cinescout

Inspiration

Location scouting is one of the most time-consuming parts of indie film pre-production — filmmakers spend hours searching for realistic filming locations, figuring out permit requirements, and benchmarking budgets against comparable productions. We wanted an agent that could do this research automatically and hand back a structured, sourced brief in seconds.

What it does

CineScout takes a scene description (e.g. "a rain-soaked neon alley for a noir chase scene, low-budget indie feature") along with a budget tier and optional region hint, and returns:

• Candidate filming locations with reasoning
• Permit and logistics notes for the region
• Comparable production budgets and precedents
• Real, live sources for every claim

How we built it

CineScout runs a three-step agentic pipeline:

1. Plan — Gemini reads the scene brief and generates a research objective plus targeted keyword search queries.
2. Research — Parallel's Search API runs those queries live against the web, returning real, current results (not cached or hardcoded data).
3. Synthesize — Gemini reads the raw research results and writes a structured, cited production brief.

The app is built in Python with Gradio for the interface, deployed on Hugging Face Spaces, and integrates the Gemini API and Parallel Search API at runtime.

Challenges we ran into

Getting the deployment pipeline working under time pressure meant working through model deprecations (Gemini model name changes), Parallel SDK parameter mismatches, dependency conflicts between Gradio and FastAPI, and free-tier hosting constraints across multiple platforms before landing on Hugging Face Spaces with a Gradio interface.

What we learned

How to orchestrate a multi-step agent where one LLM call's output becomes the next tool's structured input, and how much research quality improves when an agent plans its own search strategy instead of using a single static query.

What's next

Adding image-based location previews, multi-region comparison in a single query, and direct integration with production budgeting tools.

Built With

• gemini
• google-cloud
• google-genai
• gradio
• huggingface-spaces
• parallel
• parallel-search-api
• python