Knowledge CLI

Manage RAG/vector store knowledge bases with advanced retrieval strategies

The knowledge command manages document knowledge bases with support for multiple vector stores, retrieval strategies, rerankers, and query modes.

Quick Start

# List knowledge sources
praisonai knowledge list

# Add a document
praisonai knowledge add document.pdf

# Query knowledge base with RAG
praisonai knowledge query "API authentication"

# Query with advanced options
praisonai knowledge query "How to authenticate?" --retrieval fusion --reranker llm

Commands

Add Documents

# Add a single file
praisonai knowledge add document.pdf

# Add a directory
praisonai knowledge add ./docs/

# Add a URL
praisonai knowledge add https://example.com/docs.html

# Add with glob pattern
praisonai knowledge add "*.pdf"

Expected Output:

📚 Adding documents to knowledge base...

✅ Added: document.pdf
   • Pages: 45
   • Chunks: 128
   • Vectors: 128

Knowledge base updated successfully!
┌─────────────────────┬──────────────┐
│ Metric              │ Value        │
├─────────────────────┼──────────────┤
│ Total Documents     │ 12           │
│ Total Chunks        │ 1,456        │
│ Storage Size        │ 24.5 MB      │
└─────────────────────┴──────────────┘

Query Knowledge Base

# Basic query
praisonai knowledge query "How to authenticate API requests"

# Query with specific vector store
praisonai knowledge query "authentication" --vector-store chroma

# Query with fusion retrieval (multi-query)
praisonai knowledge query "How to authenticate?" --retrieval fusion

# Query with LLM reranking
praisonai knowledge query "API auth" --reranker llm

# Query with sub-question decomposition
praisonai knowledge query "What is Python and how to install it?" --query-mode sub_question

# Full advanced query
praisonai knowledge query "authentication flow" \
  --vector-store chroma \
  --retrieval fusion \
  --reranker llm \
  --index-type hybrid \
  --query-mode sub_question

Expected Output:

🔍 Querying knowledge base...

Found 5 relevant results:

1. [score: 0.95] api-docs.pdf (page 23)
   "Authentication is handled via Bearer tokens. Include the token
   in the Authorization header: Authorization: Bearer <token>"

2. [score: 0.87] security-guide.md (section 3.2)
   "All API requests must be authenticated. Unauthenticated requests
   will receive a 401 Unauthorized response."

3. [score: 0.82] quickstart.txt (line 45)
   "To get started, first obtain an API key from the dashboard..."

Search Knowledge Base

Unlike query, search returns raw ranked results with no LLM answer generation.

praisonai knowledge search "capital of France" --user-id myuser

See Knowledge CLI for the full flag reference, and Search Results for the SearchResult shape returned when calling Knowledge.search() from Python.

List Documents

praisonai knowledge list

Expected Output:

📚 Knowledge Base Contents:

┌────┬─────────────────────┬──────────┬────────┬─────────────────────┐
│ #  │ Document            │ Type     │ Chunks │ Added               │
├────┼─────────────────────┼──────────┼────────┼─────────────────────┤
│ 1  │ api-docs.pdf        │ PDF      │ 234    │ 2024-12-15 10:30    │
│ 2  │ security-guide.md   │ Markdown │ 45     │ 2024-12-15 10:32    │
│ 3  │ quickstart.txt      │ Text     │ 12     │ 2024-12-15 10:33    │
│ 4  │ faq.md              │ Markdown │ 28     │ 2024-12-16 09:15    │
│ 5  │ architecture.pdf    │ PDF      │ 156    │ 2024-12-16 14:20    │
└────┴─────────────────────┴──────────┴────────┴─────────────────────┘

Total: 5 documents, 475 chunks

Show Knowledge Base Info

praisonai knowledge info

Expected Output:

📊 Knowledge Base Information:

┌─────────────────────────┬────────────────────────────┐
│ Property                │ Value                      │
├─────────────────────────┼────────────────────────────┤
│ Location                │ .praison/knowledge/        │
│ Vector Store            │ ChromaDB                   │
│ Embedding Model         │ text-embedding-3-small     │
│ Total Documents         │ 5                          │
│ Total Chunks            │ 475                        │
│ Total Vectors           │ 475                        │
│ Storage Size            │ 12.3 MB                    │
│ Last Updated            │ 2024-12-16 14:20:00        │
└─────────────────────────┴────────────────────────────┘

Clear Knowledge Base

# Clear all documents
praisonai knowledge clear

Expected Output:

⚠️  This will delete all documents from the knowledge base.
Are you sure? (y/N): y

🗑️  Clearing knowledge base...
✅ Knowledge base cleared successfully!

