feat(public-api): expose all router-facing APIs at cleveractors package level; update README #48

2026-06-12T14:26:14Z

hurui200320 commented

2026-06-12 14:26:14 +00:00

Summary

This PR completes Wave C8 (issue #17): finalising the cleveractors v2.1.0 public API surface by documenting all router-facing exports in README.md and fixing a pre-existing test-isolation bug in the coverage pipeline.

All code changes for C1–C7 (imports, __all__, version bump, smoke tests) were already merged to master in prior commits. This PR delivers the one remaining subtask: the "Router-facing API" README section, plus the bug fix.

Changes

`README.md` — New "Router-facing API" section (AC7)

Added a comprehensive "Router-facing API" section after the Quick start block, including:

Summary table listing all 8 router-facing exports (validate_dict, merge_configs, create_executor, Executor, ActorResult, NodeUsage, ExecutionError, ConfigurationError) with one-line descriptions. merge_configs correctly described as accepting "zero or more" dicts. create_executor row updated to "…per-request credentials, and optional execution limits and pricing table" (reflecting the optional parameters).
End-to-end usage example following the ADR-2024 "Library API Contract" pattern (validate at upload time, execute at request time, bill from ActorResult.nodes). Includes import yaml, uses merge_configs to build platform_limits, references NodeUsage in the billing loop, and defines actor_yaml_text as a placeholder to avoid an undefined variable reference.
Individual subsections for each API group with focused code snippets.
create_executor heading updated to create_executor(config_dict, credentials, limits=None, pricing=None) to show that limits and pricing are optional.
Production warning added near create_executor: "Always pass limits and pricing in production. When omitted, no execution-limit enforcement is performed. The router MUST pass both on every request per ADR-2024."
ActorResult field list includes all 5 fields: response, prompt_tokens, completion_tokens, nodes, and state (opaque client-carried graph state for stateless resumption per ADR-2026). The state annotation updated to dict[str, Any] | None. The state= input parameter for execute() is now documented to complete the round-trip.
Streaming documentation for Executor.execute_stream() and last_result billing, including:
- Corrected last_result semantics: timeout removed from "Early abandonment" — timeout raises ExecutionError(kind="timeout") and populates last_result before re-raising.
- try/except ExecutionError block added to the streaming example showing last_result access in the except clause.
ExecutionError.kind/.reason reference table with recommended HTTP status codes (ADR-2029: budget_exhausted → 429, missing_pricing_entry → 500).
ADR citation corrected: the executor-per-request pattern is attributed to ADR-2024 (Library API Contract), not ADR-2026 (per-request credential injection).
"Package structure" table updated: cleveractors.templates row now lists only actually-exported names (BaseTemplate, GenericTemplate, TemplateType, TemplateParameter, TemplateRegistry, ComponentReference, InstantiationContext) — previously listed non-exported AgentTemplate, GraphTemplate, StreamTemplate.

"Package structure" and "Key exports" tables updated to reflect the v2.1.0 surface:

NodeType added to the cleveractors.langgraph row (previously omitted).
TemplateRegistry is now re-exported from cleveractors.templates.__init__ so the Key exports import block is runnable without ImportError.

`src/cleveractors/runtime.py` — `Executor.init` signature fix + DRY cleanup

limits and pricing parameters now have matching defaults limits: dict[str, Any] | None = None, pricing: dict[str, Any] | None = None with or {} substitution inside __init__, matching the create_executor factory signature. This prevents TypeError: missing 2 required positional arguments when users construct Executor directly (as documented in README's "Exports at a glance" table).
Validation updated: if limits is not None and not isinstance(limits, dict) (was not isinstance(limits, dict)) — None is now accepted and treated as {}.
Redundant double normalization removed: create_executor previously did limits or {} and pricing or {} before passing to Executor.__init__, which also did limits or {} and pricing or {}. The or {} in create_executor has been removed — all None-handling now lives exclusively in Executor.__init__ (single source of truth, DRY).
Explicit type annotations added: self.limits: dict[str, Any] and self.pricing: dict[str, Any] now have explicit class-level annotations per CONTRIBUTING.md mandate.
create_executor docstring updated: Args: entries for limits and pricing now include the ADR-2024 warning: "⚠️ Per ADR-2024, the router MUST pass both on every request. Omitting either disables timeout / depth / cost / call-count enforcement."
Module docstring updated to show create_executor(config_dict, credentials, limits=None, pricing=None) and note that omitting limits/pricing disables all enforcement.

