API

Key Functions:

• processStream(stream, callbacks) - Processes streaming responses and calls handlers with NEW messages from the stream in Brightsy format
• formatMessages(messages) - Converts Brightsy format messages to Vercel AI SDK format with parts for rendering
• generateUUID() - Generates PostgreSQL-compliant UUIDs for message IDs

Important: processStream doesn't return messages - it calls your handlers (onMessagesUpdate, onComplete) with the NEW messages from the stream. Always combine with your prior message history ([...priorMessages, ...streamedMessages]) in the handlers before formatting for display.

Using formatMessages to Render Messages

The formatMessages function converts Brightsy format messages (with content arrays) to Vercel format messages (with parts arrays) for rendering:

import { formatMessages } from '@brightsy/client';

// Brightsy format messages (from API or processStream)
const brightsyMessages = [
  {
    id: 'msg-1',
    role: 'user',
    content: [{ text: 'Hello!' }],
    createdAt: new Date().toISOString()
  },
  {
    id: 'msg-2',
    role: 'assistant',
    content: [{ text: 'Hi! How can I help you?' }],
    tool_calls: [
      {
        id: 'call-1',
        type: 'function',
        function: {
          name: 'search',
          arguments: '{"query":"example"}'
        }
      }
    ],
    createdAt: new Date().toISOString()
  },
  {
    id: 'msg-3',
    role: 'tool',
    tool_call_id: 'call-1',
    name: 'search',
    content: '{"results":["result1","result2"]}',
    createdAt: new Date().toISOString()
  }
];

// Convert to Vercel format with parts
const formattedMessages = formatMessages(brightsyMessages);

// Render the formatted messages
formattedMessages.forEach(message => {
  console.log(`[${message.role}]:`, message.id);
  
  message.parts.forEach(part => {
    if (part.type === 'text') {
      console.log('  Text:', part.text);
    } else if (part.type === 'tool-invocation') {
      console.log('  Tool:', part.toolInvocation.toolName);
      console.log('  Args:', part.toolInvocation.args);
      if (part.toolInvocation.result) {
        console.log('  Result:', part.toolInvocation.result);
      }
    } else if (part.type === 'reasoning') {
      console.log('  Reasoning:', part.reasoning);
    } else if (part.type === 'source') {
      console.log('  Source:', part.source.url, part.source.title);
    } else if (part.type === 'image_url') {
      console.log('  Image:', part.image_url.url);
    } else if (part.type === 'file') {
      console.log('  File:', part.file);
    }
  });
});

Output:

[user]: msg-1
  Text: Hello!
[assistant]: msg-2
  Text: Hi! How can I help you?
  Tool: search
  Args: { query: "example" }
  Result: { results: ["result1", "result2"] }

Key Points:

• formatMessages merges consecutive messages of the same role and tool/assistant messages
• Tool calls and their results are combined into single tool-invocation parts with state: 'result'
• The function handles text, tool invocations, reasoning, sources, images, and file attachments
• Usage information is accumulated when messages are merged

API Structure

The BrightsyClient uses a consistent property-based API pattern for all resource access:

CMS APIs (properties):

• client.cma - Content Management API
• client.cda - Content Delivery API
• client.cpa - Content Preview API

Collection APIs (properties):

• client.agents - Agent management
• client.scenarios - Scenario management
• client.schedules - Schedule management
• client.webhooks - Webhook management
• client.apps - App management
• client.files - File management
• client.images - Image operations
• client.mcpServers - MCP server management
• client.apikeys - API key management
• client.pageTypes - Page type management
• client.libs - Component library management
• client.sites - Site management

Agent APIs (methods, require IDs):

• client.agent(agentId) - Access a specific agent
• client.scenario(scenarioId) - Access a specific scenario

Chat Completions (Vercel AI SDK-compliant):

• client.chat.completions(agentId) - Stateless completions (provide full message history)
• client.chat.completions(agentId).chat(chatId) - Stateful workspace chat (add messages to existing chat, server maintains history)
  • .create({ message }) - Add a message and get response
  • .getMessages() - Get all messages from the chat
  • .deleteMessage(messageId) - Delete a message from the chat

Stream Processing Utilities:

• processStream(stream, callbacks) - Process streaming responses and accumulate messages in Brightsy format
• formatMessages(messages) - Convert Brightsy format messages to Vercel AI SDK format with parts
• generateUUID() - Generate PostgreSQL-compliant UUIDs

