CreativAI — AI Video Intelligence & Visual Search Platform

User Management

Get User Info

Retrieve information about the authenticated user including credits and usage.

Try It Out

Validate API key to enable

⚠️ Please enter and validate your API key in the configuration section at the top of the page to test this endpoint.

Responses

200Successfully retrieved user information

{
  "success": true,
  "data": {
    "user_id": "hHvNk7hN5Kg3UaQEP7XnFOhswBG2",
    "credits": 22.17,
    "total_indexed_hours": 0.25,
    "search_requests": 0,
    "total_clips_analyzed": 38
  },
  "error": null
}

401Unauthorized - Invalid API key

{
  "success": false,
  "data": null,
  "error": "Invalid API key provided"
}

Example Request

curl -X GET "https://creativai-apis.com/api/v2/users/get_users_info" \
  -H "x-api-key: your_api_key_here" \
  -H "Content-Type: application/json"

Collections

Collection Groups

Collection Sharing

Collection Tasks

Video Upload

S3 Transfer

Cloud Storage Import

Indexing

Video Management

Search

Data Plates

Knowledge Extraction

Extract structured data from video segments using natural-language queries. Follow these four steps to go from an indexed collection to a populated data plate.

A data plate is a structured table built from video segments. Start by creating one linked to your indexed collection.

curl -X POST "https://creativai-apis.com/api/v2/data-plates" \
  -H "x-api-key: your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "collection_id": "marketing_videos_a1b2c3d4",
    "name": "Product Feature Analysis"
  }'

Example Response

{
  "success": true,
  "data": {
    "plate_id": "plate_b2c3d4e5-f6g7-8901-hijk-lmnop1234567",
    "collection_id": "marketing_videos_a1b2c3d4",
    "name": "Product Feature Analysis",
    "status": "empty",
    "created_at": "2026-07-05T10:00:00Z"
  }
}

Live Stream

Analyze live video streams in real time. Connect any RTMP-compatible source (encoder, browser camera, IP camera) and get AI-generated insights as the stream runs.

Create a session before attaching a video source. Optionally specify a collection, a live insights query, and the analysis model mode.

curl -X POST "https://creativai-apis.com/api/v2/live-stream/sessions" \
  -H "x-api-key: your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Lobby Camera",
    "collection_name": "Operations Live Feeds",
    "user_query": "detect safety incidents",
    "model": "multimodal"
  }'

Example Response

{
  "data": {
    "session_id": "ls_a1b2c3d4-e5f6-7890-ghij-klmn12345678",
    "status": "created",
    "name": "Lobby Camera",
    "collection_id": "operations_live_feeds_a9b8c7d6",
    "ingest_url": "rtmp://ingest.creativai.io/live/ls_a1b2c3d4",
    "stream_key": "sk_xyz123",
    "model": "multimodal",
    "created_at": "2026-07-05T10:00:00Z"
  }
}

Live Stream Quickstart (Sessions + Polling)

Use this flow for Live Stream integrations. It uses live stream session APIs end-to-end, including readiness and indexing polling.

Step 1

Create session: POST /api/v2/live-stream/sessions

Step 2

Poll readiness: GET /sessions/{session_id}/mediamtx-status

Step 3

Read session state: GET /sessions/{session_id}

Step 4

Poll indexing: GET /sessions/{session_id}/indexing-jobs

Recommended polling cadence

While status is waiting, poll /worker-status and /mediamtx-status every 10 seconds.
Once all_ready is true, start or attach your source and keep polling /sessions/{session_id} for canonical status.
During active or final indexing passes, poll /indexing-jobs every 5–10 seconds.
Use POST /stop, POST /resume, and DELETE /sessions/{session_id} for lifecycle control.

Live Stream Protocol Start Endpoints

Use one of these when you need protocol-specific stream startup instead of generic session lifecycle APIs.

Protocol	Endpoint	Use when
Auto	`POST /api/v2/live-stream/stream`	You want backend auto handling of source/protocol.
RTMP	`POST /api/v2/live-stream/stream/rtmp`	Ingesting from encoders that push RTMP.
RTSP	`POST /api/v2/live-stream/stream/rtsp`	Pulling from IP camera RTSP feeds.
SRT	`POST /api/v2/live-stream/stream/srt`	Low-latency reliable ingest over unstable networks.
HLS	`POST /api/v2/live-stream/stream/hls`	Pulling from .m3u8 stream URLs.
WebRTC	`POST /api/v2/live-stream/stream/webrtc`	Browser/device real-time publish using WHIP/WHEP.
YouTube	`POST /api/v2/live-stream/stream/youtube`	Pulling from YouTube URL (with optional cookies).

Agentic Chat

Agentic Chat lets you query your video collection using natural language. The agent autonomously searches, extracts data, creates plates, and streams its reasoning in real time via SSE. Authentication uses Firebase JWT tokens, not x-api-key.

