ChromaDB 0.6.3 telemetry spam on every operation in mempalace 3.1.0

[fubak]

Bug

Every ChromaDB operation in mempalace 3.1.0 prints a telemetry error to stderr:

Failed to send telemetry event ClientStartEvent: capture() takes 1 positional argument but 3 were given
Failed to send telemetry event CollectionGetEvent: capture() takes 1 positional argument but 3 were given

The actual functionality works — this is ChromaDB's internal telemetry calling posthog with the wrong signature — but it floods logs and hook output, making real errors harder to spot.

Steps to Reproduce

import chromadb
client = chromadb.PersistentClient(path='~/.mempalace/palace')
coll = client.get_collection('mempalace_drawers')
coll.count()
# → Failed to send telemetry event ClientStartEvent: capture() takes 1 positional argument but 3 were given
# → Failed to send telemetry event CollectionGetEvent: capture() takes 1 positional argument but 3 were given

Same noise appears in the MCP server logs, hook output logs (~/.mempalace/hook_state/hook.log), and every CLI command that touches ChromaDB.

Fix

Disable ChromaDB telemetry explicitly when constructing the client:

import chromadb
from chromadb.config import Settings
client = chromadb.PersistentClient(
    path=palace_path,
    settings=Settings(anonymized_telemetry=False)
)

Or pin posthog to a version compatible with chromadb 0.6.3's telemetry calls.

Environment

• mempalace 3.1.0
• chromadb 0.6.3
• posthog 7.10.3

[goog]

after i upgrade to 3.1.0, why chromadb downgrade from 1.1.1 to 0.6.3

Downloading chroma_hnswlib-0.7.6-cp311-cp311-win_amd64.whl (151 kB)
Downloading opentelemetry_instrumentation_fastapi-0.55b1-py3-none-any.whl (12 kB)
Downloading opentelemetry_instrumentation-0.55b1-py3-none-any.whl (31 kB)
Downloading opentelemetry_instrumentation_asgi-0.55b1-py3-none-any.whl (16 kB)
Downloading opentelemetry_util_http-0.55b1-py3-none-any.whl (7.3 kB)
Downloading asgiref-3.11.1-py3-none-any.whl (24 kB)
Downloading wrapt-1.17.3-cp311-cp311-win_amd64.whl (38 kB)
Installing collected packages: wrapt, opentelemetry-util-http, chroma-hnswlib, asgiref, opentelemetry-instrumentation, opentelemetry-instrumentation-asgi, opentelemetry-instrumentation-fastapi, chromadb, mempalace
  Attempting uninstall: wrapt
    Found existing installation: wrapt 2.1.1
    Uninstalling wrapt-2.1.1:
      Successfully uninstalled wrapt-2.1.1
  Attempting uninstall: chromadb
    Found existing installation: chromadb 1.1.1
    Uninstalling chromadb-1.1.1:
      Successfully uninstalled chromadb-1.1.1
  Attempting uninstall: mempalace
    Found existing installation: mempalace 3.0.0
    Uninstalling mempalace-3.0.0:
      Successfully uninstalled mempalace-3.0.0
Successfully installed asgiref-3.11.1 chroma-hnswlib-0.7.6 chromadb-0.6.3 mempalace-3.1.0 opentelemetry-instrumentation-0.55b1 opentelemetry-instrumentation-asgi-0.55b1 opentelemetry-instrumentation-fastapi-0.55b1 opentelemetry-util-http-0.55b1 wrapt-1.17.3

[web3guru888]

Can confirm the telemetry noise — we see the same capture() takes 1 positional argument but 3 were given on every ChromaDB operation.

The fix in the issue description is correct. We use this in our integration:

from chromadb.config import Settings
client = chromadb.PersistentClient(
    path=palace_path,
    settings=Settings(anonymized_telemetry=False)
)

This should be a one-line fix upstream in wherever MemPalace constructs the ChromaDB client. Disabling telemetry by default is the right call for a library that runs inside MCP servers and automated pipelines — the noise-to-signal ratio in logs matters when you're debugging real issues.

@goog — the downgrade from 1.1.1 to 0.6.3 is the same root cause as #457. MemPalace 3.1.0 declares a dependency range that resolves to 0.6.3. If you had a newer chromadb installed independently, pip downgrades it to satisfy the constraint.