Type check the whole project instead of seam/resources with error codes disabled

[razor-x]

Why

The mypy step checked seam/resources with two error codes disabled:

uv run mypy seam/resources --disable-error-code=arg-type --disable-error-code=import-not-found

That covered 34 of the 94 files in the package and left every hand-written module unchecked. One of them was broken in a way that affected every downstream user: seam/__init__.py carried a blanket # type: ignore, so despite the package shipping py.typed, the documented import failed for anyone type checking their own code.

$ cat app.py
from seam import Seam
$ mypy app.py
app.py:1: error: Module "seam" has no attribute "Seam"  [attr-defined]

Nothing in this repository could catch that, because the file was never checked.

What changed

just lint now runs uv run mypy seam test with nothing disabled. Getting there meant fixing what the flags were hiding — 2150 errors across 76 files, from four root causes:

Root cause	Errors	Fix
Generated from_dict reads every field with dict.get, so fields the payload omits arrive as None behind types that rule None out	1720	Optional field types + from_dict takes the decoded payload as Any
Generated json_payload = {} had its type inferred from whichever parameter was written first	288	Annotate as Dict[str, Any]
SeamHttpClient overrides request to return the decoded body, but post/get kept the inherited Response return type	120	Override both verb helpers
Optional handling in the hand-written modules	~22	Individual fixes

--disable-error-code=import-not-found is dropped outright. It suppressed nothing — it was only needed because mypy was pointed at a subtree; checking the package resolves those imports.

Bugs this surfaced

• seam/__init__.py hid the package from type checkers, breaking from seam import Seam for downstream users (above).
• SeamPaginator.flatten_to_list read pagination.has_next_page without the None guard its sibling flatten already had, so a page whose response carried no pagination raised AttributeError instead of ending the loop.
• SeamHttpClient._handle_response compared response.status_code without accounting for it being unset, which would raise TypeError rather than handling the response.
• SeamActionAttemptFailedError read action_attempt.error.message unguarded, which would raise AttributeError over the actual failure.
• SeamActionAttemptTimeoutError documented its timeout as str while every caller passes a float.
• HTTP exceptions typed request_id as str, though it comes from a response header that may be absent.

Note on the generated Optional types

This is the breaking part, which is why it is aimed at beta.

Resource fields that the API may omit or send as null are now typed Optional. Runtime behavior is unchanged — those fields could already be None, the annotations just denied it — but code that reads them under a type checker may now need a None check.

Nested objects are Optional regardless of what the schema says. The schema is not a reliable guide here: an action attempt documents both error and result as required, yet a pending one carries neither. Constructing them unconditionally would fail on those payloads, so from_dict keeps its None fallback and the field admits it.

No field gains a default, so the dataclass constructors are unchanged.

Two invariants are beyond what the checker can follow and are ignored in place with a comment: the auth option guards return True only once the values they check are set, and Seam borrows Routes.__init__ to attach the route namespaces even though the two are siblings under AbstractRoutes rather than parent and child.

Verification

• mypy seam test — clean, 109 files, no disabled error codes
• pylint ./seam ./test — 10.00/10
• black --check ., rstcheck README.rst — clean
• pytest --cov=./seam — 83 passed
• tsc, eslint, prettier --check — clean for the codegen changes
• Generated code was produced by npm run generate, not edited by hand

---

Generated by Claude Code