Frequently Asked Questions

Use Bunyan to chat with available models, ask questions about documents, summarize and compare content, and use tools from the Playground. Save a recurring setup as a Blueprint. Start with the Quick Start.

For your own applications, Bunyan provides an LLM API and Document Intelligence API.

Open Settings → Organizations and select Switch To beside your organization. Then choose the intended project from the project selector at the top of the page.

If the organization is missing, ask an owner to check your membership and account email. If a project is missing, ask its administrator or an organization owner to review your access. See Organization Settings.

Check your selected organization and project first. Access also depends on organization membership, project permissions, and the resource's sharing settings. Selecting a project does not grant access to everything in it.

Organization roles are Owner, Editor, and Viewer. Project access levels are Admin, Editor, and Viewer. Ask the resource owner or your administrator to review your access. See Project Settings.

Open the model menu below the Playground message box and choose an available configuration. Supported capabilities depend on that configuration. If no suitable model appears, ask your organization administrator to review your access and available configurations.

For an API request, use the configuration name available to you in Bunyan. See LLM Endpoints.

Attach a file with the Playground's paperclip button when you want to use it in that conversation. Wait for Ready, then select the attachment with @.

Use Document Collections to organize documents for repeated use. Wait for each document to show Complete, then select the collection or document with @. See Ask Questions About Documents and Collections.

Collection uploads accept text-based PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), text, Markdown, and HTML files. Upload one local file at a time, up to 100 MiB. Scanned or image-only documents are not supported by the collection upload interface.

Playground attachments have separate limits: up to five per conversation, with a maximum of 20 MiB per file. Check attachment formats and limits before uploading. Document Intelligence has its own API file and processing limits.

Yes. Open a collection and use Upload → upload from web or Upload → upload from external source. Web content must be accessible to Bunyan. For S3-compatible storage, you need the endpoint, bucket, and credentials for the files you want to import.

Imports continue in the background. Refresh the collection's document list to check which documents completed. See Upload Documents.

Refresh the collection's document list and check its status. An upload confirmation means the file was received; wait for Complete before using it as a source. For a Playground attachment, wait for Ready.

If processing fails, check that the file opens, contains readable text, and meets the upload requirements. If processing remains unchanged, share the collection name, document title, and visible status with your administrator. If a completed document is missing from suggestions, check the selected project and your collection access. See Document Processing.

Type @, find the document or collection, and select it from the suggestions. Typing its name as plain text does not select it. Ask a specific question and state whether the answer should use only the selected sources.

Review supporting passages and source references against the original document. If the result is incomplete, narrow the question or identify the section you need. See Ask Questions About Documents.

Open the folder-shaped tools button below the message box, or type @, then select an available tool or source. The paperclip button is for uploading attachments.

Tool availability depends on your workspace and permissions. For a missing connected tool, check that its MCP connection is enabled, authorization is complete, and both the connection and tool are shared with you. See Playground Tools and MCP Connections.

Create a Blueprint to save instructions, a preferred model and language, and selected tools or sources. Choose an active Blueprint in the Playground before sending your request.

A Blueprint does not grant access to its selected resources. Everyone using it needs access to its model, sources, and tools.

Use the LLM API to call your available model configurations through an OpenAI-compatible client. Its base URL is https://inference-v2.bunyan.ai/v1/; send your Bunyan key as Authorization: Bearer YOUR_API_KEY.

Use Document Intelligence to parse documents, summarize content, extract structured data, or run agentic tasks. Its base URL is https://app.bunyan.ai/rag-agents/api/v1/document-intelligence; send your Bunyan key as X-API-Key: YOUR_API_KEY.

Manage your projects, configurations, and keys in Bunyan. See API Reference for examples.

Open API Keys in Bunyan, create a key for the intended project, and copy its value immediately. Store it in your server environment or a secret manager. Later listings show a masked value.

Use Rotate to create a replacement and revoke the old key, then update your application. Use Delete to revoke a key you no longer need. See Create an API Key.

Check the API's base URL and authentication header, then confirm that your key is active and unexpired. For LLM calls, verify the configuration name and your access to it.

Document Intelligence also requires a plan that includes the API, the scopes required by the operation, and project access for the key owner. Check the response's error message before retrying. See LLM troubleshooting and Document Intelligence access requirements.

Open Usage and Cost and start with Summary. Select a date range and, in an organization workspace, a project. Organization owners can also filter by user.

Choose a category for more detail, such as APIs for key activity or Models for model consumption. Read the units, plan limits, and reset dates shown on the dashboard. See Usage & Cost.

Narrow the date range and compare Daily Spend with Totals Over Time. Check Spend Per Project, Top Keys under APIs, and Top Models under Models to find the contributors. Review failed requests if activity rises unexpectedly.

Keep the same filters when comparing panels. If a panel reports incomplete spend data, treat that breakdown as partial. See Usage Breakdown.