Show Statistics

praisonai knowledge stats

Expected Output:

📊 Knowledge Base Statistics:
  workspace: /path/to/project
  vector_store: chroma
  retrieval_strategy: basic
  reranker: none
  index_type: vector
  query_mode: default
  document_count: 5

Export Knowledge Base

# Export to default timestamped file
praisonai knowledge export

# Export to specific file
praisonai knowledge export backup.json

# Export to specific path
praisonai knowledge export /path/to/knowledge_backup.json

Expected Output:

✅ Exported 12 documents to knowledge_export_20241226_103045.json

Import Knowledge Base

# Import from JSON file
praisonai knowledge import backup.json

# Import from specific path
praisonai knowledge import /path/to/knowledge_backup.json

Expected Output:

✅ Imported 12 documents from backup.json

Help

praisonai knowledge help

Expected Output:

Knowledge Commands:
  praisonai knowledge add <file|dir|url>   - Add document(s) to knowledge base
  praisonai knowledge query <question>     - Query knowledge base with RAG
  praisonai knowledge list                 - List indexed documents
  praisonai knowledge clear                - Clear knowledge base
  praisonai knowledge stats                - Show knowledge base statistics
  praisonai knowledge export <file.json>   - Export knowledge base to JSON file
  praisonai knowledge import <file.json>   - Import knowledge base from JSON file

Options:
  --workspace <path>           - Workspace directory (default: current)
  --vector-store <name>        - Vector store: chroma, pinecone, qdrant, weaviate, memory
  --retrieval <strategy>       - Retrieval: basic, fusion, recursive, auto_merge
  --reranker <name>            - Reranker: simple, llm, cross_encoder, cohere
  --index-type <type>          - Index: vector, keyword, hybrid
  --query-mode <mode>          - Query: default, sub_question, summarize
  --session <id>               - Session ID for persistence
  --db <path>                  - Database path for persistence

Advanced Options Reference

Vector Stores

Store	Description	Requirements
memory	In-memory (default, no persistence)	None
chroma	ChromaDB local vector store	pip install chromadb
pinecone	Pinecone cloud vector store	PINECONE_API_KEY
qdrant	Qdrant vector database	pip install qdrant-client
weaviate	Weaviate vector database	pip install weaviate-client

Retrieval Strategies

Strategy	Description
basic	Simple vector similarity search (default)
fusion	Multi-query with Reciprocal Rank Fusion
recursive	Depth-limited recursive retrieval
auto_merge	Merges adjacent chunks from same document

Rerankers

Reranker	Description	Requirements
simple	Keyword-based scoring (default)	None
llm	LLM-based relevance scoring	OPENAI_API_KEY
cross_encoder	Cross-encoder model	pip install sentence-transformers
cohere	Cohere Rerank API	COHERE_API_KEY

Index Types

Type	Description
vector	Vector similarity index (default)
keyword	BM25 keyword index
hybrid	Combined vector + keyword

Query Modes

Mode	Description
default	Standard RAG query
sub_question	Decomposes complex questions
summarize	Summarizes retrieved context

Supported File Types

Type	Extensions	Description
PDF	.pdf	PDF documents with text extraction
Markdown	.md, .mdx	Markdown files
Text	.txt	Plain text files
Code	.py, .js, .ts, etc.	Source code files
HTML	.html, .htm	Web pages
JSON	.json	JSON documents
CSV	.csv	Tabular data
URL	http://, https://	Web pages

Use Cases

Documentation Search

# Add project documentation
praisonai knowledge add ./docs/

# Query documentation
praisonai "How do I configure the database?" --knowledge

Code Understanding

# Add codebase
praisonai knowledge add ./src/

# Ask about code
praisonai "Explain the authentication flow" --knowledge

Research Assistant

# Add research papers
praisonai knowledge add ./papers/

# Query research
praisonai knowledge search "machine learning optimization techniques"

Using Knowledge with Agents

Once documents are added, agents can automatically query the knowledge base:

# Enable knowledge for agent
praisonai "Answer based on the documentation" --knowledge

Expected Output:

📚 Knowledge base enabled (5 documents)

╭─ Agent Info ─────────────────────────────────────────────────────────────────╮
│  👤 Agent: DirectAgent                                                       │
│  Role: Assistant                                                             │
│  📚 Knowledge: 5 documents, 475 chunks                                      │
╰──────────────────────────────────────────────────────────────────────────────╯

╭────────────────────────────────── Response ──────────────────────────────────╮
│ Based on the documentation, here's how to configure the database:            │
│                                                                              │
│ 1. Set the DATABASE_URL environment variable...                             │
│ [Response based on knowledge base content]                                   │
╰──────────────────────────────────────────────────────────────────────────────╯

📖 Sources:
  • api-docs.pdf (page 12)
  • quickstart.txt (section 2)

Best Practices

Organize documents by topic in separate directories for better search relevance.

Large documents are automatically chunked. Very large knowledge bases may increase response latency.

Document Quality

Use well-structured documents with clear headings

Regular Updates

Keep knowledge base updated with latest documentation

Specific Queries

Use specific search terms for better results

Organize by Topic

Group related documents together

Related

• Knowledge Concept
• RAG Features
• Fast Context CLI