Proposal: Improve static type checking across modules (via mypy)

[ArmaanjeetSandhu]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

I'd like to propose improving the overall static type checking coverage in the LangChain codebase. While type hints are present in many places, running mypy still surfaces a number of type errors and inconsistencies.

This work would involve:

• Identifying modules with frequent or critical type issues
• Resolving those issues with correct and meaningful type annotations
• Avoiding excessive use of Any or # type: ignore unless necessary
• Submitting scoped pull requests, module-by-module, to keep the work reviewable and focused

Motivation

This proposal stems from running mypy on the latest version of LangChain and encountering various errors and missing type info. For example, libs/langchain/langchain has these errors:

libs\langchain\langchain\output_parsers\yaml.py:5: error: Library stubs not installed for "yaml"  [import-untyped]
libs\langchain\langchain\chains\base.py:11: error: Library stubs not installed for "yaml"  [import-untyped]
libs\langchain\langchain\evaluation\string_distance\base.py:30: error: Cannot find implementation or library stub for module named "rapidfuzz"  [import-not-found]
libs\langchain\langchain\evaluation\embedding_distance\base.py:63: error: Incompatible import of "OpenAIEmbeddings" (imported name has type "type[langchain_community.embeddings.openai.OpenAIEmbeddings]", local name has type "type[langchain_openai.embeddings.base.OpenAIEmbeddings]")  [assignment]
libs\langchain\langchain\evaluation\embedding_distance\base.py:63: note: Error code "assignment" not covered by "type: ignore" comment
libs\langchain\langchain\evaluation\embedding_distance\base.py:124: error: Incompatible import of "OpenAIEmbeddings" (imported name has type "type[langchain_community.embeddings.openai.OpenAIEmbeddings]", local name has type "type[langchain_openai.embeddings.base.OpenAIEmbeddings]")  [assignment]
libs\langchain\langchain\evaluation\embedding_distance\base.py:124: note: Error code "assignment" not covered by "type: ignore" comment
libs\langchain\langchain\evaluation\parsing\json_schema.py:46: error: Library stubs not installed for "jsonschema"  [import-untyped]
libs\langchain\langchain\evaluation\parsing\json_schema.py:46: note: Hint: "python3 -m pip install types-jsonschema"
libs\langchain\langchain\evaluation\parsing\json_distance.py:49: error: Cannot find implementation or library stub for module named "rapidfuzz"  [import-not-found]
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:327: error: Signature of "evaluate_run" incompatible with supertype "RunEvaluator"  [override]
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:327: note:      Superclass:
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:327: note:          def evaluate_run(self, run: Run, example: Example | None = ..., evaluator_run_id: UUID | None = ...) -> EvaluationResult | EvaluationResults
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:327: note:      Subclass:
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:327: note:          def evaluate_run(self, run: Run, example: Example | None = ...) -> EvaluationResult
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:341: error: Signature of "aevaluate_run" incompatible with supertype "RunEvaluator"  [override]
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:341: note:      Superclass:
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:341: note:          def aevaluate_run(self, run: Run, example: Example | None = ..., evaluator_run_id: UUID | None = ...) -> Coroutine[Any, Any, EvaluationResult | EvaluationResults]
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:341: note:      Subclass:
libs\langchain\langchain\smith\evaluation\string_run_evaluator.py:341: note:          def aevaluate_run(self, run: Run, example: Example | None = ...) -> Coroutine[Any, Any, EvaluationResult]
libs\langchain\langchain\chains\llm_math\base.py:168: error: Cannot find implementation or library stub for module named "numexpr"  [import-not-found]