`src/cleveractors/templates/init.py` — Export `TemplateRegistry` + sorted `all`

Added TemplateRegistry to the import block and __all__ so that from cleveractors.templates import TemplateRegistry works without raising ImportError. __all__ is now sorted alphabetically for consistency with the root-level cleveractors/__init__.py.

`features/environment.py` — Root logger level, handler level, and handler set test-isolation fix

Bug: ReactiveCleverAgentsApp.__init__ sets logging.getLogger().setLevel(log_level) globally based on the verbose parameter (verbose=0 → CRITICAL, verbose=1 → ERROR), and also sets each handler's level. Tests that create ReactiveCleverAgentsApp contaminated the logging state for later scenarios in the same slipcover process, causing 5 BDD scenarios in execute_stream.feature to fail in nox -s coverage_report (while passing in nox -s unit_tests due to process isolation differences).

Fix (extended): Extended the save/restore to snapshot and restore both the root logger level, all handler levels, AND the handler set itself:

before_scenario: saves context._root_log_original_level (root logger level), context._root_log_original_handler_levels (list of (handler, level) tuples for all handlers), and context._root_log_original_handlers (set of handlers at scenario start).
after_scenario: restores all three. Computes set(current_handlers) - set(original_handlers) and removes any handlers added during the scenario (e.g. StreamHandler added by ReactiveCleverAgentsApp.__init__). This makes the workaround fully hermetic against handler addition, not just level mutation. Added # Swallow intentionally: test isolation must not fail the test run. comments to the bare except Exception: pass blocks per CONTRIBUTING.md guidance.

Note: A follow-up issue (#49) has been filed to address the production-side root cause in ReactiveCleverAgentsApp.__init__ (move verbosity logic to a named child logger logging.getLogger("cleveractors") instead of mutating the root logger).

`features/templates_coverage.feature` + `features/steps/templates_steps.py` — New smoke test for `TemplateRegistry` re-export

Added BDD scenario verifying that from cleveractors.templates import TemplateRegistry succeeds and the imported class is the same object as cleveractors.templates.registry.TemplateRegistry. This follows the same pattern used by other top-level re-export smoke tests (e.g., execution_limits.feature:32–34).

Fixes applied from review:

Removed context: Any annotations from the three new step functions to match the existing local style in templates_steps.py (no annotation on context).
Narrowed except Exception to except ImportError in step_tc_import_template_registry per project convention.
Removed (tc) suffix from all three step names (none of them actually conflict with existing steps).

`features/runtime_coverage.feature` + `features/steps/runtime_coverage_steps.py` — New `Executor.init` None-handling tests

Added three BDD scenarios that construct Executor directly (not via create_executor) to exercise the None → {} code paths in Executor.__init__:

Executor(config_dict, credentials) — no limits/pricing kwargs at all → both default to {}
Executor(config_dict, credentials, limits=None) — explicit limits=None → limits defaults to {}
Executor(config_dict, credentials, pricing=None) — explicit pricing=None → pricing defaults to {}

These complement the existing create_executor with None limits and pricing scenario, which only exercises the factory path (where None was previously pre-converted before reaching __init__).

`CHANGELOG.md`

Fixed structural defect: ~22 new-feature entries that were misplaced under ### Fixed (after the second ### Added header was removed) have been moved back to ### Added. Entries like Executor.execute_stream(), create_executor(), ActorResult and NodeUsage types, merge_configs(), validate_dict(), Registry HTTP Client, Registry Client-Side Cache, Per-request credential injection, Structured ExecutionError fields, etc. are all new features, not bug fixes.
Merged duplicate ### Fixed sections: There were two ### Fixed sections; they have been merged into one canonical section.
Updated README entry to cite ADR-2026 (previously missing despite README documenting ActorResult.state attributed to ADR-2026).
Updated ### Fixed entry for the root-logger fix to mention handler set restoration (the fix now restores root logger level, handler levels, AND removes leaked handlers).

