AlgoScan

skip-checks: true

* feat: adds runtime validation using Pydantic schemas

Implements runtime validation of API client responses using Pydantic schemas.
This provides an additional layer of sanity checks on top of generated API clients.

The schemas are generated from OpenAPI specifications using a new script.
Adds a new `poe` task to easily regenerate schemas.
Adds tests to validate the generated schemas.

Pydantic is added as a development dependency.

Validation schemas are optional, requiring `pydantic>=2.0` to be installed.

* chore: generate the validation schemas

* chore: move schemas to tests folder to avoid polluting prod packages exposed to end users

* fix: unblock check-python after alpha merge

* fix: enforce required fields, fix docs, restore targeted lint rules

- Generator: required fields no longer get default=None (Pydantic
  actually validates them now)
- Generator: only emit `from typing import Any` when used
- Generator: add timeout and error handling to fetch_spec
- Docs: fix stale import paths in README.md and VALIDATION.md after
  schema move to tests/fixtures/
- Lint: revert global PLC0415/FBT001/FBT002/PLW1641 suppression,
  restore targeted noqa comments on specific lines
- Tests: provide all required fields, add test_missing_required_fields
- Snapshots: update mock server snapshots to match current HAR data

* feat: add schema validation to polytests, aligned with TS approach

Mirrors the TypeScript variant's pattern of calling Schema.parse(result)
in every polytest to validate real API responses against OAS-generated
schemas, closing the gap between spec and implementation.

Schema generator improvements:
- Use IdentifierSanitizer-compatible _to_snake() for field naming
  (handles camelCase, Python builtins, x-algokit-field-rename)
- Handle string/enum types as RootModel[str]
- Handle byte arrays (uint8[]) as RootModel[str] (base64)
- Handle opaque empty schemas as RootModel[str]

Polytest changes:
- Added validate_with_schema() to 55 polytests across algod, indexer,
  and kmd clients
- Serializer now handles Python Enum values correctly
- Updated snapshots for enum serialization changes
- 1 known OAS spec mismatch documented (apps-local-states required
  but API returns null)

* chore: deduplicate RootModel templates, fix missing newlines at EOF

- Refactor generate_schema() to extract _root_model() and
  _is_byte_array() helpers, eliminating 3 copies of identical
  RootModel[str] template
- Fix missing newline at EOF in 6 docs/config files (pre-commit)

* fix: update snapshots for enum serialization (repr → value)

Snapshots now use enum .value ("noop") instead of repr
("<OnCompletion.NOOP: 'noop'>") to match the corrected
_dataclass_to_dict serializer. Only enum formatting changed —
no mock server data diffs.

* refactor: consolidate 211 schema files into 3 modules

Replace per-class file structure (algod/82 files + indexer/72 files +
kmd/54 files + 3 __init__.py) with one module per client (algod.py,
indexer.py, kmd.py).

Benefits:
- 211 files → 3 files
- Eliminates __init__.py boilerplate (imports, model_rebuild, __all__)
- Eliminates forward reference strings — topological sort enables
  direct class references
- from __future__ import annotations handles remaining edge cases
- No ruff per-file overrides needed (was 9 suppressed rules)
- Import paths unchanged: `from tests.fixtures.schemas.algod import X`

Generator improvements:
- Topological sort of schemas by dependency order
- PascalCase class names (ed25519PublicKey → Ed25519PublicKey)
- Docstring escaping and truncation for ruff compliance
- Cleaner module assembly — shared imports at top