fix(graph): resolve Literal collector items to the types of their values

[kevin9327]

Summary

A node output typed as a Literal (e.g. max_seq_len: Literal[256, 512] on the FLUX / FLUX.2 model loaders) can't be fed into a Collect node. Collector validation treats the Literal's values as types and calls issubclass(256, ...), so validation crashes with TypeError: issubclass() arg 1 must be a class instead of accepting or rejecting the edge. On the enqueue path this happens while the Batch body is parsed, so the API returns a 500 rather than a 422 (the same failure mode discussed in #9858).

_resolve_item_input_types now resolves a Literal[...] item to the types of its values (Literal[256, 512] → int, Optional[Literal["a", "b"]] → str), so the existing subtype checks work as intended:

• before: literal.value → collect.item → integer_collection.collection → TypeError (500 on enqueue)
• after: accepted; wiring the same items into a string collection is rejected with the normal InvalidEdgeError / 422 "Collector outputs must connect to a collection input with a matching type"

Related Issues / Discussions

• Found while tracing the 500s in fix(webv2): stop retrying an enqueue that keeps failing with a 500 #9858.
• Complementary to fix(graph): reject mismatched collector items for union collection inputs #9853, which guards the issubclass in _validate_collector_output_types for Union[list[X], X] inputs. With that guard alone a Literal item would be rejected even for a matching int collection. This PR fixes the item side, and the two don't touch the same lines.

QA Instructions

Added a small test_literal invocation to tests/test_nodes.py and four tests to tests/test_node_graph.py:

• Literal items into an int collection are accepted
• Literal items into a string collection get a clean InvalidEdgeError
• Optional[Literal[...]] items into a string collection are accepted
• a Batch holding the mismatched graph raises ValidationError (not TypeError)

All four fail on main with TypeError: issubclass() arg 1 must be a class and pass with the fix.

pytest tests/test_node_graph.py tests/test_nodes.py tests/test_graph_execution_state.py tests/test_session_queue.py
410 passed, 2 xfailed
ruff check .          # All checks passed
ruff format --check   # clean for the changed files

Review

No material findings. Only collector item-type resolution changes; non-Literal types resolve exactly as before.

Checklist

• The PR has a short but descriptive title, suitable for a changelog
• Meaningful regression coverage added / updated where needed; obsolete tests/code removed
• Persisted-state and API changes include required migrations / compatibility validation (n/a)
• Relevant performance/efficiency opportunities considered; material claims have evidence
• Material review findings resolved and relevant checks rerun
• Documentation added / updated (if applicable) (n/a)
• Updated What's New copy (if doing a release after this PR) (n/a)

---

AI-assisted: I used an AI assistant to help trace this and draft the change. I reviewed the code and tests and ran them locally.

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	⚠️ Failed 2026-10-17T02:14:56.526213Z	f590bd6	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.