PlanOpticon

Video analysis and knowledge extraction CLI. Processes recordings into structured knowledge graphs with entities, relationships, and insights.

Knowledge Graph Query Skill

PlanOpticon can build and query knowledge graphs from video content. If you see knowledge_graph.db or knowledge_graph.json files in the workspace, you can query them to understand what was discussed.

Auto-detection

Look for these files (checked automatically): - knowledge_graph.db — SQLite graph database (preferred) - knowledge_graph.json — JSON export (fallback)

Common locations: project root, results/, output/, knowledge-base/.

Quick commands

# Show graph stats (entity/relationship counts)
planopticon query

# List entities filtered by type
planopticon query "entities --type technology"
planopticon query "entities --type person"

# Search entities by name
planopticon query "entities --name python"

# See what connects to an entity
planopticon query "neighbors Alice"

# List relationships
planopticon query "relationships --source Alice"

# Natural language (requires API key)
planopticon query "What technologies were discussed?"
planopticon query "Who are the key people mentioned?"

# Output as JSON or Mermaid diagram
planopticon query --format json stats
planopticon query --format mermaid "neighbors Alice"

# Interactive REPL
planopticon query -I

When to use

• Direct mode (stats, entities, neighbors, relationships): No API key needed. Fast, deterministic. Use for structured lookups.
• Agentic mode (natural language questions): Requires an API key (OPENAI_API_KEY, ANTHROPIC_API_KEY, etc.). Use when the user asks open-ended questions about the content.

Python API

from video_processor.integrators.graph_query import GraphQueryEngine
from video_processor.integrators.graph_discovery import find_nearest_graph

path = find_nearest_graph()
engine = GraphQueryEngine.from_db_path(path)
result = engine.stats()
print(result.to_text())