Usage Examples

Chat Completions

// Stateless chat completion
const response = await client.chat.completions('agent-id').create({
  messages: [
    {
      role: 'user',
      content: [
        { text: 'Hello, how can you help me?' }
      ]
    }
  ],
  stream: false,
  tools: [
    {
      type: 'function',
      function: {
        name: 'get_weather',
        description: 'Get the current weather',
        parameters: {
          type: 'object',
          properties: {
            location: { type: 'string' }
          }
        }
      }
    }
  ],
  max_tokens: 16000,
  context: {
    user: { name: 'John' },
    pageData: { /* ... */ }
  }
});

// Response is in Brightsy format - convert to parts for rendering
const formattedMessages = formatMessages([response.choices[0].message]);
formattedMessages.forEach(message => {
  message.parts.forEach(part => {
    if (part.type === 'text') {
      console.log('Text:', part.text);
    } else if (part.type === 'tool-invocation') {
      console.log('Tool:', part.toolInvocation.toolName);
      console.log('Args:', part.toolInvocation.args);
      console.log('Result:', part.toolInvocation.result);
    } else if (part.type === 'reasoning') {
      console.log('Reasoning:', part.reasoning);
    }
  });
});

// Streaming chat completions
const stream = await client.chat.completions('agent-id').create({
  messages: [
    {
      role: 'user',
      content: [{ text: 'Tell me a story' }]
    }
  ],
  stream: true
});

// Process stream using processStream utility
import { processStream, formatMessages } from '@brightsy/client';
import type { ChatMessage } from '@brightsy/client';

// Keep track of prior messages (messages that existed before streaming started)
const priorMessages = [...existingMessages]; // Your existing message history

// Process the stream - handlers are called as messages arrive
await processStream(stream, {
  onMessagesUpdate: (streamedMessages: ChatMessage[]) => {
    // Called whenever new content arrives during streaming
    // streamedMessages contains only the NEW messages from this stream
    // Combine with priorMessages for complete history
    const allMessages = [...priorMessages, ...streamedMessages];
    
    // Convert to parts format for rendering
    const formattedMessages = formatMessages(allMessages);
    
    // Update UI with formattedMessages
    setMessages(formattedMessages); // Update your UI state
  },
  onComplete: (streamedMessages: ChatMessage[]) => {
    // Called when stream completes
    // streamedMessages contains all NEW messages from this stream
    const allMessages = [...priorMessages, ...streamedMessages];
    const formattedMessages = formatMessages(allMessages);
    
    // Final UI update
    setMessages(formattedMessages);
    setStreaming(false);
    console.log('Stream complete, total new messages:', streamedMessages.length);
  },
  onError: (error: Error) => {
    // Called if an error occurs during streaming
    console.error('Stream error:', error);
    setStreaming(false);
  }
});


// Stateful workspace chat - adds messages to existing chat
// The workspace maintains the message history on the server
const chatResponse = await client.chat.completions('agent-id').chat('chat-id').create({
  message: {
    role: 'user',
    content: [{ text: 'Continue our conversation' }]
  },
  stream: false
});

// Get all messages from a workspace chat
const messages = await client.chat.completions('agent-id').chat('chat-id').getMessages();
console.log('Chat messages:', messages);

// Delete a message from workspace chat
await client.chat.completions('agent-id').chat('chat-id').deleteMessage('message-id');

Message Formats:

• Brightsy Format: Used internally and for API requests. Messages have role, content (array), tool_calls, tool_call_id, etc.
• Vercel Format: Used for rendering. Messages have role, parts (array of MessagePart objects). Convert using formatMessages().

Example Message Structure:

// Brightsy format (stored in state, sent to API)
const brightsyMessage = {
  id: 'msg-123',
  role: 'user',
  content: [{ text: 'Hello' }],
  createdAt: new Date().toISOString()
};

// Vercel format (for rendering)
const vercelMessage = formatMessages([brightsyMessage])[0];
// Result: { id: 'msg-123', role: 'user', parts: [{ type: 'text', text: 'Hello' }], ... }

Agents

// List all agents
const agents = await client.agents.list();

// Get a specific agent
const agent = await client.agents.get('agent-id');

CMS (Content Management API)

The CMS API provides three endpoints:

