Chat and Voice

The chat interface is a synchronous interface and does not enter the task list. Points will be deducted immediately according to the background price if the chat is successful, and refunds will be automatically made if the upstream fails.

Send message

curl -X POST https://api.girllab.io/v1/chat/messages \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "message":"Hello",
    "persona":"You are a concise travel assistant.",
    "temperature":0.8,
    "top_p":0.8,
    "max_tokens":2048
  }'

If conversation_id is not passed for the first time, the new session ID will be given in the return value. The context is maintained by passing this ID back on subsequent sends. The personas and parameters passed in for the first time will be saved to the session.

Session Management

Operations	Requests
Session List	GET /v1/chat/conversations
Message History	GET /v1/chat/conversations/{id}/messages
Save settings	PATCH /v1/chat/conversations/{id}/config
Delete session	DELETE /v1/chat/conversations/{id}

The return structure of message history is:

{
  "data": {
    "config": {"persona":"...","temperature":0.8,"top_p":0.8,"max_tokens":2048},
    "messages": [{"id":1,"role":"user","content":"Hello","created_at":"..."}]
  }
}

Speech to text

POST /v1/chat/transcribe, incoming { "audio": "data:audio/webm;base64,..." }. Pure base64 is also supported, with a maximum decoded size of 10MB; text and language are returned synchronously.

Text to Speech

POST /v1/chat/speech, incoming { "text": "Hello", "voice": "Cherry" }. Synchronously return MP3 data URI: { "audio": "data:audio/mpeg;base64,..." }, which can be directly assigned to src of web page <audio>.