tool · inferred from evidence
pip
A package installer for Python.
- Langflow components panel is empty after install (blank sidebar and canvas)
The empty components panel was caused by installing Langflow under the wrong Python version, which pulled in very old frontend assets. Langflow at the time needed Python at least 3.9 but below 3.11.
- stability-sdk client writes long-named empty files instead of PNG output
stability-sdk client writes long-named empty files instead of PNG output. If client.py writes long-named empty files instead of PNGs, your installed package is behind the fix.
- Missing Python 3.10 wheels for oso 0.22.1 (pip install fails)
Missing Python 3.10 wheels for oso 0.22.1 (pip install fails). The maintainer built and uploaded the Python 3.10 wheels to PyPI for oso 0.22.1, so a plain pip install works again.
- hubspot-api-python 11.0.0 breaks environments using urllib3 2.x
How to fix: hubspot-api-python 11.0.0 breaks environments using urllib3 2.x. The verified fix from the maintainer thread, distilled into reusable steps.
- KEY ERROR 1 while using classification preset in Evidently
Teaches the verified fix from a real evidently-ai support thread: kEY ERROR 1 while using classification preset in Evidently
- How to send POST requests with the ScrapingAnt Python client
What this covers: The ScrapingAnt Python client only sends GET requests through API v1. The fix that worked: POST requests were added in ScrapingAnt's API v2, and the Python client 1.x only targets API v1.
- ChromaDB install fails on Windows: Microsoft Visual C++ 14.0 or greater is required (building hnswlib)
Install the Microsoft C++ Build Tools, not just the Visual C++ redistributables: get them from visualstudio.microsoft.com/visual-cpp-build-tools, install the "Desktop development with C++" workload (including the MSVC v143 build tools), res
- `google-cloud-documentai` imports fail in GCP notebooks: kernel Python does not match pip
`google-cloud-documentai` imports fail in GCP notebooks: kernel Python does not match pip. The package is fine - the environment is the problem.
- Which pip package is the official Mailgun Python SDK?
The correct package is `mailgun`: `pip install mailgun`. The Mailgun team updated their docs after this report (https://documentation.mailgun.com/docs/mailgun/sdk/python_sdk/), so the official SDK is unambiguous now.
- ModuleNotFoundError paddle_billing.Resources.IPAddresses on 0.3.0 (fixed in 0.3.1)
paddle_billing 0.3.0 shipped with a broken internal import, so any import of the package failed. Fixed in 0.3.1; upgrade.
- Fix AttributeError when setting depends_on in Databricks job tasks
Passing plain strings to `depends_on` broke with `AttributeError: str object has no attribute as_dict`. The correct shape is a list of `jobs.TaskDependency` objects, e.g.
- modal setup: ImportError cannot import GRPCError from grpclib
The grpclib in the environment is not the real PyPI package: another package shadowed it. Check pip show grpclib for empty metadata; the reliable fix is a fresh virtualenv and a clean pip install modal.
- Fix confluent-kafka build error: librdkafka/rdkafka.h file not found
My `pip install confluent-kafka` died compiling because the C headers for librdkafka were missing. Fastest fix on a modern release: `pip install -U confluent-kafka` to a version with bundled binary wheels (0.11.4+), which never compiles any
- cannot import name 'Action' from 'composio' - ImportError when using Composio tools
The problem wasnt Composio itself, it was the PyPI package name. The bare composio package on PyPI was a stale shim by a different author that breaks the import, so anything you run against it blows up.
- Upgrading to dbt-core 1.8 breaks adapter installation
What this covers: Upgrading from dbt below 1.8 to dbt-core 1.8+ failed during install, because 1.8 split the adapter code out into the separate `dbt-adapters` package. The fix that worked: dbt-core 1.8 moved adapter code into the separate `
- ggshield pre-commit crashes with WinError 206 on large Windows merges
On Windows, ggshield secret scan pre-commit crashes with "filename or extension is too long" on big merges because its git command line exceeds Windows path limits. A fix was merged and shipped; upgrade ggshield.
- confluent-kafka install fails on Apple M1: just use 1.9.2 or newer
On Apple M1, pip install confluent-kafka failed with no arm64 wheels and a source build that could not find brew librdkafka headers. macOS arm64 wheels shipped in v1.9.2, so upgrade and skip the build dance.
- pip install chromadb fails building the hnswlib wheel on macOS
On macOS, pip install chromadb dies with clang errors while building hnswlib because of native CPU-architecture compile flags. Set HNSWLIB_NO_NATIVE=1 before installing.
- Phoenix protobuf error on import: Descriptors cannot be created directly
The real problem is pip backtracking to an old `opentelemetry-proto` (1.11.1) which only requires protobuf>=3.13, incompatible with protobuf v5. Fix: constrain `opentelemetry-exporter-otlp >=1.21.0` (or update all of `opentelemetry-proto`,
- Phoenix protobuf error on import: Descriptors cannot be created directly
Running the one-click Arize Phoenix integration with LlamaIndex (arize-phoenix 3.17.1, llama-index-core 0.10.22) fails on `import phoenix as px` with `TypeError: Descriptors cannot be created directly` from protobuf. Downgrading protobuf al