Quality Gate Results

Gate	Result
`nox -s lint`	✅ Pass
`nox -s typecheck`	✅ Pass (0 errors, 1 pre-existing warning for optional `langchain_google_genai`)
`nox -s unit_tests`	✅ 2600 scenarios, 0 failed
`nox -s coverage_report`	✅ 96.8% (threshold: 96.5%), 2600 scenarios, 0 failed
`nox -s integration_tests`	✅ 227 tests, 227 passed

Design References

ADR-2024 — cleveractors-core as Actor Runtime Library (defines the end-to-end router pattern; executor-per-request)
ADR-2026 — Per-Request Credential Injection (opaque ActorResult.state for stateless resumption)
ADR-2027 — ActorResult Type and Token Aggregation (billing contract)
ADR-2029 — Actor Execution Limits and Budget Enforcement (ExecutionError.kind/reason)

Known Limitations / Deferred Items

Issue #49: ReactiveCleverAgentsApp.__init__ still mutates the root logger level and handler levels globally. The test-side workaround in features/environment.py is now fully hermetic (saves/restores root level, handler levels, AND removes leaked handlers). The production-side fix (moving verbosity logic to a child logger) is tracked in issue #49.

Closes #17

## Summary This PR completes Wave C8 (issue #17): finalising the `cleveractors` v2.1.0 public API surface by documenting all router-facing exports in `README.md` and fixing a pre-existing test-isolation bug in the coverage pipeline. All code changes for C1–C7 (imports, `__all__`, version bump, smoke tests) were already merged to `master` in prior commits. This PR delivers the one remaining subtask: the "Router-facing API" README section, plus the bug fix. ## Changes ### `README.md` — New "Router-facing API" section (AC7) Added a comprehensive "Router-facing API" section after the Quick start block, including: - **Summary table** listing all 8 router-facing exports (`validate_dict`, `merge_configs`, `create_executor`, `Executor`, `ActorResult`, `NodeUsage`, `ExecutionError`, `ConfigurationError`) with one-line descriptions. `merge_configs` correctly described as accepting "zero or more" dicts. `create_executor` row updated to "…per-request credentials, and optional execution limits and pricing table" (reflecting the optional parameters). - **End-to-end usage example** following the ADR-2024 "Library API Contract" pattern (validate at upload time, execute at request time, bill from `ActorResult.nodes`). Includes `import yaml`, uses `merge_configs` to build `platform_limits`, references `NodeUsage` in the billing loop, and defines `actor_yaml_text` as a placeholder to avoid an undefined variable reference. - **Individual subsections** for each API group with focused code snippets. - **`create_executor` heading** updated to `create_executor(config_dict, credentials, limits=None, pricing=None)` to show that `limits` and `pricing` are optional. - **Production warning** added near `create_executor`: "Always pass `limits` and `pricing` in production. When omitted, no execution-limit enforcement is performed. The router MUST pass both on every request per ADR-2024." - **`ActorResult` field list** includes all 5 fields: `response`, `prompt_tokens`, `completion_tokens`, `nodes`, and `state` (opaque client-carried graph state for stateless resumption per ADR-2026). The `state` annotation updated to `dict[str, Any] | None`. The `state=` input parameter for `execute()` is now documented to complete the round-trip. - **Streaming documentation** for `Executor.execute_stream()` and `last_result` billing, including: - Corrected `last_result` semantics: `timeout` removed from "Early abandonment" — timeout raises `ExecutionError(kind="timeout")` and populates `last_result` before re-raising. - `try/except ExecutionError` block added to the streaming example showing `last_result` access in the except clause. - **`ExecutionError.kind`/`.reason` reference table** with recommended HTTP status codes (ADR-2029: `budget_exhausted` → 429, `missing_pricing_entry` → 500). - **ADR citation corrected**: the executor-per-request pattern is attributed to ADR-2024 (Library API Contract), not ADR-2026 (per-request credential injection). - **"Package structure" table** updated: `cleveractors.templates` row now lists only actually-exported names (`BaseTemplate`, `GenericTemplate`, `TemplateType`, `TemplateParameter`, `TemplateRegistry`, `ComponentReference`, `InstantiationContext`) — previously listed non-exported `AgentTemplate`, `GraphTemplate`, `StreamTemplate`. "Package structure" and "Key exports" tables updated to reflect the v2.1.0 surface: - `NodeType` added to the `cleveractors.langgraph` row (previously omitted). - `TemplateRegistry` is now re-exported from `cleveractors.templates.__init__` so the Key exports import block is runnable without `ImportError`. ### `src/cleveractors/runtime.py` — `Executor.__init__` signature fix + DRY cleanup - `limits` and `pricing` parameters now have matching defaults `limits: dict[str, Any] | None = None, pricing: dict[str, Any] | None = None` with `or {}` substitution inside `__init__`, matching the `create_executor` factory signature. This prevents `TypeError: missing 2 required positional arguments` when users construct `Executor` directly (as documented in README's "Exports at a glance" table). - Validation updated: `if limits is not None and not isinstance(limits, dict)` (was `not isinstance(limits, dict)`) — `None` is now accepted and treated as `{}`. - **Redundant double normalization removed**: `create_executor` previously did `limits or {}` and `pricing or {}` before passing to `Executor.__init__`, which also did `limits or {}` and `pricing or {}`. The `or {}` in `create_executor` has been removed — all `None`-handling now lives exclusively in `Executor.