Create an agentic chat session scoped to your indexed collection. The session holds conversation history and agent state across multiple messages.

⚠️ Agentic Chat uses Firebase ID token (JWT) for authentication — not x-api-key. Obtain it with getIdToken() from the Firebase client SDK.

curl -X POST "https://creativai-apis.com/api/v2/agentic-chat/sessions" \
  -H "Authorization: Bearer <your_firebase_id_token>" \
  -H "Content-Type: application/json" \
  -d '{
    "collection_id": "tv_shows_collection_multimodal",
    "title": "Find food scenes"
  }'

Example Response

{
  "data": {
    "session_id": "sess_xyz789",
    "collection_id": "tv_shows_collection_multimodal",
    "title": "Find food scenes",
    "created_at": "2026-07-05T10:00:00Z",
    "message_count": 0
  }
}

SSE Event Reference (Event → What your app should do)

Event	When it appears	Client action
`thinking`	Model is reasoning	Append to reasoning panel (optional)
`answer_delta`	Token-by-token output	Append streamed text to the current assistant message
`answer`	Final assembled answer payload	Replace partial text with final answer and flush buffer
`node_enter` / `node_exit`	Pipeline step starts/completes	Update step indicator progress UI
`search_feedback_required`	Agent needs relevance guidance	Show feedback chips; then call `POST /resume` with action `search_feedback`
`confirmation_required`	Agent needs explicit approval	Collect user response; resume with action `confirmation_response`
`job_started` / `job_progress` / `job_completed`	Background jobs (KE/YouTube indexing)	Update background job list and progress bars
`youtube_search_candidates_ready`	Candidate selection step is ready	Open candidate picker and continue YouTube flow APIs
`visualization_ready`	Charts/visuals generated	Attach chart ids / visualization payload to message UI
`state_sync`	Canonical session state checkpoint	Replace local context/jobs/token usage with server state
`complete`	Normal end of stream	Stop loader and keep the stream result as final
`error`	Stream/runtime error event	Show error state and offer reconnect/resume

YouTube Search (Agentic)

Data Plate Verification

User Management

Get User Info

Retrieve information about the authenticated user including credits and usage.

Try It Out

Validate API key to enable

⚠️ Please enter and validate your API key in the configuration section at the top of the page to test this endpoint.

Responses

200Successfully retrieved user information

{
  "success": true,
  "data": {
    "user_id": "hHvNk7hN5Kg3UaQEP7XnFOhswBG2",
    "credits": 22.17,
    "total_indexed_hours": 0.25,
    "search_requests": 0,
    "total_clips_analyzed": 38
  },
  "error": null
}

401Unauthorized - Invalid API key

{
  "success": false,
  "data": null,
  "error": "Invalid API key provided"
}

Example Request

curl -X GET "https://creativai-apis.com/api/v2/users/get_users_info" \
  -H "x-api-key: your_api_key_here" \
  -H "Content-Type: application/json"

Collections

Collection Groups

Collection Sharing

Collection Tasks

Video Upload

S3 Transfer

Cloud Storage Import

Indexing

Video Management

Search

Data Plates

Knowledge Extraction

Extract structured data from video segments using natural-language queries. Follow these four steps to go from an indexed collection to a populated data plate.

A data plate is a structured table built from video segments. Start by creating one linked to your indexed collection.

curl -X POST "https://creativai-apis.com/api/v2/data-plates" \
  -H "x-api-key: your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "collection_id": "marketing_videos_a1b2c3d4",
    "name": "Product Feature Analysis"
  }'

Example Response

{
  "success": true,
  "data": {
    "plate_id": "plate_b2c3d4e5-f6g7-8901-hijk-lmnop1234567",
    "collection_id": "marketing_videos_a1b2c3d4",
    "name": "Product Feature Analysis",
    "status": "empty",
    "created_at": "2026-07-05T10:00:00Z"
  }
}

Live Stream

Analyze live video streams in real time. Connect any RTMP-compatible source (encoder, browser camera, IP camera) and get AI-generated insights as the stream runs.

Create a session before attaching a video source. Optionally specify a collection, a live insights query, and the analysis model mode.

curl -X POST "https://creativai-apis.com/api/v2/live-stream/sessions" \
  -H "x-api-key: your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Lobby Camera",
    "collection_name": "Operations Live Feeds",
    "user_query": "detect safety incidents",
    "model": "multimodal"
  }'

Example Response

{
  "data": {
    "session_id": "ls_a1b2c3d4-e5f6-7890-ghij-klmn12345678",
    "status": "created",
    "name": "Lobby Camera",
    "collection_id": "operations_live_feeds_a9b8c7d6",
    "ingest_url": "rtmp://ingest.creativai.io/live/ls_a1b2c3d4",
    "stream_key": "sk_xyz123",
    "model": "multimodal",
    "created_at": "2026-07-05T10:00:00Z"
  }
}

Live Stream Quickstart (Sessions + Polling)

Use this flow for Live Stream integrations. It uses live stream session APIs end-to-end, including readiness and indexing polling.

