Knowledge Bases

​

Overview

​

Creating a Knowledge Base

1. Navigate to Control Hub → Knowledge Bases
2. Click Create Knowledge Base
3. Configure the following settings:

​

Basic Settings

• Name: A descriptive name for your knowledge base
• Description: Optional description of the content and purpose
• Embedding Model: Choose the AI model for generating search embeddings

​

Structured Data Extraction (Optional)

• Extraction Model: AI model to use for structured data extraction
• Extraction Schema: JSON Schema defining what data to extract from each document

​

Supported File Types

​

File Processing Pipeline

1. Upload: Secure cloud storage
2. Content Extraction: Text extracted from various formats
3. Structured Extraction: AI extracts data based on your schema (if configured)
4. Intelligent Chunking: Documents split while preserving structure
5. Vector Embeddings: AI generates searchable embeddings
6. Storage: Everything indexed for fast retrieval

​

Structured Data Extraction

​

What is Structured Extraction?

• Financial Reports: Extract revenue, costs, and key metrics
• Contracts: Pull out dates, parties, and terms
• Research Papers: Extract methodology, results, and conclusions
• Product Specs: Standardize technical specifications

​

Setting Up Extraction

1. Choose an Extraction Model: Select from your available chat models
2. Define JSON Schema: Specify the structure you want extracted

​

Example JSON Schema

{
  "type": "object",
  "properties": {
    "title": {
      "type": "string",
      "description": "Document title or heading"
    },
    "key_metrics": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "metric": {"type": "string"},
          "value": {"type": "number"},
          "unit": {"type": "string"}
        }
      },
      "description": "Important numerical data points"
    },
    "summary": {
      "type": "string",
      "description": "Brief summary of main points"
    }
  },
  "required": ["title", "summary"]
}

​

Viewing Extracted Data

• View extracted JSON data in the file details
• Use the data in custom workflows
• Export for analysis in other tools

​

Agent Integration

​

Enabling Knowledge Base Access

1. Edit your agent configuration
2. Enable the Search Knowledge Base tool
3. Select which knowledge bases the agent can access

​

Advanced Search Features

​

File Filtering

{
  "query": "quarterly revenue",
  "knowledge_base_id": 123,
  "include_file_ids": [45, 67],  // Search only these files
  "exclude_file_ids": [12, 34],  // Exclude these files
  "top_k": 10
}

​

Search Types

• Vector Search: Semantic similarity using AI embeddings
• Text Search: Traditional keyword matching
• Hybrid Search: Combines both approaches for best results

​

Best Practices

​

Organization

• Separate by Topic: Create different knowledge bases for different subjects
• Regular Updates: Remove outdated documents to maintain search quality
• Descriptive Names: Use clear, searchable file names

​

Structured Extraction

• Start Simple: Begin with basic schemas and expand over time
• Test with Sample Files: Validate your schema works before bulk uploads
• Use Required Fields: Mark essential fields as required in your schema

​

Performance

• Optimize File Sizes: Larger files take longer to process
• Monitor Processing: Check the processing status of uploaded files
• Use File Filtering: Help agents focus on relevant documents

​

Managing Knowledge Bases

​

File Management

• Upload Status: Monitor processing progress for each file
• Reprocess Files: Re-run extraction if you update your schema
• Delete Files: Remove individual files without affecting others

​

Knowledge Base Settings

• Update Schema: Modify extraction schema for future uploads
• Change Models: Switch embedding or extraction models as needed
• Access Control: Configure which agents can access the knowledge base

​

Troubleshooting

• Processing Failures: Check file format compatibility
• Poor Search Results: Consider adjusting your embedding model
• Extraction Issues: Validate your JSON schema syntax

​

API Access

# Search a knowledge base
curl -X POST "https://asteragents.com/api/kb/search" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "revenue trends",
    "knowledge_base_id": 123,
    "top_k": 5
  }'

​

Pricing Notes

• File processing and storage are included in your plan
• Embedding generation uses your model provider credits
• Structured extraction uses your chat model credits
• No additional fees for search operations