__init__` (single source of truth, DRY). - **Explicit type annotations added**: `self.limits: dict[str, Any]` and `self.pricing: dict[str, Any]` now have explicit class-level annotations per CONTRIBUTING.md mandate. - **`create_executor` docstring updated**: `Args:` entries for `limits` and `pricing` now include the ADR-2024 warning: "⚠️ Per ADR-2024, the router MUST pass both on every request. Omitting either disables timeout / depth / cost / call-count enforcement." - Module docstring updated to show `create_executor(config_dict, credentials, limits=None, pricing=None)` and note that omitting `limits`/`pricing` disables all enforcement. ### `src/cleveractors/templates/__init__.py` — Export `TemplateRegistry` + sorted `__all__` Added `TemplateRegistry` to the import block and `__all__` so that `from cleveractors.templates import TemplateRegistry` works without raising `ImportError`. `__all__` is now sorted alphabetically for consistency with the root-level `cleveractors/__init__.py`. ### `features/environment.py` — Root logger level, handler level, and handler set test-isolation fix **Bug:** `ReactiveCleverAgentsApp.__init__` sets `logging.getLogger().setLevel(log_level)` globally based on the `verbose` parameter (`verbose=0` → CRITICAL, `verbose=1` → ERROR), and also sets each handler's level. Tests that create `ReactiveCleverAgentsApp` contaminated the logging state for later scenarios in the same slipcover process, causing 5 BDD scenarios in `execute_stream.feature` to fail in `nox -s coverage_report` (while passing in `nox -s unit_tests` due to process isolation differences). **Fix (extended):** Extended the save/restore to snapshot and restore both the root logger level, all handler levels, AND the handler set itself: - `before_scenario`: saves `context._root_log_original_level` (root logger level), `context._root_log_original_handler_levels` (list of `(handler, level)` tuples for all handlers), and `context._root_log_original_handlers` (set of handlers at scenario start). - `after_scenario`: restores all three. Computes `set(current_handlers) - set(original_handlers)` and removes any handlers added during the scenario (e.g. `StreamHandler` added by `ReactiveCleverAgentsApp.__init__`). This makes the workaround fully hermetic against handler addition, not just level mutation. Added `# Swallow intentionally: test isolation must not fail the test run.` comments to the bare `except Exception: pass` blocks per CONTRIBUTING.md guidance. **Note:** A follow-up issue (#49) has been filed to address the production-side root cause in `ReactiveCleverAgentsApp.__init__` (move verbosity logic to a named child logger `logging.getLogger("cleveractors")` instead of mutating the root logger). ### `features/templates_coverage.feature` + `features/steps/templates_steps.py` — New smoke test for `TemplateRegistry` re-export Added BDD scenario verifying that `from cleveractors.templates import TemplateRegistry` succeeds and the imported class is the same object as `cleveractors.templates.registry.TemplateRegistry`. This follows the same pattern used by other top-level re-export smoke tests (e.g., `execution_limits.feature:32–34`). **Fixes applied from review:** - Removed `context: Any` annotations from the three new step functions to match the existing local style in `templates_steps.py` (no annotation on `context`). - Narrowed `except Exception` to `except ImportError` in `step_tc_import_template_registry` per project convention. - Removed `(tc)` suffix from all three step names (none of them actually conflict with existing steps). ### `features/runtime_coverage.feature` + `features/steps/runtime_coverage_steps.py` — New `Executor.__init__` None-handling tests Added three BDD scenarios that construct `Executor` directly (not via `create_executor`) to exercise the `None → {}` code paths in `Executor.__init__`: 1. `Executor(config_dict, credentials)` — no `limits`/`pricing` kwargs at all → both default to `{}` 2. `Executor(config_dict, credentials, limits=None)` — explicit `limits=None` → `limits` defaults to `{}` 3. `Executor(config_dict, credentials, pricing=None)` — explicit `pricing=None` → `pricing` defaults to `{}` These complement the existing `create_executor with None limits and pricing` scenario, which only exercises the factory path (where `None` was previously pre-converted before reaching `__init__`). ### `CHANGELOG.md` - **Fixed structural defect**: ~22 new-feature entries that were misplaced under `### Fixed` (after the second `### Added` header was removed) have been moved back to `### Added`. Entries like `Executor.execute_stream()`, `create_executor()`, `ActorResult and NodeUsage types`, `merge_configs()`, `validate_dict()`, `Registry HTTP Client`, `Registry Client-Side Cache`, `Per-request credential injection`, `Structured ExecutionError fields`, etc. are all new features, not bug fixes. - **Merged duplicate `### Fixed` sections**: There were two `### Fixed` sections; they have been merged into one canonical section. - Updated README entry to cite ADR-2026 (previously missing despite README documenting `ActorResult.state` attributed to ADR-2026). - Updated `### Fixed` entry for the root-logger fix to mention handler set restoration (the fix now restores root logger level, handler levels, AND removes leaked handlers). ## Quality Gate Results | Gate | Result | |------|--------| | `nox -s lint` | ✅ Pass | | `nox -s typecheck` | ✅ Pass (0 errors, 1 pre-existing warning for optional `langchain_google_genai`) | | `nox -s unit_tests` | ✅ 2600 scenarios, 0 failed | | `nox -s coverage_report` | ✅ 96.8% (threshold: 96.5%), 2600 scenarios, 0 failed | | `nox -s integration_tests` | ✅ 227 tests, 227 passed | ## Design References - **ADR-2024** — cleveractors-core as Actor Runtime Library (defines the end-to-end router pattern; executor-per-request) - **ADR-2026** — Per-Request Credential Injection (opaque `ActorResult.state` for stateless resumption) - **ADR-2027** — ActorResult Type and Token Aggregation (billing contract) - **ADR-2029** — Actor Execution Limits and Budget Enforcement (ExecutionError.kind/reason) ## Known Limitations / Deferred Items - **Issue #49**: `ReactiveCleverAgentsApp.__init__` still mutates the root logger level and handler levels globally. The test-side workaround in `features/environment.py` is now fully hermetic (saves/restores root level, handler levels, AND removes leaked handlers). The production-side fix (moving verbosity logic to a child logger) is tracked in issue #49. Closes #17

