Skip to content
Command Line Interface

Command Line Interface

The SDK ships two management commands. warmup_rag pre-builds RAG pipelines so the first chat request doesn’t pay the indexing cost, and refresh_integrations refreshes MCP integration tool lists.

warmup_rag: Pre-warm RAG Indexes 

python manage.py warmup_rag

Warms up every agent that has a rag_provider configured, across all memories. Agents without a RAG provider are skipped.

Options:

Option Description
--agent <class_name> Only warm up a specific agent (class name, e.g. PirateBasicAgent)
--memory <memory_id> Only warm up a specific memory
--force-rebuild Rebuild indexes from scratch, deleting existing ones

Examples:

# Everything
python manage.py warmup_rag

# One agent
python manage.py warmup_rag --agent PirateBasicAgent

# One memory
python manage.py warmup_rag --memory <memory-uuid>

# Full rebuild (persistent backends like Qdrant recreate from scratch)
python manage.py warmup_rag --force-rebuild

When to Use It

  • Before deploying a new RAG setup: so indexes are ready when traffic arrives.
  • After document changes: to keep retrieval current.
  • As a scheduled job: e.g. a cron entry or django-tasks task that re-warms nightly.

You can also trigger the same work from code:

from django_ai_sdk import Agent

agent = await AgentService.get(agent_id)
await Agent.warmup(agent, memory_id=None)            # build + cache
await Agent.reindex(agent, memory_id=None, force_rebuild=False)
Agent.clear_rag_cache(agent)

See the RAG section of the Agents guide for the full lifecycle.

Contributor? The CLI Implementation manual page documents how the command works internally.
How It Works