The Platform API

Agent / Chat

Manage your videos, ask questions, and build applications with the Platform API. Explore the operations below for request details, responses, and examples.

Agent / Chat

Send messages to the AI agent for analysis. The agent has access to project videos and documents, and can examine frames, compare visuals, and search project documents to answer your question.

POST/api/v1/projects/:projectId/agent/processSend message for AI processing

Send message for AI processing

Request body:

{
  "message": "What companies signed the NDA?",
  "streaming": true,
  "useUnifiedSSE": true,
  "sessionId": "unique-session-id",
  "videoId": ["VIDEO_ID_1"]
}

message (required) — the user query

streaming — enable streaming response via SSE

useUnifiedSSE — use unified SSE channel (recommended)

sessionId — session ID for multi-turn memory

videoId — specific video(s). If omitted, all project videos are used.

arrayId — use a video array instead of individual IDs

curl -X POST https://platform.endlessai.com/api/v1/projects/PROJECT_ID/agent/process \
  -H "X-Api-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Describe what happens in the first 10 seconds",
    "sessionId": "session-001"
  }'

import requests

resp = requests.post(
    f"{BASE}/projects/{PROJECT_ID}/agent/process",
    headers={"X-Api-Key": API_KEY, "Content-Type": "application/json"},
    json={
        "message": "Describe what happens in the first 10 seconds",
        "sessionId": "session-001"
    },
    timeout=300
)
print(resp.json())

const res = await fetch(
  `${BASE}/projects/${projectId}/agent/process`,
  {
    method: "POST",
    headers: { "X-Api-Key": API_KEY, "Content-Type": "application/json" },
    body: JSON.stringify({
      message: "Describe what happens in the first 10 seconds",
      sessionId: "session-001"
    })
  }
);
const result = await res.json();

Response shape (non-streaming):

{
  "success": true,
  "sessionId": "session-001",
  "summary": {
    "iterations": 1,
    "processingTimeMs": 4200
  },
  "timestamp": "2026-02-23T10:30:00.000Z"
}

summary gives you iteration count, token totals, and processing time for the run.

POST/api/v1/projects/:projectId/agent/askAsk a question and poll for the answer

Ask a question and poll for the answer

Starts a run and returns 202 with a run_id to poll. Pass video_id to ask about one video; otherwise the project's videos are used. One run at a time per credential — a second ask while one is running returns 409 with the active run_id. Returns 402 with code: "budget_stopped" when the project's token budget is paused.

curl -X POST https://platform.endlessai.com/api/v1/projects/PROJECT_ID/agent/ask \
            -H "X-Api-Key: YOUR_KEY" \
            -H "Content-Type: application/json" \
            -d '{"question": "When did the first truck arrive?", "video_id": "VIDEO_ID"}'

          # Then poll the run
          curl https://platform.endlessai.com/api/v1/projects/PROJECT_ID/agent/runs/RUN_ID \
            -H "X-Api-Key: YOUR_KEY"

import time, requests

          headers = {"X-Api-Key": API_KEY}
          run = requests.post(
              f"{BASE}/projects/{PROJECT_ID}/agent/ask",
              headers=headers,
              json={"question": "When did the first truck arrive?", "video_id": VIDEO_ID},
          ).json()

          while True:
              result = requests.get(
                  f"{BASE}/projects/{PROJECT_ID}/agent/runs/{run['run_id']}",
                  headers=headers,
              ).json()
              if result["status"] != "running":
                  break
              time.sleep(5)
          print(result.get("answer") or result.get("error"))

const headers = { "X-Api-Key": API_KEY, "Content-Type": "application/json" };
          const { run_id } = await (await fetch(
            `${BASE}/projects/${projectId}/agent/ask`,
            {
              method: "POST",
              headers,
              body: JSON.stringify({ question: "When did the first truck arrive?", video_id: videoId }),
            }
          )).json();

          let result;
          do {
            await new Promise((r) => setTimeout(r, 5000));
            result = await (await fetch(
              `${BASE}/projects/${projectId}/agent/runs/${run_id}`,
              { headers }
            )).json();
          } while (result.status === "running");
          console.log(result.answer ?? result.error);

// Response (202 Accepted)
          {
            "run_id": "RUN_ID",
            "status": "running",
            "poll": "/api/v1/projects/PROJECT_ID/agent/runs/RUN_ID"
          }

GET/api/v1/projects/:projectId/agent/runs/:runIdGet an ask run

Get an ask run

status is running, complete, error or timeout. A run is kept for one hour. A completed run may later gain a correction field; the original answer is never changed.

{
            "run_id": "RUN_ID",
            "project_id": "PROJECT_ID",
            "status": "complete",
            "question": "When did the first truck arrive?",
            "answer": "The first truck backs into bay 2 at 00:04:12.",
            "error": null,
            "started_at": "2026-10-01T09:30:00.000Z",
            "completed_at": "2026-10-01T09:30:41.000Z"
          }

POST/api/v1/qna/agentAsk a question about a video, collection or video list

Ask a question about a video, collection or video list

Project key only; answers within the request. Send question and exactly one of videoId, arrayId or videoIds. Reuse the returned sessionId to ask follow-up questions with the same memory.

// Request
          {
            "question": "Did any forklift enter the pedestrian lane?",
            "arrayId": "ARRAY_ID",
            "sessionId": "session-001"
          }

          // Response
          {
            "success": true,
            "sessionId": "session-001",
            "content": "Yes — twice: at 00:12 and at 03:47.",
            "summary": { "iterations": 4, "processingTimeMs": 18342 },
            "timestamp": "2026-10-01T09:30:00.000Z"
          }