Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.brainstormer.io/llms.txt

Use this file to discover all available pages before exploring further.

This page explains the building blocks of Brainstormer. Understanding these concepts will help you navigate the dashboard and use the API effectively.

​
Agent

An agent is an AI assistant you configure and deploy. Each agent has:
  • A model — the underlying LLM that generates responses
  • A system prompt — instructions that define the agent’s role, tone, and behavior
  • An optional knowledge base — documents and URLs the agent uses to answer questions
  • An optional voice — a custom TTS voice for spoken responses
Agents are scoped to your organization. You can create as many as you need, and each one maintains its own conversations and configuration independently. You interact with agents through the dashboard chat interface or the REST API.

​
Model

The model is the large language model (LLM) that powers an agent’s responses. Brainstormer connects to 300+ models via OpenRouter, including:
  • OpenAI: GPT-4, GPT-4o
  • Anthropic: Claude 3 Haiku, Claude 3 Sonnet, Claude 3 Opus
  • Google: Gemini 2.0 Flash, Gemini Pro
  • Meta: Llama models
  • And many more
You choose a model when you create an agent and can change it at any time. Different models have different capabilities, speeds, and costs. The model you pick affects how credits are consumed per conversation.

​
Knowledge base

A knowledge base is a collection of content — documents, PDFs, and URLs — that an agent uses to answer questions accurately. When a user sends a message, Brainstormer searches the knowledge base for relevant content and includes it in the model’s context before generating a response. This technique is called retrieval-augmented generation (RAG). Knowledge bases support:
  • PDF, Word, and plain text documents
  • Public URLs and RSS feeds
  • Multimodal content including images and audio
Responses generated using knowledge base content include citation markers (e.g. [1], [2]) with interactive tooltips that show the source document and the relevant passage. An organization can have multiple knowledge bases, and a single agent can be linked to one.

​
Conversation

A conversation is a chat session between a user and an agent. Brainstormer preserves the full message history within a conversation, so the agent has context from earlier turns when it generates each new response. When you call POST /v1/chat without a conversation_id, a new conversation is created and its ID is returned. Pass that ID in subsequent calls to continue the same thread. Conversations are stored and visible in the dashboard under Chat History.

​
Credit

Credits are the billing unit in Brainstormer. Your organization has a credit wallet, and interactions debit credits based on what the agent does:
  • Model token usage (prompt and completion tokens)
  • Voice synthesis (text-to-speech)
  • Knowledge base queries (vector search)
You can view your current balance, top up credits, and see a detailed usage breakdown on the Billing page. The credit ledger is append-only, so every debit and credit is fully auditable.
Credits are consumed at the organization level, not per user. All agents and team members draw from the same wallet.

​
Organization

An organization is the workspace that groups everything in Brainstormer: your agents, knowledge bases, team members, and credit wallet. When you sign up, you create an organization automatically. Within an organization you can:
  • Invite team members and assign roles
  • Share agents and knowledge bases across the team
  • Monitor aggregate usage and credits
  • Manage API applications and keys
If you need separate environments (e.g. production vs. staging), consider using separate organizations.

​
API key

An API key is a credential that allows external applications to call the Brainstormer REST API on behalf of your organization. Keys look like: 
brs_live_xxxxxxxxxxxxxxxxxx
You generate keys from the Developer page, organized under Applications. Each application can have multiple keys — for example, one for production and one for development. Pass your key in the Authorization header of every API request: 
Authorization: Bearer brs_live_xxxxxxxxxxxxxxxxxx
A revoked or deleted key stops working immediately. Rotate keys regularly and never expose them in client-side code.
API keys have access to all agents and knowledge bases in your organization. Treat them like passwords.

​
Voice

A voice is a custom text-to-speech (TTS) model cloned from audio samples you provide. Once created, a voice can be attached to an agent so that its responses are spoken aloud. To create a voice:
  1. Upload 5–25 audio samples of the speaker (MP3 or WAV).
  2. Brainstormer trains a voice model from those samples.
  3. Attach the voice to an agent from the agent’s settings.
Voice synthesis consumes credits and is billed per session. The dashboard includes a voice testing interface where you can preview synthesis quality before deploying.
Voice cloning requires you to have the rights to clone the voice you’re uploading. Only upload recordings of yourself or recordings you have explicit permission to use.