#Hi, I'm getting a consistent 500 Internal Server Error on POST /v1/convai/agents/{agent_id}/simulat

1 messages · Page 1 of 1 (latest)

hexed cradleBOT
#
AI-generated Answer for bubbly_armadillo_40021

Short answer

  • Availability on Creator: Yes. ElevenLabs’ API is included in all plans (including Creator), and the Agents “Simulate conversation” endpoints are part of the public API. (help.elevenlabs.io)
  • Known issue right now: No. The ElevenLabs status page currently reports “We’re fully operational.” (status.elevenlabs.io)

Quick checks (to avoid 500s)

  • Endpoint and base URL: POST https://api.elevenlabs.io/v1/convai/agents/{agent_id}/simulate-conversation (note the hyphens in simulate-conversation). (elevenlabs.io)
  • Auth header: Include your API key in the xi-api-key header. Example: -H "xi-api-key: $ELEVENLABS_API_KEY". (elevenlabs.io)
  • Minimal valid body: Provide simulation_specification with at least simulated_user_config, e.g.:
    { "simulation_specification": { "simulated_user_config": { "first_message": "Hello, how can I help you today?", "language": "en" } } } (elevenlabs.io)
  • If you still see errors: Try the streaming variant POST /v1/convai/agents/{agent_id}/simulate-conversation/stream, which surfaces validation errors as 422s more explicitly. (elevenlabs.io)

Background

  • The Simulate Conversation endpoints were publicly introduced and documented in May 2025; current docs show the hyphenated paths. If you’re using an older, underscore path from early changelogs, switch to the hyphenated version. (elevenlabs.io)

If you can share one failing request-id and the exact JSON you’re sending, I can compare it field-by-field against the spec from the docs above.

Sources:
How much does it cost to use the API? – ElevenLabs
ElevenLabs Status
Simulate conversation | ElevenLabs Documentation
API Authentication | ElevenLabs Documentation
Simulate conversation | ElevenLabs Documentation
Stream simulate conversation | ElevenLabs Documentation
May 12, 2025 | ElevenLabs Documentation

heady mulch
#

Le probleme c'est que prompt provoque un 500, et sans prompt le user simule fait n'importe quoi. C'est peut-etre un bug dans l'API simulation. Poste ca sur le Discord ElevenLabs :

▎ I got the simulate-conversation endpoint working (200 OK) but the simulated_user_config.prompt field causes a 500 error. Without it, the simulated user ignores first_message and persona — it behaves
generically instead of following the scenario.

â–Ž Working payload (but user doesn't follow instructions):
â–Ž {"simulation_specification": {"simulated_user_config": {"first_message": "Urgence!", "language": "fr"}, "dynamic_variables": {...}}}

â–Ž Broken payload (500):
▎ {"simulation_specification": {"simulated_user_config": {"prompt": "Tu es un client stressé...", "language": "fr"}, "dynamic_variables": {...}}}

â–Ž Agent ID: agent_4901kmjw1tq2fjmah9p0nhrsvg81, plan: creator