• CDA (Content Delivery API): Returns published records (read-only)
• CMA (Content Management API): Returns draft records with full CRUD operations
• CPA (Content Preview API): Returns draft records (read-only)

// Get published records (CDA)
const records = await client.cda.from('blog-posts')
  .where('status', 'eq', 'published')
  .and('category', 'eq', 'technology')
  .orderBy('created_at', 'desc')
  .page(1)
  .pageSize(10)
  .get();

// Access results
records.data.forEach(record => {
  console.log(record.data.title);
});

// Pagination info
console.log('Total:', records.pagination.total);
console.log('Page:', records.pagination.page, 'of', records.pagination.page_count);

// Create a record (CMA only)
const newRecord = await client.cma.from('blog-posts')
  .save({
    data: {
      title: 'My Blog Post',
      content: 'Post content here'
    }
  });

// Update a record (CMA only)
await client.cma.from('blog-posts')
  .id('record-id')
  .save({
    data: { title: 'Updated Title' }
  });

// Delete a record (CMA only)
await client.cma.from('blog-posts')
  .id('record-id')
  .delete();

// Advanced filtering with grouping
const records = await client.cda.from('articles')
  .where('status', 'eq', 'published')
  .andOpenGrouping()
    .where('featured', 'eq', true)
    .or('views', 'gt', 1000)
  .closeGrouping()
  .get();

// Field selection
const records = await client.cda.from('articles')
  .select('title', 'summary', 'author')
  .get();

// Filter by tags
const records = await client.cda.from('articles')
  .tags('featured', 'technology')
  .where('status', 'eq', 'published')
  .get();

// Aggregations
const stats = await client.cda.from('articles')
  .where('status', 'eq', 'published')
  .count()
  .sum('views')
  .avg('rating')
  .get();

// Aggregations with grouping
const categoryStats = await client.cda.from('articles')
  .groupBy('category', 'author')
  .where('status', 'eq', 'published')
  .count()
  .sum('views')
  .get();

// Results will be grouped by category and author
categoryStats.data.forEach(group => {
  console.log(`Category: ${group.category}, Author: ${group.author}`);
  console.log(`Count: ${group.count}`);
  console.log(`Total views: ${group.views_sum}`);
});

File Manager

// Upload a file
const file = await client.files.upload({
  file: fileBuffer,
  filename: 'document.pdf',
  path: 'documents/'
});

// List files
const files = await client.files.list({
  path: 'documents/'
});

// Delete a file
await client.files.delete('documents/document.pdf');

Files

// Upload a file
const file = await client.files.upload({
  file: fileBuffer,
  filename: 'document.pdf',
  path: 'documents/'
});

// List files
const files = await client.files.list({
  path: 'documents/'
});

// Delete a file
await client.files.delete('documents/document.pdf');

Images

// Analyze an image
const analysis = await client.images.analyze({
  imageUrl: 'https://example.com/image.jpg',
  prompt: 'Describe this image'
});

// Generate an image
const image = await client.images.generate({
  prompt: 'A sunset over mountains'
});

Scenarios

// List scenarios
const scenarios = await client.scenarios.list();

// Get a scenario
const scenario = await client.scenarios.get('scenario-id');

// Execute a scenario (uses method since it requires scenarioId)
const result = await client.scenario('scenario-id').execute({
  input: 'Process this data'
});

Sites

// List sites
const sites = await client.sites.list();

// Get a site
const site = await client.sites.get('site-id');

// List pages for a site
const pages = await client.sites.pages('site-id').list();

// Get a page
const page = await client.sites.pages('site-id').get('page-id');

Other Collections

// Schedules
const schedules = await client.schedules.list();
const schedule = await client.schedules.get('schedule-id');

// Webhooks
const webhooks = await client.webhooks.list();
const webhook = await client.webhooks.get('webhook-id');

// Apps
const apps = await client.apps.list();
const app = await client.apps.get('app-id');

// MCP Servers
const servers = await client.mcpServers.list();
const server = await client.mcpServers.get('server-id');

// API Keys
const keys = await client.apikeys.list();
const key = await client.apikeys.get('key-id');

// Page Types
const pageTypes = await client.pageTypes.list();
const pageType = await client.pageTypes.get('page-type-id');

// Component Libraries
const libs = await client.libs.list();
const lib = await client.libs.get('lib-id');

Endpoints

Agents

Agents are AI-powered assistants that can be accessed through the Agents section. They support chat completions, multi-turn conversations, and file attachments.