hurui200320 added 1 commit 2026-06-12 14:26:14 +00:00

feat(public-api): expose all router-facing APIs at cleveractors package level; update README

CI / lint (pull_request) Successful in 1m21s

Details

CI / typecheck (pull_request) Successful in 1m21s

Details

CI / security (pull_request) Successful in 1m20s

Details

CI / quality (pull_request) Successful in 56s

Details

CI / build (pull_request) Successful in 54s

Details

CI / integration_tests (pull_request) Successful in 1m23s

Details

CI / unit_tests (pull_request) Successful in 3m27s

Details

CI / coverage (pull_request) Successful in 3m7s

Details

CI / status-check (pull_request) Successful in 3s

Details

14b59ca201

Added 'Router-facing API' section to README.md documenting all 7 router-facing
exports introduced in waves C1–C7: validate_dict, merge_configs, create_executor,
Executor, ActorResult, NodeUsage, ExecutionError, and ConfigurationError.

The new section includes:
- A summary table listing each export with a one-line description.
- A complete end-to-end usage example following the ADR-2024 API contract
  (validate_dict → merge_configs → create_executor → execute → bill from nodes).
- Individual subsections for each API with focused code snippets.
- Streaming documentation for Executor.execute_stream() and last_result.
- A reference table for ExecutionError.kind and .reason values with recommended
  HTTP status codes per ADR-2029.

'Package structure' table updated to list all new v2.1.0 exports. 'Key exports'
section reorganised to show the router-facing surface first, then the legacy
CLI-facing surface, then sub-module imports for advanced use.

