Command Reference

Infrastructure

# Check all services
python scripts/start_test_services.py --status

# Start all services
python scripts/start_test_services.py

# List Incus containers
incus list

Tests

# Semantic graph tests
uv run pytest tests/semantic_graph/ -v --timeout=120

# Full suite (batched executor, NEVER xdist)
python scripts/run_tests.py

# Subset with gates skipped
python scripts/run_tests.py tests/semantic_graph/ --skip-gates

orchgraph (Memgraph HA)

HA is managed natively by the Raft coordinators — there is no HTTP HA controller. Connect the routing driver at neo4j://10.53.1.160:7691 (auto-routes to MAIN/replicas); query cluster state over a Bolt connection to any coordinator on port 7691.

# Cluster state / current MAIN — SHOW INSTANCES on a coordinator (auth=None)
python3 - <<'PY'
from neo4j import GraphDatabase
d = GraphDatabase.driver("bolt://10.53.1.160:7691", auth=None)
with d.session() as s:
    for rec in s.run("SHOW INSTANCES"):
        print(dict(rec))
d.close()
PY

Failover/failback are handled automatically by the coordinators (Raft); no manual curl endpoint. See Memgraph HA Cluster Reference.

Decommissioned 2026-07-10: the old HTTP HA controller (memgraph-ha on :8080, with /main /status /failover /failback) and its data nodes (memgraph-1/2/3) were deleted and replaced by orchgraph.

Wiki.js CLI (wiki)

# List / read pages
wiki list
wiki get home
wiki get infrastructure/memgraph-cluster
wiki search memgraph

# Create / update a page from a markdown file
wiki create infrastructure/redis --title "Redis Queue" --file infrastructure/redis.md
wiki update infrastructure/redis --file infrastructure/redis.md

# Bulk import all markdown files
python3 scripts/bulk-import-wiki.py

wiki-kimi CLI

wiki-kimi remember --title "Decision X" --tags decision,project:orchestrata-ultimate
wiki-kimi recall --query memgraph
wiki-kimi handoff --message "Next: implement controller failover"
wiki-kimi validate
wiki-kimi index

Graph Indexer

# Rebuild the Memgraph HA graph index
python3 graph-indexer/indexer.py

# Watch for changes
python3 graph-indexer/indexer.py --watch

# Load old orchestrata-ultimate sessions from context history
python3 graph-indexer/load_old_sessions.py

RAG Pipeline

# Ingest all wiki pages into the vector store
wiki-rag-ingest

# Query across all projects
wiki-rag-query "How does the Memgraph HA controller work?" --top-k 5

# Query within a single project
wiki-rag-query "What are the current blockers?" --project orchestrata-ultimate --top-k 5

# Expand with graph traversal
wiki-rag-query "Explain the repository pattern" --graph-expand --graph-depth 1 --graph-limit 10

Todoist Integration (todoist-kimi)

# List projects and tasks
todoist-kimi projects
todoist-kimi tasks

# Add a task
todoist-kimi add "Fix Neo4j connection retry" --due today --priority 3

# Close a task
todoist-kimi close <task-id>

# Sync active wiki blockers to Todoist
todoist-kimi sync-blockers --project orchestrata-ultimate

Session

There is no bare session binary — invoke the CLI wrapper directly (it auto-detects the project root by walking up for context.json/context.md):

python3 ~/.claude/tools/session_cli.py status
python3 ~/.claude/tools/session_cli.py begin --purpose "Purpose"
python3 ~/.claude/tools/session_cli.py log-work --desc "Did a thing"
python3 ~/.claude/tools/session_cli.py log-artifact --path path/to/file.md --lines 42 --summary "Created page"
python3 ~/.claude/tools/session_cli.py set-handoff --decisions "..." --blockers "..." --active-work "..."
python3 ~/.claude/tools/session_cli.py end --upcoming "Next task"

See claude/tools/session_cli.md for the full command reference.

GitLab Push

cd ~/workspace/knowledge-wiki
unset GITLAB_TOKEN
git add -A && git commit -m "backup: $(date -I)" && git push origin master

The local/tracked branch is master (origin/master), not main — origin/main on GitLab is a stale, disconnected branch last touched 2026-07-09. Pushing to main silently misses the real history.