We don't publish
your competitive advantage.

IFWhen the npm registry does not have the latest version of @modelcontextprotocol/server-memory (e.g., 0.6.3) despite the fix being committed in the repository.

THENUse a git-based dependency in your package.json to reference the specific commit hash containing the fix. For example, `"@modelcontextprotocol/server-memory": "github:modelcontextprotocol/servers#04209ec"`. Alternatively, wait for the maintainers to cut a release and trigger the publish CI.

IFImporting litellm.proxy raises ModuleNotFoundError for 'enterprise' after PR #10321.

THENPin litellm to version 1.67.2 or lower in your requirements until a patched version (1.67.4.post1) is released. Check available versions with 'pip index versions litellm'.

IFUser encounters ModuleNotFoundError for 'tiktoken' when running CrewAI.

THENInstall the optional embeddings or tools extras: `uv pip install 'crewai[embeddings]'` or `uv pip install 'crewai[tools]'`. If wheel build fails for tiktoken, ensure Rust compiler is installed or use `uv pip install tiktoken --prefer-binary`.

IFUpgrading mcp-use to 1.24.1 introduces @mcp-use/inspector@2.1.0 which has hard dependencies on @langchain packages, causing ERR_MODULE_NOT_FOUND in bundlers like Next.js.

THENPin mcp-use to version 1.22.3 in package.json to avoid the problematic transitive dependency. Alternatively, wait for a fix where langchain becomes an optional peer dependency or is moved to a separate package.

IFLatest LiteLLM version adds dependency on 'pondpond' which requires 'madoka' (no prebuilt wheels) and fails in environments without C++ compiler (e.g., clean Python Docker image).

THENPin LiteLLM to version 1.77.1 or earlier to avoid the problematic dependency. Alternatively, install system build tools (e.g., g++, build-essential) before pip install, or use a Docker image that includes a C++ compiler. For long-term stability, monitor upstream for wheels or alternative packages.

IFImporting from langchain (e.g., `from langchain import OpenAI` or `from langchain.chat_models import ChatOpenAI`) fails with `ModuleNotFoundError: No module named 'langchain_community'` after upgrading to langchain v0.2.

THENInstall the `langchain-community` package separately using `pip install langchain-community`. Additionally, update any imports that rely on community modules (e.g., `OpenAI`, `ChatOpenAI`) to import from `langchain_community` instead of `langchain` if needed, as langchain v0.2 split out community integrations into a separate package.

IFvLLM installation fails with dependency resolution error because required PyTorch nightly build (torch==2.9.0.dev20250804+cu128) is no longer available.

THENReplace the outdated PyTorch nightly dependency with a direct wheel URL for PyTorch 2.9.0+cu128 (e.g., from download.pytorch.org/whl/cu128) and install vLLM from the gpt-oss wheels index using `uv pip install --pre vllm --extra-index-url https://wheels.vllm.ai/gpt-oss/ --index-strategy unsafe-best-match`.

IFImporting litellm.proxy fails with ModuleNotFoundError: No module named 'enterprise'.

THENPin litellm to version 1.67.2 or earlier as a temporary workaround. The fix is available in version 1.67.4.post1, so upgrade to that or later once released. Alternatively, downgrade using `pip install litellm==1.67.2`.

IFQwen2.5-VL model fails with ValueError when using vLLM due to upstream transformations in the transformers library.

THENPin the transformers library to a known working commit by running 'pip install --upgrade git+https://github.com/huggingface/transformers.git@336dc69d63d56f232a183a3e7f52790429b871ef' until the upstream breaking change is resolved.

IFAdding llama-index via poetry fails with error 'llama-index requires Python >=3.8.1,<3.12' when the project's Python version range includes 3.12 or higher.

THENUpdate your project's pyproject.toml to restrict Python to a version below 3.12, e.g. 'python = ">=3.11,<3.12"'. Then run 'poetry lock --no-update' to update the lock file without changing dependencies, followed by 'poetry install' and 'poetry add llama-index@latest'.

IFA Python library pins exact dependency versions in pyproject.toml, causing forced downgrades and conflicts for downstream consumers.

THENMove exact version pins out of pyproject.toml into a separate constraints file (e.g., constraints.txt) used only for Docker builds or lockfiles. Restore range-based dependency specs (e.g., pydantic>=2.5,<3) in pyproject.toml to express compatibility without locking environments.

IFnpm v7+ fails with ERESOLVE unable to resolve dependency tree during bootstrap or install due to strict peer dependency resolution.

THENAdd `legacy-peer-deps=true` to your .npmrc file or pass `--legacy-peer-deps` flag to npm install/lerna bootstrap. This reverts to npm v6 behavior, ignoring peer dependency conflicts. Alternatively, update conflicting dependencies to compatible versions.