Summary

mempalace status (any version that ships the current miner.py::status) fails with an uncaught Chroma InternalError on any palace that has more than ~32,766 drawers.

Symptom

$ mempalace status
Traceback (most recent call last):
  File ".../mempalace", line 7, in <module>
    sys.exit(main())
  File ".../mempalace/cli.py", line 681, in main
    dispatch[args.command](args)
  File ".../mempalace/cli.py", line 217, in cmd_status
    status(palace_path=palace_path)
  File ".../mempalace/miner.py", line 844, in status
    r = col.get(limit=total, include=["metadatas"]) if total else {"metadatas": []}
  File ".../mempalace/backends/chroma.py", line 65, in get
    return self._collection.get(**kwargs)
  File ".../chromadb/api/models/Collection.py", line 161, in get
    get_results = self._client._get(...)
  File ".../chromadb/api/rust.py", line 440, in _get
    rust_response = self.bindings.get(...)
chromadb.errors.InternalError: Error executing plan:
  Internal error: error returned from database:
  (code: 1) too many SQL variables

Exit code: 1.

Root cause

miner.py::status() calls:

total = col.count()
r = col.get(limit=total, include=["metadatas"]) if total else {"metadatas": []}

Chroma's backend issues a SQL statement whose host-parameter count grows with the limit. Once the limit exceeds SQLite's default SQLITE_MAX_VARIABLE_NUMBER (32766 on most standard SQLite builds), SQLite refuses the statement and Chroma surfaces InternalError: too many SQL variables.

The inflection point is exactly 32766: limits up to and including 32766 succeed; 32767+ fail.

Reproducer

On any environment with a palace that has more than 32 766 drawers — which is roughly what you get after mining a few months of conversations with the auto-save hooks enabled — simply:

mempalace status

Minimal programmatic repro:

from mempalace.config import MempalaceConfig
from mempalace.palace import get_collection
col = get_collection(MempalaceConfig().palace_path)
print(col.count())               # > 32766
col.get(limit=col.count(), include=["metadatas"])
# → chromadb.errors.InternalError: too many SQL variables

Proposed fix

Paginate col.get() at a batch size well below the SQLite limit (10 000 is safe) and aggregate metadatas. Diff is 9 added / 3 removed lines, no behavior change for smaller palaces. PR forthcoming.

Why this matters

mempalace status is the primary "is my palace healthy?" command. Power users with active auto-save tend to cross 32 K drawers within weeks, and when the command starts failing the traceback doesn't hint at the real cause (variable limit), which made it hard to diagnose.

Environment

• mempalace 3.3.1 (pip)
• chromadb 1.5.8
• Python 3.12.12
• macOS 15.1 (arm64), miniforge