Protobox

Knowledge Base API

RAG-powered knowledge management for AI agents

The Knowledge Base API enables you to build RAG (Retrieval-Augmented Generation) powered AI agents. Upload documents, search with hybrid vector+text queries, and let your agents access business knowledge programmatically.

Base URL

https://platform.protobox.ai/api/v1/knowledge

Authentication

All endpoints require authentication via Bearer token or API key:

# Bearer token (from login)
Authorization: Bearer <access_token>

# API key (for MCP/integrations)
X-API-Key: <api_key>
X-Workspace-ID: <workspace_id>

Endpoints

MethodEndpointDescription
POST/knowledgeCreate manual entry
GET/knowledgeList entries
GET/knowledge/:idGet entry
PATCH/knowledge/:idUpdate entry
DELETE/knowledge/:idDelete entry
POST/knowledge/searchSearch knowledge
POST/knowledge/uploadUpload URL/text
POST/knowledge/upload/fileUpload file
GET/knowledge/tasksList tasks
GET/knowledge/tasks/:taskIdGet task status
GET/knowledge/configGet KB config
PATCH/knowledge/configUpdate KB config
GET/knowledge/statsGet statistics
POST/knowledge/foldersCreate folder
GET/knowledge/foldersList folders
GET/knowledge/folders/:idGet folder
PATCH/knowledge/folders/:idUpdate folder
DELETE/knowledge/folders/:idDelete folder
POST/knowledge/:id/moveMove to folder
GET/knowledge/:id/sync-configGet sync config
PATCH/knowledge/:id/sync-configUpdate sync config
POST/knowledge/:id/syncTrigger sync

Search Modes

Combines vector semantic search with text matching using Reciprocal Rank Fusion (RRF).

Best for: Most queries - balances meaning and exact matches

{
  "query": "What's your return policy for electronics?",
  "mode": "hybrid",
  "limit": 5
}

Pure semantic search using OpenAI embeddings (1536 dimensions).

Best for: Conceptual questions where exact words don't matter

{
  "query": "Can I get my money back?",
  "mode": "vector",
  "limit": 5
}

Full-text search with MongoDB text indexes (BM25-style).

Best for: Finding specific terms, product names, codes

{
  "query": "SKU-12345 warranty",
  "mode": "text",
  "limit": 5
}

Source Types

SourceDescriptionUse Case
manualText entries created directlyQuick FAQ entries, announcements
documentUploaded files (PDF, DOCX, etc.)Product manuals, policies
webpageCrawled web pagesHelp center articles, blog posts
faqStructured Q&A pairsCommon customer questions
api_docsAPI documentationTechnical support

Supported File Types

CategoryFormats
DocumentsPDF, DOCX, TXT
WebHTML, URLs (auto-crawled)
MarkdownMD, MDX

Processing Pipeline

1. Upload 2. Process 3. Store 4. Search File/URL/Text Parse Content Chunk Text Generate Embeddings S3 Storage MongoDB + Atlas Vector Search Text Search RRF Fusion

Rate Limits

OperationLimit
Create/Update100/min
Search200/min
Upload20/min
File size10 MB

MCP System Tools

When using the Knowledge Base via MCP, these system tools are available:

ToolDescription
kb_searchSearch with hybrid/vector/text modes
kb_addCreate a manual entry
kb_listList entries with pagination
kb_getGet a single entry by ID
kb_statsGet KB statistics

What's Next?

Create Entry

Add knowledge manually

Search

Query your knowledge base

Upload Files

Ingest documents and URLs

SDK Reference

Use the TypeScript SDK

On this page