cache; not intended for manual use.

​
Examples

/conversation/new?ask=Hello%20from%20the%20address%20bar

/conversation/new?model=<model-id>&ask=Summarize%20this%20page

/conversation/new?mode=image&model=<image-model-id>

/conversation/new?online=true&ask=What%27s%20new%20today%3F

/conversation/new?source=opensearch&ask=Define%20vector%20search

​
Notes

  • model takes precedence over mode (a model selection can switch the mode).
  • There is no separate q parameter today - use ask instead.
  • Web search provider and depth are not configurable via URL; they follow your Settings.

​
Media URLs

Base URL
  • /media
Mode shortcuts
  • /media/image -> redirects to /media?mode=image
  • /media/video -> redirects to /media?mode=video
  • /media/audio -> redirects to /media?mode=audio

​
Parameters

  • mode - Selects the media mode.
    • Allowed values: image, video, audio.
  • model - Selects a media model by id.
    • If the model matches a known image/video/audio model, mode is switched automatically.
    • When audioMode=stt, model may also be a speech-to-text model id.
  • audioMode - Selects audio sub-mode (only in mode=audio).
    • Allowed values: tts, stt.
  • prompt - Prefills the prompt and is consumed on load (removed from the URL).

​
Examples

/media?mode=image&model=<image-model-id>

/media?mode=video&model=<video-model-id>&prompt=Make%20this%20cinematic

/media?mode=audio&audioMode=stt&model=<stt-model-id>

​
Notes

  • model takes precedence over mode (a model selection can switch the mode).
  • If a model id is unknown or incompatible with the selected mode, it is ignored.
  • For speech-to-text, include both audioMode=stt and a compatible model id.

​
OpenSearch Integration

An OpenSearch descriptor is available at: 
/opensearch.xml
If you are wiring this into a custom browser keyword/search engine, use the conversation URL with ask: 
/conversation/new?source=opensearch&ask={searchTerms}
(Replace {searchTerms} with the URL-encoded user query.)