Attachments

Delete an Attachment

DELETE /agents/{agentId}/chat/completion/workspace/{chatId}/messages/{messageId}/attachments/{filename}

Deletes a specific attachment from a message.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)
• messageId - The unique identifier of the message (string, required)
• filename - The name of the file to delete (string, required)

Responses:

• 200 OK - File deleted successfully
• 204 No Content - For CORS preflight requests

Upload an Attachment

File uploads use a two-step process for security and efficiency:

Step 1: Get Upload URL

POST /agent/{agentId}/upload-url

Gets a signed upload URL for secure file upload.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema:

{
  "filename": "example.pdf",
  "contentType": "application/pdf",
  "messageId": "msg_123456789",
  "chatId": "chat_987654321"
}

Properties:

• filename - The name of the file to upload (string, required)
• contentType - The MIME type of the file (string, optional)
• messageId - The unique identifier of the message (string, required)
• chatId - The unique identifier of the chat (string, required)

Responses:

• 200 OK - Returns upload URL and file information

{
  "uploadUrl": "https://supabase-storage-url.com/signed-upload-url",
  "fileUrl": "https://attachments.brightsy.ai/account/agents/agent-id/chats/chat-id/messages/message-id/filename.pdf",
  "path": "account/agents/agent-id/chats/chat-id/messages/message-id/filename.pdf"
}

Step 2: Upload File to Signed URL

PUT {uploadUrl}

Upload the file directly to the signed URL returned from Step 1.

Request Body:

• Content Type: The content type specified in Step 1
• Body: Raw file binary data

Responses:

• 200 OK - File uploaded successfully

Complete Upload Example:

// Step 1: Get upload URL
const uploadResponse = await fetch('/api/v1beta/agent/agent-123/upload-url', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    filename: 'document.pdf',
    contentType: 'application/pdf',
    messageId: 'msg-456',
    chatId: 'chat-789'
  })
});

const { uploadUrl, fileUrl } = await uploadResponse.json();

// Step 2: Upload file to signed URL
const fileUploadResponse = await fetch(uploadUrl, {
  method: 'PUT',
  headers: {
    'Content-Type': 'application/pdf'
  },
  body: fileBuffer
});

// File is now accessible at fileUrl

Messages

Create or Update Message

POST /agent/{agentId}/chat/completion/workspace/{chatId}/message/{messageId}

Creates or updates a message within a chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)
• messageId - The unique identifier of the message (string, required)

Request Body:

• Content Type: application/json
• Schema: A Message object

Responses:

• 200 OK - Message created/updated successfully

Delete Message

DELETE /agent/{agentId}/chat/completion/workspace/{chatId}/message/{messageId}

Deletes a specific message.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)
• messageId - The unique identifier of the message (string, required)

Responses:

• 200 OK - Message deleted successfully
• 204 No Content - For CORS preflight requests

List Messages

GET /agent/{agentId}/chat/completion/workspace/{chatId}/message

Retrieves a list of messages within a chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Responses:

• 200 OK - Returns a list of messages

Add Message to Chat

POST /agent/{agentId}/chat/completion/workspace/{chatId}/message

Creates a new message within a chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Request Body:

• Content Type: application/json
• Schema: A Message object

Responses:

• 200 OK - Message created successfully
• 204 No Content - For CORS preflight requests

Chats

Get Chat Details

GET /agent/{agentId}/chat/completion/workspace/{chatId}/message

Retrieves details of a specific chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Responses:

• 200 OK - Returns chat details

Create Chat

POST /agent/{agentId}/chat/completion/workspace/{chatId}/message

Creates a new message and chat if needed.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Request Body:

• Content Type: application/json
• Schema: An object containing a message property of type string

Responses:

• 200 OK - Chat created successfully
• 204 No Content - For CORS preflight requests

Agent Messages

Create Workspace Chat

POST /agent/{agentId}/chat/completion/workspace

Creates a message for an agent and starts a new workspace-based chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema: A Message object

Responses:

• 200 OK - Message created successfully
• 204 No Content - For CORS preflight requests

Create Completion

POST /agents/{agentId}/chat/completion

Creates a completion where the client provides the entire messages array.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema:

