Execution via API

Invoke an assistant, continue conversations and select reasoning effort through the public API.

Send messages to POST /api/v1/assistants/{identifier}/messages with a valid API key. Get the assistant identifier from the console’s Get Code dialog or the assistant listing endpoint.

Start a conversation

This JavaScript example uses asynchronous mode. The response includes the chatUid that identifies the new conversation:

const baseUrl = 'https://api.devic.ai';
const assistantId = process.env.DEVIC_ASSISTANT_ID;
const headers = {
  Authorization: `Bearer ${process.env.DEVIC_API_KEY}`,
  'Content-Type': 'application/json',
};

const response = await fetch(
  `${baseUrl}/api/v1/assistants/${assistantId}/messages?async=true`,
  {
    method: 'POST',
    headers,
    body: JSON.stringify({
      message: 'Review this proposal',
      conversationReasoningEffort: 'low',
      reasoningEffort: 'high',
    }),
  },
);
if (!response.ok) throw new Error(`Request failed: ${response.status}`);
const result = await response.json();
const { chatUid } = result.data ?? result;

Choose effort levels supported by the assistant’s model, or omit them to inherit its defaults. See reasoning effort for model discovery, inheritance and resets.

Continue a conversation

Include the same chatUid on subsequent messages:

await fetch(`${baseUrl}/api/v1/assistants/${assistantId}/messages?async=true`, {
  method: 'POST',
  headers,
  body: JSON.stringify({
    chatUid,
    message: 'Explain the main risks',
  }),
});

The saved conversation effort remains low; the earlier message’s high override applied only to that turn. To restore the assistant default, include conversationReasoningEffort: null in a subsequent request.

Read the result

Use the realtime endpoint to follow asynchronous progress, then read the saved conversation history. Without async=true, the message endpoint waits for completion and returns the messages produced by the turn. API responses use the standard data envelope.

The CLI handles asynchronous polling by default:

devic assistants chat "$DEVIC_ASSISTANT_ID" -m "Review this proposal" \
  --conversation-reasoning-effort low --reasoning-effort high
devic assistants chats get "$DEVIC_ASSISTANT_ID" "$CHAT_UID"

Parameters

Field	Description
message	Text to send.
chatUid	Existing conversation identifier; omit to start a new one.
reasoningEffort	Effort for this user turn. Null or omission inherits defaults.
conversationReasoningEffort	Saved conversation effort. Null restores assistant inheritance.
model, provider	Optional model/provider overrides; the effort must be compatible.
metadata	Custom metadata for your integration.
tenantId, subtenantId	Tenant scope, when using multi-tenancy.

See the message API reference for the full schema, file attachments, response types and errors. Keep API keys on your server; use tenant sessions for browser integrations.