sprang-chat

Ask any question about the codebase using the knowledge graph.

Answer questions grounded entirely in .sprang/knowledge-graph.json.

1. Check .sprang/knowledge-graph.json exists. If not, tell the user to run /sprang-analyze first.
2. Read project metadata (project_name, description, languages) to anchor context.
3. Call sprang_query with keywords from the question to find relevant nodes — use multiple search terms if needed.
4. For each matched node, call sprang_node to get its 1-hop neighborhood (in/out degree, layer, neighbors).
5. Follow edges to trace the dependency chain as far as needed to fully answer.
6. Find which architectural layers the matched nodes belong to.
7. Answer grounded in graph data — cite summary fields directly, reference file paths, name layers.
8. Offer follow-ups: /sprang-explain <file> for deep dive, /sprang-diff if the topic involves recent changes.

$ARGUMENTS