{
  "messages": [
    {
      "role": "user",
      "content": "Hello, how can you help me?"
    }
  ],
  "context": {
    "user": { "name": "John" },
    "customData": { /* any data accessible to the agent */ }
  },
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get the current weather for a location",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {
              "type": "string",
              "description": "The city name"
            }
          },
          "required": ["location"]
        }
      }
    }
  ],
  "max_tokens": 16000
}

Properties:

• messages - Array of message objects with role and content (array, required)
• context - Additional context data available to the agent (object, optional)
• tools - Array of function tools the agent can call (array, optional)
• max_tokens - Maximum tokens for the response, overrides agent default (number, optional)

Responses:

• 200 OK - Completion generated successfully
• 204 No Content - For CORS preflight requests

Note: This endpoint does not store the conversation in Brightsy. No chatId is used as the conversation is managed by the client.

Cancel Chat Completion

POST /agent/{agentId}/chat/completion/cancel

Cancels an ongoing chat completion request.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema:

{
  "request_id": "req_123456789",
  "chat_id": "chat_987654321",
  "reason": "user_cancellation"
}

Properties:

• request_id - The unique identifier of the request to cancel (string, required)
• chat_id - The unique identifier of the chat (string, optional - only needed for workspace-based chats)
• reason - The reason for cancellation (string, optional, defaults to "user_cancellation")

Responses:

• 200 OK - Request cancelled successfully

{
  "message": "Request cancelled successfully",
  "cancelled_at": "2024-01-15T10:30:00.000Z",
  "request_id": "req_123456789"
}

• 400 Bad Request - Invalid request parameters
• 401 Unauthorized - User does not have access to agent
• 404 Not Found - Agent not found

Error Responses:

{
  "error": {
    "message": "request_id is required"
  }
}

Note: If a request has already been cancelled, the endpoint will return a 200 status with the message "Request already cancelled".

CMS (Content Management System)

The CMS allows you to manage structured content with custom record types, relationships, and localization. Access the CMS through the CMS section.

Record Types

Record types define the structure of your content. They can be created and managed through the CMS interface.

Records

The BrightsyClient provides a fluent API for querying CMS records with three endpoints:

API Endpoints:

• client.cda - Content Delivery API (read-only published records)
• client.cma - Content Management API (full CRUD on draft records)
• client.cpa - Content Preview API (read-only latest draft version)

Query Methods:

• .from(slug) - Specify record type by slug (alias: .recordType(slug))
• .where(field, operator, value) - Add filter condition
• .and(field, operator, value) - Add AND filter condition
• .or(field, operator, value) - Add OR filter condition
• .openGrouping() - Open parenthesis for filter grouping
• .andOpenGrouping() - AND with opening parenthesis
• .orOpenGrouping() - OR with opening parenthesis
• .closeGrouping() - Close parenthesis
• .select(...fields) - Select specific fields to return
• .groupBy(...fields) - Alias for select, used with aggregates for grouping
• .orderBy(field, direction) - Sort results ('asc' or 'desc')
• .page(number) - Set page number
• .pageSize(size) - Set page size
• .count(field?) - Add count aggregate
• .sum(field) - Add sum aggregate
• .avg(field) - Add average aggregate
• .min(field) - Add min aggregate
• .max(field) - Add max aggregate
• .tags(...tagNames) - Filter by tags
• .id(id) - Specify record ID for single record operations

Filter Operators:

• eq - Equals
• neq - Not equals
• gt - Greater than
• gte - Greater than or equal
• lt - Less than
• lte - Less than or equal
• like - Contains (case-sensitive)
• ilike - Contains (case-insensitive)
• in - In array
• notIn - Not in array
• begins - Begins with
• ends - Ends with

// Query published records
const records = await client.cda.from('articles')
  .where('status', 'eq', 'published')
  .and('category', 'eq', 'technology')
  .orderBy('published_at', 'desc')
  .get();

// Access results
records.data.forEach(record => {
  console.log(record.data);
});

// Pagination
console.log(records.pagination.total); // Total count
console.log(records.pagination.page); // Current page
console.log(records.pagination.page_count); // Total pages
console.log(records.pagination.page_size); // Page size

// Complex filtering with grouping
// Query: status = 'published' AND (featured = true OR views > 1000)
const records = await client.cda.from('articles')
  .where('status', 'eq', 'published')
  .andOpenGrouping()
    .where('featured', 'eq', true)
    .or('views', 'gt', 1000)
  .closeGrouping()
  .get();

