Skip to content
PifeatureOther

Get A24 movie recommendation via search and memory

Session from agent_traces on Hugging Face by merve · Apache-2.0 · condensed by Coders Talk

1msession
0interventions
by library · 1d
GOAL

Merve asked her assistant to recommend a good A24 studio movie and settled on one to watch.

  1. Prompt · +0m
    Can you recommend me a good movie from A24 studio
  2. Agent did · +0m
    Searched memory and web for A24 movie recommendations, tried fetching the A24 films catalog page, hit a web search API error and an unclear fetch result
  3. Fail · +0m
    Web search API returned a 503 high-demand error, and the A24 site fetch did not return a usable catalog
  4. Agent did · +0m
    Fell back to recommending A24 films from memory, grouped by mood: Everything Everywhere All at Once, The Green Knight, Hereditary, Midsommar, Lady Bird, The Whale
  5. Prompt · +1m
    I'll do Green Knight, fanxx
  6. Agent did · +1m
    Searched memory for viewing preferences and streaming habits, then answered with where to watch The Green Knight (Max, Kanopy, digital rental) and a viewing tip
  7. Outcome · +1m
    Session ended with Merve picking The Green Knight and receiving streaming options to watch it
VERDICT

I'd probably just ask directly for a shortlist with where-to-watch info up front, since the web search failed and it had to fall back on memory anyway. The fallback to memory-based recommendations was what actually saved the answer when the search tools errored out.

I tried this · 0

Discussion · 0

How Builds work