Summary

• viking_fs.ls() defaults to node_limit=1000, which silently truncates results for directories with more than 1000 entries
• index_resource calls ls() without specifying node_limit, so large memory directories (e.g. 1800+ files) get silently truncated
• Files beyond the 1000 limit are never vectorized, resulting in incomplete vector indexes and missing search results

Root cause

viking_fs.ls() signature (line 1743):

async def ls(self, uri, ..., node_limit=1000)

index_resource (line 354) calls it with default:

files = await viking_fs.ls(uri, ctx=ctx)  # node_limit defaults to 1000

Fix

Pass node_limit=10000 explicitly:

files = await viking_fs.ls(uri, ctx=ctx, node_limit=10000)

Impact

Directories with up to 10,000 files will now be fully indexed. This fixes silent data loss during reindexing of large memory stores.

Test plan

• Create a directory with >1000 files, run index_resource
• Verify all files are vectorized (not just the first 1000)
• Check vector count matches file count

🤖 Generated with Claude Code