Key Concepts
These are the main building blocks you will use in Bunyan.
Organizations and Projects
An organization brings together a team, its subscription, and shared work. A project groups related resources and has its own team access levels.
Choose your active organization and project before creating resources. Organization roles, project roles, and resource sharing determine what you can see and use.
Model Configurations
A model configuration is an available model connection you can select in Bunyan. Access and supported capabilities depend on the configuration.
- Language models generate responses in the Playground and support tasks such as summarization and analysis.
- Embedding models prepare document content for search. Select one when creating a collection.
Choose a language model in the Playground's model menu. A Blueprint can save a preferred model for a recurring task. See Playground.
Playground and Conversations
The Playground is where you chat with Bunyan, select sources, and use tools. A conversation keeps your messages and follow-up questions together.
Bunyan's agentic chat can use available tools to complete a request. Describe the task, specify the sources you want, and ask for the output format you need. See How Chat Works.
Collections and Attachments
A collection groups documents for repeated use. After upload, Bunyan prepares each document for search; wait for its status to show Complete.
An attachment is a file uploaded directly to a Playground conversation with the paperclip button. Wait for it to show Ready before sending your question.
Use @ and select the matching document, collection, or attachment from the suggestions to identify the source you want Bunyan to use.
See Data Management and Ask Questions About Documents.
Retrieval-Augmented Generation (RAG)
RAG means retrieving relevant content from your documents and using it to help answer a question. It lets you ask about information in a handbook, contract, or collection of reports.
A response may use selected passages rather than every part of a collection. Review any source references against the original document, and narrow your question if the answer is incomplete.
Blueprints
A Blueprint saves instructions, a preferred model and language, and selected tools or sources. Select it in the Playground to reuse that setup for a recurring task.
Blueprint instruction files guide the assistant's behavior. They are separate from searchable documents stored in collections.
See Blueprints.
Tools and MCP Connections
Tools let Bunyan search collections, read documents, search the web, run a workflow, or use a connected service.
An MCP connection connects Bunyan to a service that exposes tools through the Model Context Protocol. The service determines which actions it offers; connection and tool access settings determine who can use them.
See Playground Tools and MCP Connections.
Workflows
A workflow defines a sequence of steps for a repeatable task. You can select an accessible workflow through the Playground's Flow tool. Running it performs its configured actions.
API Keys and Public APIs
An API key lets your application authenticate to Bunyan's public APIs. Create it in Bunyan, then use the header required by the API:
| Public API | Purpose | Authentication |
|---|---|---|
| LLM API | Call available LLM configurations. | Authorization: Bearer YOUR_API_KEY |
| Document Intelligence | Parse, summarize, extract structured data, and run agentic tasks. | X-API-Key: YOUR_BUNYAN_API_KEY |
See Create an API Key and API Reference for setup and examples.