Skip to content

knowledge.mdx

Latest commit

 

History

History
312 lines (242 loc) · 7.22 KB

knowledge.mdx

File metadata and controls

312 lines (242 loc) · 7.22 KB
title Knowledge Base
description Add documents to your agents' knowledge base for RAG-powered responses on elizaOS Cloud.

import { Callout, Steps, Tabs, Cards } from "nextra/components";

Knowledge Base

Enhance your agents with document-based knowledge using Retrieval-Augmented Generation (RAG).

Stable

Overview

The Knowledge Base allows you to:

  • Upload documents (PDF, TXT, MD, etc.)
  • Automatically chunk and embed content
  • Query semantically for relevant context
  • Connect to agents for enhanced responses

Quick Start

Dashboard

Navigate to Dashboard → Knowledge for the visual interface.

API

<Tabs items={['cURL', 'JavaScript', 'Python']}> <Tabs.Tab>

# Upload a document
curl -X POST "https://cloud.milady.ai/api/v1/knowledge/upload-file" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@document.pdf" \
  -F "agentId=agent_abc123"

</Tabs.Tab> <Tabs.Tab>

const formData = new FormData();
formData.append('file', file);
formData.append('agentId', 'agent_abc123');

const response = await fetch('https://cloud.milady.ai/api/v1/knowledge/upload-file', {
method: 'POST',
headers: {
'Authorization': 'Bearer YOUR_API_KEY',
},
body: formData,
});

const result = await response.json();
console.log('Document uploaded:', result.id);

</Tabs.Tab> <Tabs.Tab>

import requests

with open('document.pdf', 'rb') as f:
    response = requests.post(
        'https://cloud.milady.ai/api/v1/knowledge/upload-file',
        headers={'Authorization': 'Bearer YOUR_API_KEY'},
        files={'file': f},
        data={'agentId': 'agent_abc123'},
    )

result = response.json()
print(f'Document uploaded: {result["id"]}')

</Tabs.Tab>

Supported Formats

Format Extension Max Size
PDF .pdf 5MB
Text .txt 5MB
Markdown .md 5MB
Word .docx 5MB
JSON .json 5MB
XML .xml 5MB
YAML .yaml 5MB
CSV .csv 5MB

Note: Maximum 5MB per file and 5MB total per batch upload.

Upload Documents

### Prepare Your Documents Ensure documents are in a supported format and within size limits.

Upload via Dashboard or API

Use the Knowledge dashboard or the upload API endpoint.

Wait for Processing

Documents are automatically chunked and embedded.

Connect to Agent

Link the knowledge base to your agent for RAG.

Upload Response

{
  "id": "doc_abc123",
  "filename": "product-manual.pdf",
  "status": "processing",
  "chunks": null,
  "createdAt": "2024-01-15T10:30:00Z"
}

Check Processing Status

curl -X GET "https://cloud.milady.ai/api/v1/knowledge/check?documentId=doc_abc123" \
  -H "Authorization: Bearer YOUR_API_KEY"
{
  "id": "doc_abc123",
  "status": "completed",
  "chunks": 42,
  "tokensUsed": 15234
}

Query Knowledge

Search your knowledge base semantically:

curl -X POST "https://cloud.milady.ai/api/v1/knowledge/query" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "How do I reset my password?",
    "agentId": "agent_abc123",
    "limit": 5
  }'

Response

{
  "results": [
    {
      "content": "To reset your password, navigate to Settings > Security > Reset Password...",
      "score": 0.92,
      "documentId": "doc_abc123",
      "metadata": {
        "page": 15,
        "section": "Account Settings"
      }
    },
    {
      "content": "Password requirements include at least 8 characters...",
      "score": 0.85,
      "documentId": "doc_abc123",
      "metadata": {
        "page": 16,
        "section": "Security"
      }
    }
  ]
}

Managing Documents

List Documents

curl -X GET "https://cloud.milady.ai/api/v1/knowledge?agentId=agent_abc123" \
  -H "Authorization: Bearer YOUR_API_KEY"

Get Document Details

curl -X GET "https://cloud.milady.ai/api/v1/knowledge/doc_abc123" \
  -H "Authorization: Bearer YOUR_API_KEY"

Delete Document

curl -X DELETE "https://cloud.milady.ai/api/v1/knowledge/doc_abc123" \
  -H "Authorization: Bearer YOUR_API_KEY"

RAG Configuration

Chunk Settings

Configure how documents are split:

Setting Default Description
chunkSize 1000 Characters per chunk
chunkOverlap 200 Overlap between chunks
minChunkSize 100 Minimum chunk size

Retrieval Settings

Configure query behavior:

Setting Default Description
limit 5 Number of results
minScore 0.7 Minimum relevance score
rerank true Re-rank results for relevance

Agent Integration

Automatic RAG

When knowledge is connected to an agent, relevant context is automatically retrieved:

// Agent with knowledge base
const response = await fetch("https://cloud.milady.ai/api/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "agent_abc123", // Agent with connected knowledge
    messages: [{ role: "user", content: "How do I reset my password?" }],
  }),
});

The agent automatically queries the knowledge base and includes relevant context in its response.

Manual RAG

Query knowledge and inject context manually:

// 1. Query knowledge
const knowledge = await fetch("https://cloud.milady.ai/api/v1/knowledge/query", {
  method: "POST",
  headers: {
    Authorization: "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    query: "password reset",
    limit: 3,
  }),
}).then((r) => r.json());

// 2. Include in chat
const context = knowledge.results.map((r) => r.content).join("\n\n");

const response = await fetch("https://cloud.milady.ai/api/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "gpt-4o",
    messages: [
      { role: "system", content: `Use this context:\n\n${context}` },
      { role: "user", content: "How do I reset my password?" },
    ],
  }),
});

Pricing

See Billing & Credits for current pricing on document uploads, embeddings, queries, and storage.

Best Practices

  • Quality Content — Use well-structured, accurate documents
  • Organize — Group related documents for better retrieval
  • Update Regularly — Keep knowledge current by refreshing documents
  • Test Queries — Verify retrieval quality before production

Next Steps

Create agents that use knowledge Integrate RAG into your applications Learn about vector embeddings