Step 1

Create session: POST /api/v2/live-stream/sessions

Step 2

Poll readiness: GET /sessions/{session_id}/mediamtx-status

Step 3

Read session state: GET /sessions/{session_id}

Step 4

Poll indexing: GET /sessions/{session_id}/indexing-jobs

Recommended polling cadence

While status is waiting, poll /worker-status and /mediamtx-status every 10 seconds.
Once all_ready is true, start or attach your source and keep polling /sessions/{session_id} for canonical status.
During active or final indexing passes, poll /indexing-jobs every 5–10 seconds.
Use POST /stop, POST /resume, and DELETE /sessions/{session_id} for lifecycle control.

Live Stream Protocol Start Endpoints

Use one of these when you need protocol-specific stream startup instead of generic session lifecycle APIs.

Protocol	Endpoint	Use when
Auto	`POST /api/v2/live-stream/stream`	You want backend auto handling of source/protocol.
RTMP	`POST /api/v2/live-stream/stream/rtmp`	Ingesting from encoders that push RTMP.
RTSP	`POST /api/v2/live-stream/stream/rtsp`	Pulling from IP camera RTSP feeds.
SRT	`POST /api/v2/live-stream/stream/srt`	Low-latency reliable ingest over unstable networks.
HLS	`POST /api/v2/live-stream/stream/hls`	Pulling from .m3u8 stream URLs.
WebRTC	`POST /api/v2/live-stream/stream/webrtc`	Browser/device real-time publish using WHIP/WHEP.
YouTube	`POST /api/v2/live-stream/stream/youtube`	Pulling from YouTube URL (with optional cookies).

Agentic Chat

Create an agentic chat session scoped to your indexed collection. The session holds conversation history and agent state across multiple messages.

⚠️ Agentic Chat uses Firebase ID token (JWT) for authentication — not x-api-key. Obtain it with getIdToken() from the Firebase client SDK.

curl -X POST "https://creativai-apis.com/api/v2/agentic-chat/sessions" \
  -H "Authorization: Bearer <your_firebase_id_token>" \
  -H "Content-Type: application/json" \
  -d '{
    "collection_id": "tv_shows_collection_multimodal",
    "title": "Find food scenes"
  }'

Example Response

{
  "data": {
    "session_id": "sess_xyz789",
    "collection_id": "tv_shows_collection_multimodal",
    "title": "Find food scenes",
    "created_at": "2026-07-05T10:00:00Z",
    "message_count": 0
  }
}

SSE Event Reference (Event → What your app should do)

Event	When it appears	Client action
`thinking`	Model is reasoning	Append to reasoning panel (optional)
`answer_delta`	Token-by-token output	Append streamed text to the current assistant message
`answer`	Final assembled answer payload	Replace partial text with final answer and flush buffer
`node_enter` / `node_exit`	Pipeline step starts/completes	Update step indicator progress UI
`search_feedback_required`	Agent needs relevance guidance	Show feedback chips; then call `POST /resume` with action `search_feedback`
`confirmation_required`	Agent needs explicit approval	Collect user response; resume with action `confirmation_response`
`job_started` / `job_progress` / `job_completed`	Background jobs (KE/YouTube indexing)	Update background job list and progress bars
`youtube_search_candidates_ready`	Candidate selection step is ready	Open candidate picker and continue YouTube flow APIs
`visualization_ready`	Charts/visuals generated	Attach chart ids / visualization payload to message UI
`state_sync`	Canonical session state checkpoint	Replace local context/jobs/token usage with server state
`complete`	Normal end of stream	Stop loader and keep the stream result as final
`error`	Stream/runtime error event	Show error state and offer reconnect/resume

Build with the visual intelligence layer.

User Management

Get User Info

Try It Out

Responses

Example Request

Collections

Collection Groups

Collection Sharing

Collection Tasks

Video Upload

Video Upload & Indexing Flow

S3 Transfer

How to Transfer Videos from Your S3 Bucket

Cloud Storage Import

Indexing

Video Management

Search

Data Plates

Knowledge Extraction

Live Stream

Live Stream Quickstart (Sessions + Polling)

Agentic Chat

YouTube Search (Agentic)

Data Plate Verification

Give your visual dataa structured brain.

Build with the visual intelligence layer.

User Management

Get User Info

Try It Out

Responses

Example Request

Collections

Collection Groups

Collection Sharing

Collection Tasks

Video Upload

Video Upload & Indexing Flow

S3 Transfer

How to Transfer Videos from Your S3 Bucket

Cloud Storage Import

Indexing

Video Management

Search

Data Plates

Knowledge Extraction

Live Stream

Live Stream Quickstart (Sessions + Polling)

Agentic Chat

YouTube Search (Agentic)

Data Plate Verification

Give your visual dataa structured brain.

Give your visual data
a structured brain.

Give your visual data
a structured brain.