--- title: "YouTube Search" short_description: "Search videos by keyword, date and duration." emoji: "🔎" colorFrom: "blue" colorTo: "indigo" sdk: "static" app_file: "index.html" pinned: false license: "mit" tags: - smolagents - tool - flashdata - youtube-search --- # YouTube Search Search YouTube by keyword, upload date, duration and other filters. Return video IDs, titles, channels and available video statistics. [API documentation](https://flashdata.dev/docs/sources/youtube_search?utm_source=hugging-face&utm_medium=integration&utm_campaign=intent_gtm_2026q4&utm_content=youtube-video-search-flashdata) · [Get an API key](https://flashdata.dev/signup?utm_source=hugging-face&utm_medium=integration&utm_campaign=intent_gtm_2026q4&utm_content=youtube_search-task) · [Pricing](https://flashdata.dev/pricing?utm_source=hugging-face&utm_medium=integration&utm_campaign=intent_gtm_2026q4&utm_content=youtube_search-task) Requires your own FlashData API key and paid Credits. The Python tool is free to install. ## Try the inputs Edit the JSON in this Space, generate a Python call, and inspect the included sample response or paste your own JSON. This page runs locally in your browser; it does not run a data query or collect your API key. ## Run with Python Use Python 3.10 or newer: ```sh python -m pip install "smolagents==1.26.0" "langchain-flashdata==0.2.1" ``` Review [tool.py](tool.py) before enabling `trust_remote_code`. Run locally with your key: ```python import getpass import os from smolagents import load_tool os.environ["FLASHDATA_API_KEY"] = getpass.getpass("FlashData API key: ") tool = load_tool("flashdata-dev/youtube-video-search-flashdata", trust_remote_code=True) result = tool(**{'max_results': 10, 'query': 'AI agents introduction'}) print(result) ``` The same loaded tool can be used with smolagents. A direct Python call does not require a model. ## Inputs | Parameter | Required / default | Details | | --- | --- | --- | | `query` | Required | YouTube search query, between 1 and 2,000 characters. | | `max_results` | 10 | Requested number of videos. minimum=1, maximum=20 | | `filters` | None | filters | See [input-schema.json](input-schema.json) for full validation rules. Optional `None` values are omitted; adapter defaults apply. ## Output Read `results.0.results`. Fields: `videoId`, `title`, `channel`, `duration`, `viewCount`. Optional fields may be absent. [example.json](example.json) contains a recorded response and its original request, not a new result for the input shown in the Space. ## Limits and cost Request 1–20 results; this adapter defaults to 10. Actual results may be fewer. Transcript retrieval is separate. Each data request uses FlashData Credits. Check [current pricing](https://flashdata.dev/pricing?utm_source=hugging-face&utm_medium=integration&utm_campaign=intent_gtm_2026q4&utm_content=youtube_search-task) before running. Any model or hosting charges are separate. For an uncertain timeout, check your FlashData Jobs before retrying; this tool does not automatically retry paid submissions.