Add test coverage for 4 new chapter pipeline steps

[Copilot]

Four production pipeline steps (voice & dialogue differentiation, human oddities, metaphor reduction, copy edit) execute on every chapter generation with no dedicated tests — regressions in template variables or prompt keys would be silent.

Changes

tests/conftest.py

• Added 4 matched cases to _canned_llm_response() so these actions no longer fall through to the generic fallback:
  • "voice" in act and "dialogue" in act
  • "human oddities" in act
  • "metaphor" in act
  • "copy edit" in act

tests/test_new_pipeline_prompts.py (new)

• Unit tests for each prompt builder verifying message structure, required variables are injected into rendered output, and determinism
• Pipeline integration tests verifying all 4 steps fire inside _run_all_chapter_agents by patching a recording wrapper onto novelforge.agents.chapter._helpers.call_llm (the actual call site) and asserting each action string appears:

def _run_and_collect_actions(self, mock_llm, mocker) -> list[str]:
    actions: list[str] = []

    def _recording(messages, *, action: str = "", json_mode: bool = False) -> str:
        actions.append(action)
        return _canned_llm_response(messages, action=action, json_mode=json_mode)

    mocker.patch("novelforge.agents.chapter._helpers.call_llm", side_effect=_recording)
    _run_all_chapter_agents(**self._PIPELINE_KWARGS)
    return actions

• _action_present() helper de-duplicates the substring-matching assertions across the 4 per-step tests and the combined test_all_four_steps_execute_in_single_run test

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• api.openai.com
  • Triggering command: /usr/bin/python3 python3 -m pytest tests/ -x -q --tb=short (dns block)
  • Triggering command: /usr/bin/python3 python3 -m pytest tests/ -q --cov=novelforge --cov-report=term-missing --cov-fail-under=85 (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)

[Copilot]

Pull request overview

Adds dedicated unit + integration tests to ensure four always-on chapter-generation refinement steps (voice/dialogue differentiation, human oddities, metaphor reduction, copy edit) remain wired correctly and don’t silently regress due to prompt/template variable changes.

Changes:

• Extend the shared mock_llm canned response router to return step-specific outputs for the four new pipeline actions.
• Add a new test module that:
  • unit-tests each prompt builder for structure, variable injection, and determinism
  • integration-tests that all four steps execute within _run_all_chapter_agents by recording action strings at the actual LLM call site.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
tests/conftest.py	Adds canned LLM responses for the four new refinement actions so tests don’t fall back to the generic response.
tests/test_new_pipeline_prompts.py	New unit + integration tests covering prompt builders and verifying the four pipeline steps execute during a full chapter pipeline run.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CyberSecDef]

This is ready to merge

[CyberSecDef]

@copilot
please review these changes and provide a merging approval response

[Copilot]

@copilot
please review these changes and provide a merging approval response

Reviewed and approved. ✅

Summary of changes (commit 378e5e2):

• tests/conftest.py — _canned_llm_response() now has specific match arms for all four new action strings (voice & dialogue differentiation, human oddities, metaphor reduction, copy edit), so they no longer silently fall through to the generic fallback.
• tests/test_new_pipeline_prompts.py (new, 25 tests) — unit tests for each prompt builder (structure, variable injection, determinism) plus pipeline integration tests that record action strings at the actual call site (novelforge.agents.chapter._helpers.call_llm) to assert each step executes inside _run_all_chapter_agents.

CI results:

• 61 targeted tests pass (0 failures)
• Full suite passes with 85.86% coverage (threshold 85%)
• CodeQL security scan: 0 alerts

Safe to merge.

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• api.openai.com
  • Triggering command: /usr/bin/python3 python3 -m pytest tests/ -q --cov=novelforge --cov-fail-under=85 (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)