fix: serialize and deserialize typing.Literal

[LK-maker-007]

Related Issues

• fixes serialize_type / deserialize_type do not support typing.Literal #12285

Proposed Changes:

serialize_type had no handling for typing.Literal, so it fell through to the generic-path logic and rendered the literal's values as if they were type names:

serialize_type(Literal["yes", "no"])   # -> "typing.Literal[yes, no]"

deserialize_type then tried to resolve yes and no as types and raised DeserializationError. A value that happened to look like a real type name silently corrupted instead of raising: Literal["int", "str"] round-tripped to Literal[int, str] (two types).

The fix special-cases typing.Literal on both sides:

• serialize: render each value with repr(), so strings keep their quotes and non-string values (int, bool, bytes, None) are emitted as valid Python literals.
• deserialize: parse the arguments with ast.literal_eval instead of the type-resolving generic path. ast.literal_eval is quote-aware, so a comma inside a string value (Literal["a, b", "c"]) no longer splits the arguments, and it is limited to safe literals (str/int/bool/None/bytes/tuples).

Scope note: this covers the literal value kinds Python's own Literal accepts and that survive a text round-trip (str, bytes, int, bool, None). Enum members as Literal values are out of scope — they can't be reconstructed from repr() alone and would need import-path handling; I left that out rather than half-support it. Happy to follow up if you want it.

How did you test it?

Unit tests in test/utils/test_type_serialization.py (serialization, deserialization, and round_trip for Literal, mirroring the existing Callable trio), covering the failing case, the silent-corruption case (Literal["int", "str"]), non-string values, a comma inside a value, and Literal nested in Optional/Union. Proven fail-without/pass-with: with the source change stashed the three new tests fail with DeserializationError; with it applied they pass.

Full file green (195 passed), plus the consumer suites: test_output_adapter.py, test_conditional_router.py, test_pipeline.py (80 passed). ruff check / ruff format --check clean, mypy clean on the changed module.

End-to-end check — this previously crashed on load, now round-trips:

p = Pipeline()
p.add_component("oa", OutputAdapter(template="{{ x }}", output_type=Literal["yes", "no"]))
Pipeline.loads(p.dumps())   # OK

Notes for the reviewer

The Literal branch is placed before the generic-args parsing on both sides on purpose: typing.get_origin(Literal[...]) is typing.Literal, and its args are values, so they must not go through the type-resolving path (_parse_generic_args / deserialize_type per arg).

Checklist

• I have read the contributors guidelines and the code of conduct.
• I have updated the related issue with new insights and changes.
• I have added unit tests and updated the docstrings.
• I've used one of the conventional commit types for my PR title: fix:.
• I have documented my code.
• I have added a release note file, following the contributors guidelines.
• I have run pre-commit hooks and fixed any issue.

[vercel]

@LK-maker-007 is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[sjrl]

@anakin87 I can take review of this one

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
haystack/utils
type_serialization.py
Project Total

This report was generated by python-coverage-comment-action

[LK-maker-007]

Applied all four: trimmed both comments, removed the test comment, and switched the release note to RST double backticks throughout (not just the flagged line). Thanks for the review.

[sjrl]

Thanks!