// Field selection
const records = await client.cda.from('articles')
  .select('title', 'summary', 'author.name')
  .where('status', 'eq', 'published')
  .get();

// Filter by tags
const records = await client.cda.from('articles')
  .tags('featured', 'technology')
  .where('status', 'eq', 'published')
  .get();

// Create record (CMA only)
await client.cma.from('articles').save({
  data: {
    title: 'New Article',
    content: 'Article content'
  },
  tags: ['featured', 'technology']
});

// Update record (CMA only)
await client.cma.from('articles').id('record-id').save({
  data: { title: 'Updated Title' }
});

// Delete record (CMA only)
await client.cma.from('articles').id('record-id').delete();

// Aggregations
const stats = await client.cda.from('articles')
  .where('status', 'eq', 'published')
  .count()
  .sum('views')
  .avg('rating')
  .get();

// Aggregations with grouping by category
const categoryStats = await client.cda.from('articles')
  .groupBy('category', 'author')
  .where('status', 'eq', 'published')
  .count()
  .sum('views')
  .get();

// Results will be grouped by category and author
categoryStats.data.forEach(group => {
  console.log(`Category: ${group.category}, Author: ${group.author}`);
  console.log(`Count: ${group.count}`);
  console.log(`Total views: ${group.views_sum}`);
});

File Manager

The File Manager provides storage and organization for files. Access it through the File Manager section in the main navigation.

Upload Files

const file = await client.files.upload({
  file: fileBuffer,
  filename: 'document.pdf',
  path: 'documents/'
});

List Files

const files = await client.files.list({
  path: 'documents/',
  limit: 50
});

Delete Files

await client.files.delete('documents/document.pdf');

Automations

The Automations section provides tools for workflow automation, including scenarios, API keys, webhooks, and schedules.

Scenarios

Scenarios are automated workflows that can be triggered by webhooks or scheduled executions.

Execute Scenario

POST /scenario/{scenarioId}

Executes a scenario.

Path Parameters:

• scenarioId - The unique identifier of the scenario (string, required)

Request Headers:

• [Signature Header Name] - The header name is defined in the scenario configuration as signature_name. For example, if signature_name is set to "X-Scenario-Signature", you would include that header in your request. While you could technically use the raw signature_secret as the header value, it's best practice to generate a signature using the signature_secret and the payload as described below.

Signature Generation:

The signature must be generated using HMAC SHA-256 with the following process:

1. Convert the entire request body to a JSON string (payload)
2. Create an HMAC SHA-256 hash using the scenario's signature_secret as the key
3. Update the hash with the payload string directly
4. Get the hex digest of the hash, which becomes the signature value

Example (Node.js):

const crypto = require('crypto');
const payload = JSON.stringify(requestBody);
const signature = crypto.createHmac('sha256', signature_secret)
  .update(payload, 'utf8')
  .digest('hex');

Security Best Practice: Using HMAC signatures as shown above is the recommended best practice for API security. The signature validates the authenticity of requests and ensures data integrity, protecting against tampering. Store your signature_secret securely and never expose it in client-side code.

Request Body:

• Content Type: application/json
• Schema: An object containing an input property of type string

Responses:

• 200 OK - Scenario executed successfully
• 401 Unauthorized - When the signature header is missing or invalid
• 204 No Content - For CORS preflight requests

API Keys

API keys are managed in Automations → API Keys. They can be scoped to specific agents, scenarios, or CMS access.

Webhooks

Webhooks are managed in Automations → Webhooks. They allow external systems to trigger scenarios or receive notifications.

Schedules

Scheduled executions can be configured in Automations → Schedules to run scenarios at specific times or intervals.

Object Models

Message Object

Represents a message with various content types.

Properties:

• id - Unique identifier for the message (string)
• content - Array of content items, each being one of:
  • Text: Object with type as "text" and text as a string
  • Image URL: Object with type as "image_url" and image_url containing url and optional detail

Example:

{
  "id": "msg_123456789",
  "content": [
    {
      "type": "text",
      "text": "Hello, this is a text message"
    },
    {
      "type": "image_url",
      "image_url": {
        "url": "https://example.com/image.jpg",
        "detail": "high"
      }
    }
  ]
}

Chat Object

Represents a chat containing messages.

