Quickstart

Host: https://api.memories.ai/datalake/v1 Auth: Authorization: sk-mai-...

This walks the full happy path: create a collection → upload a video → poll the operation → read results. You need an API key first — create one in the Developer Console, then set it here along with the base path:

BASE=https://api.memories.ai/datalake/v1
AUTH="Authorization: sk-mai-xxxx"

Create a collection

Pick a vector model; optionally enable safety detection or face recognition.

COL=$(curl -s -X POST "$BASE/collections" -H "$AUTH" -H "Content-Type: application/json" \
  -d '{"name":"demo","enabled_detectors":["safety_detector"]}' \
  | jq -r .id)

Upload a video by URL

Returns 202 with a video_id and an operation.

R=$(curl -s -X POST "$BASE/videos" -H "$AUTH" -H "Content-Type: application/json" \
  -d '{"collection_id":"'$COL'","source_url":"https://example.com/demo.mp4","fps":1.0}')
VID=$(echo $R | jq -r .video_id); OP=$(echo $R | jq -r .operation)

Poll the operation until done

Indexing runs preprocess → index → derive.

watch -n 5 "curl -s $BASE/operations/$OP -H '$AUTH' | jq '{done, progress}'"

Read the results

Once ready, pull details, summary, captions, safety events, or any time-slice.

curl -s "$BASE/videos/$VID" -H "$AUTH" | jq            # details (duration / AI title / tags)
curl -s "$BASE/videos/$VID/summary" -H "$AUTH"          # AI summary
curl -s "$BASE/videos/$VID/caption" -H "$AUTH"          # full captions
curl -s "$BASE/videos/$VID/events" -H "$AUTH"           # safety events (if a detector is on)
curl -s "$BASE/moments/$VID@10.0-20.0?expand=caption,frame,clip" -H "$AUTH"   # any time-slice

The same flow in Python

import time, requests

BASE = "https://api.memories.ai/datalake/v1"
HEADERS = {"Authorization": "sk-mai-..."}

col = requests.post(f"{BASE}/collections", headers=HEADERS,
    json={"name": "demo", "enabled_detectors": ["safety_detector"]}).json()

r = requests.post(f"{BASE}/videos", headers=HEADERS,
    json={"collection_id": col["id"], "source_url": "https://example.com/demo.mp4", "fps": 1.0}).json()
vid, op = r["video_id"], r["operation"]

while True:                                        # poll to done
    o = requests.get(f"{BASE}/operations/{op}", headers=HEADERS).json()
    if o["done"]:
        break
    time.sleep(5)

print(requests.get(f"{BASE}/videos/{vid}/summary", headers=HEADERS).json())
print(requests.get(f"{BASE}/moments/{vid}@10.0-20.0?expand=caption,frame", headers=HEADERS).json())

Next steps

Search

Query the collection and get moments back.

Live Stream

Attach an rtsp/rtmp source and search it live.

Webhooks

Get notified instead of polling.

Errors & Rate Limits

The full error envelope and retry guidance.

Get Started

API Reference

Agents & Tools

Resources

The same flow in Python

Next steps

Search

Live Stream

Webhooks

Errors & Rate Limits

​The same flow in Python

​Next steps

Search

Live Stream

Webhooks

Errors & Rate Limits

The same flow in Python

Next steps