Also fixed a pre-existing test-isolation bug in features/environment.py:
ReactiveCleverAgentsApp.__init__ sets the root logger level globally (verbose=0
→ CRITICAL, verbose=1 → ERROR, etc.). Under slipcover, this contaminated
subsequent scenarios that relied on WARNING-level log records propagating to the
root logger (execute_stream.feature:192–225). The fix saves and restores the root
logger level across each scenario via before_scenario/after_scenario hooks.
All 2556 BDD scenarios and 199 Robot integration tests pass; coverage 96.9%.

ISSUES CLOSED: #17

hurui200320 added this to the v2.1.0 milestone 2026-06-12 14:26:22 +00:00

hurui200320 added the

Type

Task

label 2026-06-12 14:26:22 +00:00

hurui200320 referenced this pull request

2026-06-12 14:29:43 +00:00

feat(public-api): expose all router-facing APIs at cleveractors package level; update README #17

hurui200320 added a new dependency 2026-06-12 16:40:10 +00:00

#17 feat(public-api): expose all router-facing APIs at cleveractors package level; update README

hurui200320 force-pushed feature/17-public-api from 14b59ca201 to 203a433120

2026-06-12 17:14:58 +00:00

Compare

hurui200320 referenced this pull request

2026-06-12 17:38:29 +00:00

refactor(application): move ReactiveCleverAgentsApp verbosity logging to child logger to stop mutating root logger #49

hurui200320 force-pushed feature/17-public-api from 203a433120 to 35ecb8d875

2026-06-12 17:45:08 +00:00

Compare

hurui200320 force-pushed feature/17-public-api from 35ecb8d875 to e0625f3cdb

2026-06-12 17:51:23 +00:00

Compare

hurui200320 force-pushed feature/17-public-api from e0625f3cdb to 5776daed1b

2026-06-12 19:02:21 +00:00

Compare

hurui200320 force-pushed feature/17-public-api from 5776daed1b to 88ca16e238

2026-06-12 19:57:22 +00:00

Compare

CI / quality (pull_request) Successful in 34s

Details

CI / lint (pull_request) Failing after 38s

Details

CI / unit_tests (pull_request) Failing after 39s

Details

CI / build (pull_request) Successful in 40s

Details

CI / typecheck (pull_request) Successful in 1m6s

Details

CI / security (pull_request) Successful in 1m4s

Details

CI / coverage (pull_request) Has been skipped

Details

CI / integration_tests (pull_request) Successful in 1m5s

Details

CI / status-check (pull_request) Failing after 3s

Details

This pull request has changes conflicting with the target branch.

CHANGELOG.md
src/cleveractors/runtime.py

View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u origin feature/17-public-api:feature/17-public-api

git checkout feature/17-public-api

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Blocks

#17 feat(public-api): expose all router-facing APIs at cleveractors package level; update README

cleveragents/cleveractors-core

Reference: cleveragents/cleveractors-core#48

feat(public-api): expose all router-facing APIs at cleveractors package level; update README #48

Summary

Changes

README.md — New "Router-facing API" section (AC7)

src/cleveractors/runtime.py — Executor.__init__ signature fix + DRY cleanup

src/cleveractors/templates/__init__.py — Export TemplateRegistry + sorted __all__

features/environment.py — Root logger level, handler level, and handler set test-isolation fix

features/templates_coverage.feature + features/steps/templates_steps.py — New smoke test for TemplateRegistry re-export

features/runtime_coverage.feature + features/steps/runtime_coverage_steps.py — New Executor.__init__ None-handling tests

CHANGELOG.md

Quality Gate Results

Design References

Known Limitations / Deferred Items

Checkout

`README.md` — New "Router-facing API" section (AC7)

`src/cleveractors/runtime.py` — `Executor.init` signature fix + DRY cleanup

`src/cleveractors/templates/init.py` — Export `TemplateRegistry` + sorted `all`

`features/environment.py` — Root logger level, handler level, and handler set test-isolation fix

`features/templates_coverage.feature` + `features/steps/templates_steps.py` — New smoke test for `TemplateRegistry` re-export

`features/runtime_coverage.feature` + `features/steps/runtime_coverage_steps.py` — New `Executor.init` None-handling tests

`CHANGELOG.md`