Properties:

• id - Unique identifier for the chat (string)
• messages - Array of Message objects

Example:

{
  "id": "chat_987654321",
  "messages": [
    {
      "id": "msg_123456789",
      "content": [
        {
          "type": "text",
          "text": "Hello, this is a text message"
        }
      ]
    }
  ]
}

Attachments

Delete an Attachment

DELETE /agents/{agentId}/chat/completion/workspace/{chatId}/messages/{messageId}/attachments/{filename}

Deletes a specific attachment from a message.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)
• messageId - The unique identifier of the message (string, required)
• filename - The name of the file to delete (string, required)

Responses:

• 200 OK - File deleted successfully
• 204 No Content - For CORS preflight requests

Upload an Attachment

File uploads use a two-step process for security and efficiency:

Step 1: Get Upload URL

POST /agent/{agentId}/upload-url

Gets a signed upload URL for secure file upload.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema:

{
  "filename": "example.pdf",
  "contentType": "application/pdf",
  "messageId": "msg_123456789",
  "chatId": "chat_987654321"
}

Properties:

• filename - The name of the file to upload (string, required)
• contentType - The MIME type of the file (string, optional)
• messageId - The unique identifier of the message (string, required)
• chatId - The unique identifier of the chat (string, required)

Responses:

• 200 OK - Returns upload URL and file information

{
  "uploadUrl": "https://supabase-storage-url.com/signed-upload-url",
  "fileUrl": "https://attachments.brightsy.ai/account/agents/agent-id/chats/chat-id/messages/message-id/filename.pdf",
  "path": "account/agents/agent-id/chats/chat-id/messages/message-id/filename.pdf"
}

Step 2: Upload File to Signed URL

PUT {uploadUrl}

Upload the file directly to the signed URL returned from Step 1.

Request Body:

• Content Type: The content type specified in Step 1
• Body: Raw file binary data

Responses:

• 200 OK - File uploaded successfully

Complete Upload Example:

// Step 1: Get upload URL
const uploadResponse = await fetch('/api/v1beta/agent/agent-123/upload-url', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    filename: 'document.pdf',
    contentType: 'application/pdf',
    messageId: 'msg-456',
    chatId: 'chat-789'
  })
});

const { uploadUrl, fileUrl } = await uploadResponse.json();

// Step 2: Upload file to signed URL
const fileUploadResponse = await fetch(uploadUrl, {
  method: 'PUT',
  headers: {
    'Content-Type': 'application/pdf'
  },
  body: fileBuffer
});

// File is now accessible at fileUrl

Messages

Create or Update Message

POST /agent/{agentId}/chat/completion/workspace/{chatId}/message/{messageId}

Creates or updates a message within a chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)
• messageId - The unique identifier of the message (string, required)

Request Body:

• Content Type: application/json
• Schema: A Message object

Responses:

• 200 OK - Message created/updated successfully

Delete Message

DELETE /agent/{agentId}/chat/completion/workspace/{chatId}/message/{messageId}

Deletes a specific message.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)
• messageId - The unique identifier of the message (string, required)

Responses:

• 200 OK - Message deleted successfully
• 204 No Content - For CORS preflight requests

List Messages

GET /agent/{agentId}/chat/completion/workspace/{chatId}/message

Retrieves a list of messages within a chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Responses:

• 200 OK - Returns a list of messages

Add Message to Chat

POST /agent/{agentId}/chat/completion/workspace/{chatId}/message

Creates a new message within a chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Request Body:

• Content Type: application/json
• Schema: A Message object

Responses:

• 200 OK - Message created successfully
• 204 No Content - For CORS preflight requests

Chats

Get Chat Details

GET /agent/{agentId}/chat/completion/workspace/{chatId}/message

Retrieves details of a specific chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Responses:

• 200 OK - Returns chat details

Create Chat

POST /agent/{agentId}/chat/completion/workspace/{chatId}/message

Creates a new message and chat if needed.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)
• chatId - The unique identifier of the chat (string, required)

Request Body:

• Content Type: application/json
• Schema: An object containing a message property of type string

Responses:

• 200 OK - Chat created successfully
• 204 No Content - For CORS preflight requests

Agent Messages

Create Workspace Chat

POST /agent/{agentId}/chat/completion/workspace

Creates a message for an agent and starts a new workspace-based chat.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema: A Message object

Responses:

• 200 OK - Message created successfully
• 204 No Content - For CORS preflight requests

Create Completion

POST /agents/{agentId}/chat/completion

Creates a completion where the client provides the entire messages array.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema:

