Query Commands

vrin query

Query the knowledge base with natural language. VRIN retrieves relevant facts from your knowledge graph, combines them with matching document chunks, and generates a reasoned response.

vrin query "What were ACME's key financial metrics in Q4?"

Options

Flag	Short	Default	Description
`--stream / --no-stream`		`--no-stream`	Stream tokens to stdout as they arrive
`--mode`	`-m`	`chat`	Response mode: `chat`, `thinking`, `research`
`--depth`	`-d`		Query depth: `basic`, `thinking`, `research`
`--model`			LLM model override (e.g., `claude-sonnet-4-20250514`)
`--session-id`	`-s`		Continue a conversation session
`--web-search`		`false`	Enable web search augmentation
`--json`			Force JSON output

Response modes

chat
thinking
research

Default mode. Fast, conversational response.

vrin query "What is ACME's revenue?" --mode chat

Shows reasoning steps before the final answer. Good for complex questions.

vrin query "Compare ACME and Beta Corp revenue trends" --mode thinking

Exhaustive analysis with maximum retrieval depth. Uses more facts and chunks.

vrin query "What are the key risk factors across our portfolio?" --mode research

Query depth

Controls how deeply VRIN searches your knowledge graph:

Depth	Behavior	Best for
`basic`	Fast vector + graph retrieval	Simple factual lookups
`thinking`	Cross-document analysis, multi-hop reasoning	Questions connecting multiple sources
`research`	Exhaustive graph traversal, maximum fact coverage	Complex analytical queries

vrin query "How do risk factors in our Q3 and Q4 reports compare?" --depth research

Streaming

Stream tokens to stdout as they arrive. Metadata is written to stderr after streaming completes.

vrin query "Summarize our competitive position" --stream

With --json, the full response is buffered and output as JSON after streaming completes:

vrin query "Summarize our competitive position" --stream --json

Conversations

Continue a multi-turn conversation by passing a session ID:

# Start a conversation
vrin query "What was ACME's Q4 revenue?" --json
# Response includes session_id

# Continue it
vrin query "How does that compare to Q3?" --session-id "sess_abc123"

See conversation commands for managing conversation history.

vrin facts

Fast fact retrieval without AI summary generation. Returns raw structured facts from the knowledge graph. Useful for quick lookups or when you want to process facts programmatically.

vrin facts "ACME revenue"

Options

Flag	Description
`--json`	Force JSON output

Example output (JSON)

vrin facts "ACME revenue" --json

{
  "ok": true,
  "data": {
    "facts": [
      {
        "subject": "ACME Corp",
        "predicate": "reported_revenue",
        "object": "$50M in Q4 2025",
        "confidence": 0.95,
        "source": "ACME Q4 2025 Earnings"
      }
    ],
    "count": 1
  }
}

Get Started

MCP Server

CLI

Python SDK

TypeScript SDK

Concepts

Guides

Query Commands

vrin query

Options

Response modes

Query depth

Streaming

Conversations

vrin facts

Options

Example output (JSON)

​vrin query

​Options

​Response modes

​Query depth

​Streaming

​Conversations

​vrin facts

​Options

​Example output (JSON)

vrin query

Options

Response modes

Query depth

Streaming

Conversations

vrin facts

Options

Example output (JSON)