Zurück zu den Videos

OpenAI

Introducing the Agents API

Launch
Plattform
YouTube
Länge
2:18
Gesprochene Sprache
Englisch
Aufrufe
220.155
Median-Schnitt
5.92s
Gesicht im Bild
57%
Dunkle Bilder
24%
OpenAI Introducing the Agents API at 0:00medium presenter
OpenAIIntroducing the Agents API
0:00
OpenAI Introducing the Agents API at 0:11close presenter
OpenAIIntroducing the Agents API
0:11
OpenAI Introducing the Agents API at 0:14close presenter
OpenAIIntroducing the Agents API
0:14
OpenAI Introducing the Agents API at 0:14medium presenter
OpenAIIntroducing the Agents API
0:14
OpenAI Introducing the Agents API at 0:16
OpenAIIntroducing the Agents API
0:16
OpenAI Introducing the Agents API at 0:23
OpenAIIntroducing the Agents API
0:23
OpenAI Introducing the Agents API at 0:26close presenter
OpenAIIntroducing the Agents API
0:26
OpenAI Introducing the Agents API at 0:30
OpenAIIntroducing the Agents API
0:30
OpenAI Introducing the Agents API at 0:30
OpenAIIntroducing the Agents API
0:30
OpenAI Introducing the Agents API at 0:32
OpenAIIntroducing the Agents API
0:32
OpenAI Introducing the Agents API at 0:33
OpenAIIntroducing the Agents API
0:33
OpenAI Introducing the Agents API at 0:40medium presenter
OpenAIIntroducing the Agents API
0:40
OpenAI Introducing the Agents API at 0:43
OpenAIIntroducing the Agents API
0:43
OpenAI Introducing the Agents API at 0:44
OpenAIIntroducing the Agents API
0:44
OpenAI Introducing the Agents API at 0:45
OpenAIIntroducing the Agents API
0:45
OpenAI Introducing the Agents API at 0:46
OpenAIIntroducing the Agents API
0:46
OpenAI Introducing the Agents API at 0:46
OpenAIIntroducing the Agents API
0:46
OpenAI Introducing the Agents API at 0:47medium presenter
OpenAIIntroducing the Agents API
0:47
OpenAI Introducing the Agents API at 0:48medium presenter
OpenAIIntroducing the Agents API
0:48
OpenAI Introducing the Agents API at 0:58medium presenter
OpenAIIntroducing the Agents API
0:58
OpenAI Introducing the Agents API at 1:00
OpenAIIntroducing the Agents API
1:00
OpenAI Introducing the Agents API at 1:02
OpenAIIntroducing the Agents API
1:02
OpenAI Introducing the Agents API at 1:06
OpenAIIntroducing the Agents API
1:06
OpenAI Introducing the Agents API at 1:10
OpenAIIntroducing the Agents API
1:10
OpenAI Introducing the Agents API at 1:14
OpenAIIntroducing the Agents API
1:14
OpenAI Introducing the Agents API at 1:18
OpenAIIntroducing the Agents API
1:18
OpenAI Introducing the Agents API at 1:22
OpenAIIntroducing the Agents API
1:22
OpenAI Introducing the Agents API at 1:29medium presenter
OpenAIIntroducing the Agents API
1:29
OpenAI Introducing the Agents API at 1:32close presenter
OpenAIIntroducing the Agents API
1:32
OpenAI Introducing the Agents API at 1:45
OpenAIIntroducing the Agents API
1:45
OpenAI Introducing the Agents API at 1:49
OpenAIIntroducing the Agents API
1:49
OpenAI Introducing the Agents API at 1:56close presenter
OpenAIIntroducing the Agents API
1:56
OpenAI Introducing the Agents API at 2:10medium presenter
OpenAIIntroducing the Agents API
2:10
OpenAI Introducing the Agents API at 2:16
OpenAIIntroducing the Agents API
2:16

Transkript von Introducing the Agents API

Von einem Sprachmodell aus dem Ton des Films transkribiert, nicht von der Marke geschrieben. Klicke auf eine Zeile, um an diese Stelle zu springen.

  1. 0:00Getting a long-running agent into production takes a lot of work, even with a capable model.
  2. 0:05You need to connect tools, track progress, manage context, and secure and maintain the infrastructure around it.
  3. 0:11Today, we're launching the Agents API to handle that infrastructure for you.
  4. 0:15The Agents API brings a hosted version of the Codex Harness to your applications.
  5. 0:20OpenAI handles orchestration, sessions, and context management so that you can stay focused on building.
  6. 0:26Let's look at an example.
  7. 0:28Suppose we want to build an agent that helps investigate incidents in our production stack.
  8. 0:32It would need access to observability data and recent code changes, along with our team's
  9. 0:37instructions for how to handle an outage. We can connect all of our necessary tools through MCPs
  10. 0:43and give the agent our investigation runbook via a skill. You control the agent's execution
  11. 0:49environment and the tools it has access to. That includes connecting a sandbox, whether it's through
  12. 0:54through OpenAI, a third-party provider, or using your own infrastructure.
  13. 0:58In a situation like this, we'd also be working through a ton of logs, likely more than we
  14. 1:04could fit in the model's context window.
  15. 1:06With programmatic tool calling, the agent can process those logs and filter the results
  16. 1:11in code, meaning fewer tokens spent passing around raw data and more spent on the information
  17. 1:16the agent needs.
  18. 1:18For the largest tasks, independent work can be delegated via multi-agent orchestration.
  19. 1:23In this case, we might want one sub-agent to examine recent changes, another to check telemetry, and then have the lead agent unify their findings.
  20. 1:32But even in a single-agent session, long-running context windows still work really well with compaction.
  21. 1:38It gives the model a summary of all prior work completed so the agent can continue its investigation.
  22. 1:44Once the work is complete, the findings should be served as a report that the on-call team can review,
  23. 1:49a likely root cause, the supporting evidence, and suggested next steps
  24. 1:54bundled into a single shareable file.
  25. 1:56And there we have it.
  26. 1:57With the Agent's API, we were able to drive an entire workflow
  27. 2:01without building or maintaining any of our own agent infrastructure.
  28. 2:05Infrastructure that will continue to get better alongside new models and new capabilities.
  29. 2:10We're excited to bring you an ever-improving harness behind one API.
  30. 2:15Happy building.
30 Zeilen · 384 Wörter