Skip to main content
Pawa Knowledge Base is a managed content store that automatically indexes your documents for semantic search and retrieval-augmented generation (RAG). Upload files (PDF, DOCX, images, HTML, PPTX, XLSX, text) and query them with natural language in English or Swahili. You can create and update your knowledge base at any time. Explore the full list of endpoint workflows in the API reference. Unlock the value of your documents by transforming them into searchable, AI-ready knowledge, enabling you to build grounded assistants that cite your own content. You can knowledge base in :
  • Your agents chat request.
  • Your RAG pipelines.
Knowledge Base Example Using Pawa AI
Document Parsing Example

​
Use Cases of Knowledge Base

📚 Company Knowledge

Centralize policies, HR docs, and manuals for employees to query in natural language.

⚖️ Legal Document Search

Retrieve clauses, dates, and rules from contracts and compliance files instantly.

🏥 Healthcare Records

Index medical reports and prescriptions for secure semantic search in digital health systems.

📊 Research & Reports

Store and retrieve insights from research papers, presentations, and business reports.

​
Supported Formats

Pawa AI can help build a knowledge base with a wide range of formats, including:
  • PDFs
  • DOCX
  • Excel (XLSX)
  • PowerPoint (PPTX)
  • Images (PNG)
  • Audio files
  • Web links / HTML

​
Pawa AI Knowledge Base Workflow

  1. Create a knowledge base container, with reference files
  2. Query with natural language using semantic retrieval to retrieve top-K passages
  3. Feed semantic results into chat for grounded RAG answers.
  4. Out of the box RAG request, just pass the knowledge_base_id as parameter.

​
Create a Knowledge Base, With multiple files.

curl --request POST \
  --url https://api.pawa-ai.com/v1/store/knowledge-base \
  --header "Authorization: Bearer $PAWA_AI_API_KEY" \
  --header 'Content-Type: multipart/form-data' \
  --form 'name=Pawa Knowledge Base' \
  --form 'description=This is a sample knowledge base for Pawa.' \
  --form knowledgeBase=@example-file

​
The Example Response

{
  "success": true,
  "message": "Knowledge base created successfully",
  "data": {
    "id": 21,
    "name": "WCF Knowledge Base",
    "kbReferenceId": "kb-5d76835d-fe3e-4793-b0d6-f2e1a668eb5d",
    "description": "This is a knowledge base for WCF, it contains information about various concerns about Workers Compesation Fund., it is used to answer questions about WCF",
    "status": "ACTIVE",
    "createdAt": "2025-08-04T16:35:37.522Z",
    "updatedAt": "2025-08-04T16:35:37.522Z"
  }
}
After creating the knowledge base, now you can anything with it, such as performing semantic retrieval, or sending your out-of-the-box RAG chat request.

​
Semantic retrieval on Knowledge Base

  curl --request POST \
  --url https://api.pawa-ai.com/v1/store/knowledge-base/semantic-retrieval \
  --header "Authorization: Bearer $PAWA_AI_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "knowledge_base_id": "21",
  "question_asked_by_user": "When WCF got started?"
}'

​
Grounded answers (RAG) on Knowledge Base

curl --request POST \
     --url https://api.pawa-ai.com/v1/chat/request \
     --header "Authorization: Bearer $PAWA_AI_API_KEY" \  
     --header 'Content-Type: application/json' \
     --data '{
      "model": "pawa-v1-blaze-20250318",
                 "messages": [
                  {
                     "role": "user",
                     "content": [
                           {
                              "type": "text",
                              "text": "What was the last years sales of the company?"
                           }
                     ]
                  }]
               "knowledgeBase": {
                  "kbReferenceId": "kb-930d251e-8a8b-4ba9-bae1-5fceb47bd654",
                  "isMust": "True"
               }}'

​
Troubleshooting

  • Low relevance: rephrase query; ensure documents are properly parsed
  • Missing files: confirm knowledge_base_id and upload success
  • Slow retrieval: reduce top_k or filter by metadata
⌘I