We don't publish
your competitive advantage.

IFA new dependency on `madoka` (via `pondpond`) in LiteLLM v1.77.x causes build failures in environments without a C++ compiler (missing `g++` or `stdexcept`).

THENPin LiteLLM to version <1.77.0 (e.g., use 1.76.x) to avoid the problematic dependency. For Docker builds, you may also install a C++ build toolchain, but pinning is the immediate fix.

IFImporting from tenacity.asyncio fails after upgrading tenacity to version 8.4.0 due to removal of the asyncio submodule.

THENPin tenacity to a version before 8.4.0 (e.g., >=8.2.0,<8.4.0) as a temporary workaround. For a permanent fix, update all import statements from 'from tenacity.asyncio import AsyncRetrying' to 'from tenacity import AsyncRetrying'.

IFImporting AsyncChromiumLoader from langchain_community raises AttributeError: 'Language' enum has no attribute 'COBOL'.

THENUpgrade langchain to a version that includes the COBOL attribute (e.g., >=0.0.354) or reinstall langchain completely. The error occurs due to a missing enum value in outdated versions. Run `pip install --upgrade langchain` to resolve.

IFPackage declares Node 18 support but a direct dependency requires Node >=20, causing engine incompatibility errors on Node 18.

THENEnsure all direct dependencies are compatible with the declared engine range in package.json. If a dependency requires a newer Node version, either update the declared minimum Node version or find an alternative compatible dependency.

IFUsing openai library v1.0.0 or later with legacy LangChain code that references openai.Completion triggers AttributeError.

THENPin openai to v0.28.1 or upgrade LangChain to a version that supports the openai v1.x API. If upgrading LangChain, ensure the code uses openai.ChatCompletion instead of the deprecated openai.Completion.

IFWhen using LangChain with tenacity library upgraded to version 8.4.0, importing tenacity fails with ModuleNotFoundError for tenacity.asyncio.

THENPin tenacity to version 8.3.0 by running 'pip install tenacity==8.3.0' or update LangChain to a version that is compatible with tenacity 8.4.0. This ensures the asyncio submodule is available.

IFUsing arize-phoenix-evals version 3.0.0 with arize-phoenix 4.35.0 causes ModuleNotFoundError for 'phoenix.evals.models'.

THENDowngrade arize-phoenix-evals to version 2.13.0 to restore the 'evals.models' module. Alternatively, upgrade arize-phoenix to a version compatible with evals 3.0.0, if available.

IFUsing langfuse with Python 3.14 causes a ConfigError because the library depends on Pydantic v1, which does not support Python 3.14.

THENUpgrade langfuse to a version that uses Pydantic v2 or later to maintain compatibility with Python 3.14. The error originates from pydantic.v1 fields module when trying to infer attribute types. Moving to Pydantic v2 resolves this class incompatibility.

IFUsing DocArrayInMemorySearch with VectorstoreIndexCreator raises a Pydantic ValidationError because pydantic v2 lacks required 'text' and 'metadata' fields.

THENDowngrade pydantic to version 1.x (e.g., 1.10.13) to ensure compatibility with LangChain's DocArray integration. Run 'pip install --upgrade pydantic==1.10.13' in your environment.

IFImporting from langchain.chains (e.g., 'from langchain.chains import *') with langchain 0.0.353 and langchain-core 0.1.5 causes ModuleNotFoundError: No module named 'langchain_core.tracers.langchain_v1'.

THENDowngrade langchain-core to version 0.1.4 via pip install langchain-core==0.1.4, or alternatively downgrade the entire langchain package to 0.0.349 with its compatible dependencies (langchain-community 0.0.1 and langchain-core 0.0.13).

IFImportError: undefined symbol __nvJitLinkComplete_12_4 when importing torch after installing vllm (e.g., vllm 0.6.3.post2.dev386) which pulls torch 2.5.1.

THENDowngrade torch to version 2.4.0 which is known to work with vllm 0.6.3.post1. Alternatively, apply workaround from PyTorch issue #111469 (e.g., set LD_PRELOAD to a compatible libnvJitLink.so) or wait for PyTorch fix in a future release.

IFError when initializing LangChain OpenAI LLM with openai package version >=1.1.1 due to deprecated imports and API changes.

THENDowngrade openai package to version 0.28.1 using 'pip install openai==0.28.1', or update the LangChain code to use the new openai client API (e.g., openai.ApiClient and openai.OpenAIError).