{
  "messages": [
    {
      "role": "user",
      "content": "Hello, how can you help me?"
    }
  ],
  "context": {
    "user": { "name": "John" },
    "customData": { /* any data accessible to the agent */ }
  },
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get the current weather for a location",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {
              "type": "string",
              "description": "The city name"
            }
          },
          "required": ["location"]
        }
      }
    }
  ],
  "max_tokens": 16000
}

Properties:

• messages - Array of message objects with role and content (array, required)
• context - Additional context data available to the agent (object, optional)
• tools - Array of function tools the agent can call (array, optional)
• max_tokens - Maximum tokens for the response, overrides agent default (number, optional)

Responses:

• 200 OK - Completion generated successfully
• 204 No Content - For CORS preflight requests

Note: This endpoint does not store the conversation in Brightsy. No chatId is used as the conversation is managed by the client.

Cancel Chat Completion

POST /agent/{agentId}/chat/completion/cancel

Cancels an ongoing chat completion request.

Path Parameters:

• agentId - The unique identifier of the agent (string, required)

Request Body:

• Content Type: application/json
• Schema:

{
  "request_id": "req_123456789",
  "chat_id": "chat_987654321",
  "reason": "user_cancellation"
}

Properties:

• request_id - The unique identifier of the request to cancel (string, required)
• chat_id - The unique identifier of the chat (string, optional - only needed for workspace-based chats)
• reason - The reason for cancellation (string, optional, defaults to "user_cancellation")

Responses:

• 200 OK - Request cancelled successfully

{
  "message": "Request cancelled successfully",
  "cancelled_at": "2024-01-15T10:30:00.000Z",
  "request_id": "req_123456789"
}

• 400 Bad Request - Invalid request parameters
• 401 Unauthorized - User does not have access to agent
• 404 Not Found - Agent not found

Error Responses:

{
  "error": {
    "message": "request_id is required"
  }
}

Note: If a request has already been cancelled, the endpoint will return a 200 status with the message "Request already cancelled".

Scenarios

Execute Scenario

POST /scenario/{scenarioId}

Executes a scenario.

Path Parameters:

• scenarioId - The unique identifier of the scenario (string, required)

Request Headers:

• [Signature Header Name] - The header name is defined in the scenario configuration as signature_name. For example, if signature_name is set to "X-Scenario-Signature", you would include that header in your request. While you could technically use the raw signature_secret as the header value, it's best practice to generate a signature using the signature_secret and the payload as described below.

Signature Generation:

The signature must be generated using HMAC SHA-256 with the following process:

1. Convert the entire request body to a JSON string (payload)
2. Create an HMAC SHA-256 hash using the scenario's signature_secret as the key
3. Update the hash with the payload string directly
4. Get the hex digest of the hash, which becomes the signature value

Example (Node.js):

const crypto = require('crypto');
const payload = JSON.stringify(requestBody);
const signature = crypto.createHmac('sha256', signature_secret)
  .update(payload, 'utf8')
  .digest('hex');

Security Best Practice: Using HMAC signatures as shown above is the recommended best practice for API security. The signature validates the authenticity of requests and ensures data integrity, protecting against tampering. Store your signature_secret securely and never expose it in client-side code.

Request Body:

• Content Type: application/json
• Schema: An object containing an input property of type string

Responses:

• 200 OK - Scenario executed successfully
• 401 Unauthorized - When the signature header is missing or invalid
• 204 No Content - For CORS preflight requests

Object Models

Message Object

Represents a message with various content types.

Properties:

• id - Unique identifier for the message (string)
• content - Array of content items, each being one of:
  • Text: Object with type as "text" and text as a string
  • Image URL: Object with type as "image_url" and image_url containing url and optional detail

Example:

{
  "id": "msg_123456789",
  "content": [
    {
      "type": "text",
      "text": "Hello, this is a text message"
    },
    {
      "type": "image_url",
      "image_url": {
        "url": "https://example.com/image.jpg",
        "detail": "high"
      }
    }
  ]
}

Chat Object

Represents a chat containing messages.

Properties:

• id - Unique identifier for the chat (string)
• messages - Array of Message objects

Example:

{
  "id": "chat_987654321",
  "messages": [
    {
      "id": "msg_123456789",
      "content": [
        {
          "type": "text",
          "text": "Hello, this is a text message"
        }
      ]
    }
  ]
}