LlamaIndex Tool: iLoveVideoEditor

This tool allows LlamaIndex agents to render MP4 videos from declarative VideoJSON specifications using the iLoveVideoEditor cloud rendering API.

Features

• 🎬 Programmatic video rendering: Turn VideoJSON specs into MP4 videos via a single tool call
• ⏳ Wait or queue: Poll until the render finishes, or queue and check status later
• 🤖 Agent-ready: Works directly with LlamaIndex agents via BaseToolSpec
• 🛡️ Error handling: Graceful error handling with informative messages

Requirements

• Python >= 3.10
• llama-index-core >= 0.13.0
• requests >= 2.31.0
• An iLoveVideoEditor API key (get one from the dashboard)

Installation

pip install llama-index-tools-ilovevideoeditor

Required Environment Variables

ILOVEVIDEOEDITOR_API_KEY=vf_live_...

Usage

from llama_index.tools.ilovevideoeditor import ILoveVideoEditorTool
from llama_index.core.agent import ReActAgent

# Initialize the tool spec (reads ILOVEVIDEOEDITOR_API_KEY from the environment)
tool_spec = ILoveVideoEditorTool()

# Convert the spec to a list of FunctionTools
tools = tool_spec.to_tool_list()

# Create an agent with the rendering tools
agent = ReActAgent.from_tools(tools, verbose=True)

# The agent can now render videos from VideoJSON specs
agent.chat(
    "Render a 3-second 1080p video with the text 'Hello from LlamaIndex' "
    "using the iLoveVideoEditor tool."
)

Direct usage without an agent:

from llama_index.tools.ilovevideoeditor import ILoveVideoEditorTool

tool = ILoveVideoEditorTool()

result = tool.render_video(
    {
        "name": "hello",
        "layers": [
            {
                "type": "text",
                "settings": {
                    "startTime": 0,
                    "duration": 3,
                    "text": "Hello from LlamaIndex",
                    "fontSize": 64,
                    "color": "#ffffff",
                },
            }
        ],
    }
)
# '{"jobId": "425ba18a-...", "status": "completed", "downloadUrl": "https://..."}'

Queue without waiting, then check status later:

queued = tool.render_video(video_json, wait_for_completion=False)
# '{"jobId": "425ba18a-...", "status": "queued"}'

status = tool.get_render_status("425ba18a-...")
# '{"jobId": "425ba18a-...", "status": "completed", "downloadUrl": "https://..."}'

Configuration options

• api_key: API key. Defaults to the ILOVEVIDEOEDITOR_API_KEY environment variable.
• base_url: API base URL. Defaults to the ILOVEVIDEOEDITOR_API_BASE environment variable or https://api.ilovevideoeditor.com.
• max_wait_seconds: Maximum time to poll for render completion (default 300).
• poll_interval_seconds: Interval between status polls (default 2).

API reference

For the full VideoJSON schema and REST API reference, see the iLoveVideoEditor API docs.

Examples

See the examples directory for more usage examples.

Development

Run tests:

make test

Run linters:

make lint

Format code:

make format