Knowledge Bases

Organize documents into searchable collections for context-aware AI chat and semantic search.

What are Knowledge Bases?

Knowledge bases are collections of documents that provide context for AI-powered chat. When you ask a question, DocLD searches the knowledge base for relevant information and provides answers with citations.

Key benefits:

• Focused context - Chat only searches relevant documents
• Better accuracy - More precise answers from curated content
• Team collaboration - Share knowledge bases with your team
• Analytics - Track what questions are being asked

Creating a Knowledge Base

From the Dashboard

1. Go to Knowledge Bases in the sidebar
2. Click Create Knowledge Base
3. Enter a name and description
4. Optionally select a template
5. Add documents

Using Templates

Templates provide pre-configured settings for common use cases:

Templates configure:

• Retrieval settings (how many results to fetch)
• Chunking strategy (how documents are split)
• Response style (concise vs detailed)

Adding Documents

Upload New Documents

Drag and drop files directly into a knowledge base, or select an existing knowledge base when uploading.

Add Existing Documents

Add documents you’ve already uploaded:

1. Open a knowledge base
2. Click Add Documents
3. Select from your document library
4. Click Add Selected

Batch Operations

Add multiple documents at once via the dashboard or API:

curl -X POST "/api/knowledge-bases/{id}/documents/batch" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{"document_ids": ["doc1", "doc2", "doc3"]}'

Chatting with Knowledge Bases

Once documents are added, start chatting:

1. Open the knowledge base
2. Click Chat or go to the Chat page
3. Select your knowledge base from the dropdown
4. Ask questions

How RAG Works

1. Query - Your question is sent to Pinecone; embeddings and search happen server-side (llama-text-embed-v2)
2. Search - Similar content is found in the knowledge base
3. Retrieve - Relevant chunks are gathered
4. Generate - AI creates an answer using the context
5. Cite - Sources are linked to the response

Chat Modes

Smart Suggestions

DocLD analyzes your knowledge base and suggests:

• Related documents - Documents that would improve coverage
• Gap analysis - Topics that aren’t well covered
• Duplicate detection - Potential duplicate documents

Access suggestions from the knowledge base page or via API.

Analytics

Track how your knowledge base is being used:

• Query volume - How many questions are asked
• Popular topics - Most common question themes
• Response quality - Confidence scores and feedback
• Document usage - Which documents are cited most

View analytics from the knowledge base page under the Analytics tab.

Settings

Configure knowledge base behavior:

Retrieval Settings

{
  "retrieval": {
    "top_k": 5,
    "threshold": 0.7,
    "reranking": true
  }
}

Chunking Settings

{
  "chunking": {
    "strategy": "semantic",
    "max_size": 1000,
    "overlap": 100
  }
}

Sharing & Collaboration

Organization Knowledge Bases

Create knowledge bases at the organization level to share with team members:

1. Create a knowledge base
2. Select an organization
3. All organization members can access it

Member Permissions

Best Practices

1. Keep focused - One topic or domain per knowledge base
2. Quality over quantity - Curated content beats volume
3. Update regularly - Keep documents current
4. Review analytics - See what’s working
5. Use templates - Start with pre-configured settings
6. Remove duplicates - Use duplicate detection

API Reference

See the Knowledge Bases API for programmatic access.