Platform concepts for VideoVector

Overview

VideoVector exposes a public workflow surface for teams that need to ingest media, define extraction logic, run prompt-based analysis, search results, and deliver outputs into surrounding systems.

This section explains the platform model from highest level to implementation detail:

• Platform model: indexes, media, prompts, prompt runs, results, connectors, exports, and webhooks.
• Prompt schema design: JSON schema rules, nested fields, repeated objects, reserved names, and semantic indexing controls.
• Prompt execution model: target selection, segmentation modes, transcription, image embeddings, segment-level outputs, and video-level synthesis.
• Search model: text, image, multimodal, filter, multi-run, SQL, and agentic search behavior.
• Workflow automation model: connectors, imports, exports, automations, and event delivery.

When to use this section

Use the concepts pages when you need to answer design questions such as:

• What is the correct unit of organization for a workflow: an index, a prompt, or a prompt run?
• When should a field live in segment-level output versus video-level synthesis?
• Which fields are searchable, filterable, or safe to exclude from semantic indexing?
• When should downstream systems use exports, webhooks, or automations?

What to read next

If you already know the platform model and want implementation steps, go to Guides. If you need endpoint-level details, go to API reference. If you want code-first integration, start with the Python SDK or MCP docs.