libs\langchain\langchain\agents\agent.py:25: error: Library stubs not installed for "yaml"  [import-untyped]
libs\langchain\langchain\agents\agent.py:25: note: Hint: "python3 -m pip install types-PyYAML"
libs\langchain\langchain\agents\agent.py:25: note: (or run "mypy --install-types" to install all missing stub packages)
libs\langchain\langchain\agents\agent.py:25: note: See https://mypy.readthedocs.io/en/stable/running_mypy.html#missing-imports
libs\langchain\langchain\evaluation\loading.py:62: error: Cannot find implementation or library stub for module named "datasets"  [import-not-found]
libs\langchain\langchain\evaluation\loading.py:138: error: Incompatible import of "ChatOpenAI" (imported name has type "type[langchain_community.chat_models.openai.ChatOpenAI]", local name has type "type[langchain_openai.chat_models.base.ChatOpenAI]")  [assignment]       
libs\langchain\langchain\evaluation\loading.py:138: note: Error code "assignment" not covered by "type: ignore" comment
libs\langchain\langchain\smith\evaluation\runner_utils.py:337: error: Argument 1 has incompatible type "dict[str, Any] | None"; expected
 "dict[Any, Any]"  [arg-type]
libs\langchain\langchain\smith\evaluation\runner_utils.py:350: error: Argument 1 to "_get_prompt" has incompatible type "dict[str, Any] | None"; expected "dict[str, Any]"  [arg-type]
libs\langchain\langchain\smith\evaluation\runner_utils.py:353: error: Argument 1 to "_get_messages" has incompatible type "dict[str, Any] | None"; expected "dict[str, Any]"  [arg-type]
libs\langchain\langchain\smith\evaluation\runner_utils.py:372: error: Argument 1 has incompatible type "dict[str, Any] | None"; expected
 "dict[Any, Any]"  [arg-type]
libs\langchain\langchain\smith\evaluation\runner_utils.py:786: error: Argument 2 to "_arun_llm" has incompatible type "dict[str, Any] | None"; expected "dict[str, Any]"  [arg-type]
libs\langchain\langchain\smith\evaluation\runner_utils.py:796: error: Argument 2 to "_arun_chain" has incompatible type "dict[str, Any] | None"; expected "dict[str, Any]"  [arg-type]
libs\langchain\langchain\smith\evaluation\runner_utils.py:938: error: Argument 2 to "_run_llm" has incompatible type "dict[str, Any] | None"; expected "dict[str, Any]"  [arg-type]
libs\langchain\langchain\smith\evaluation\runner_utils.py:948: error: Argument 2 to "_run_chain" has incompatible type "dict[str, Any] | None"; expected "dict[str, Any]"  [arg-type]
libs\langchain\langchain\chains\loading.py:9: error: Library stubs not installed for "yaml"  [import-untyped]
libs\langchain\langchain\chains\loading.py:224: error: Cannot find implementation or library stub for module named "langchain_experimental.llm_bash.base"  [import-not-found]
libs\langchain\langchain\chains\loading.py:341: error: Cannot find implementation or library stub for module named "langchain_experimental.pal_chain"  [import-not-found]
libs\langchain\langchain\chains\loading.py:404: error: Cannot find implementation or library stub for module named "langchain_experimental.sql"  [import-not-found]
libs\langchain\langchain\agents\loading.py:8: error: Library stubs not installed for "yaml"  [import-untyped]
libs\langchain\langchain\chains\openai_functions\citation_fuzzy_match.py:33: error: Library stubs not installed for "regex"  [import-untyped]
libs\langchain\langchain\chains\openai_functions\citation_fuzzy_match.py:33: note: Hint: "python3 -m pip install types-regex"
libs\langchain\langchain\memory\entity.py:118: error: Cannot find implementation or library stub for module named "upstash_redis"  [import-not-found]
libs\langchain\langchain\memory\entity.py:209: error: Cannot find implementation or library stub for module named "redis"  [import-not-found]

Proposal (If applicable)

I'd be happy to start this process by selecting a small, self-contained module, resolving its type errors, and opening a PR.

Before doing so, I'd like to confirm:

• Whether this work would be welcomed by the maintainers
• If there's a preferred area of the codebase to focus on first

Looking forward to your feedback!