test(backend): assert process_audio threads the game-system context into summarise; share same_task_session (#457) #646

Merged

claude-bot merged 2 commits from test/457-process-audio-harness into main 2026-09-14 04:51:09 +00:00

Conversation 0 Commits 2 Files changed 17 +380 -274

claude-bot commented 2026-09-14 04:12:18 +00:00

Collaborator

Copy link

Test-only. Nothing under webapp/backend/app/ changes.

1. One same_task_session fixture

Celery task bodies open their own session with task_session(). The db fixture never really commits (commit is aliased to flush, then rolled back), so without a patch a task can't see any row the test set up. It returns early and the test passes without reaching the code under test. That's how #400 and #401 went wrong.

The patch was hand-copied into 14 files:

• 10 module-level same_task_session fixtures
• 7 per-test closures in test_reminder_tasks.py / test_previously_on_recap.py
• 2 more under a different name (_yield_db) in test_availability.py and test_summary_recompose.py, which a search for the fixture name missed

All were behaviourally identical: each patches only app.database.task_session with a bare yield db. They're now one fixture in conftest.py, with a docstring explaining the silent-pass failure it prevents.

Refactor mutation check: with the shared fixture made a no-op, 14 tests across 5 converted files went red; they're green again with it restored. The at-risk tests in test_availability.py guard this by construction: with no patch, the poller finds no session and call_count == 1 fails.

2. summarise gets the resolved game-system context through process_audio

test_process_audio_system_context.py drives process_audio.run(...) end to end, with only the provider calls stubbed (the seam from test_hallucination_filter.py). It intercepts audio_service.summarise and asserts on the argument it was actually called with.

• Free-form campaign: receives "Homebrew Steampunk System" unchanged.
• Linked registry campaign: receives the dnd5e prompt_hint, and not the raw column.

Both cases run the real resolve_system_prompt_context against an in-memory Campaign (and, for the linked case, GameSystem); neither branch needs a database. Together they tell resolving the context apart from simply passing campaign.game_system through.

Mutation check: passing None at the summarise(...) call site in reminder_tasks.py fails both tests (assert None == 'Homebrew Steampunk System', assert None == 'D&D 5th Edition characters are built from…'); restored, git diff webapp/backend/app is empty.

3. The gap note is gone

The "summarise() call site" note in test_system_aware_prompts.py is removed, and the module docstring now points to the new test.

Verification

• Full backend suite: 3476 passed, 13 skipped.
• The two follow-up files rerun: 83 passed.
• ruff format --check / ruff check clean under CI's pinned 0.4.4.

Against #457's acceptance criteria

• A shared fixture patches task_session; the hand-copied duplicates are removed.
• A test drives process_audio end to end with external seams stubbed. The seam fixture in test_failed_windows.py already did this; the new module adds a case aimed at summarise.
• summarise receives the resolved context, for linked and free-form campaigns.
• The note in test_system_aware_prompts.py is removed.

Item 4 (sequencing guards for #421/#425/#431/#432/#397) was optional and isn't in this PR.

Closes #457

🤖 Generated with Claude Code

Test-only. Nothing under `webapp/backend/app/` changes. ## 1. One `same_task_session` fixture Celery task bodies open their own session with `task_session()`. The `db` fixture never really commits (`commit` is aliased to `flush`, then rolled back), so without a patch a task can't see any row the test set up. It returns early and the test passes without reaching the code under test. That's how #400 and #401 went wrong. The patch was hand-copied into **14 files**: - 10 module-level `same_task_session` fixtures - 7 per-test closures in `test_reminder_tasks.py` / `test_previously_on_recap.py` - 2 more under a different name (`_yield_db`) in `test_availability.py` and `test_summary_recompose.py`, which a search for the fixture name missed All were behaviourally identical: each patches only `app.database.task_session` with a bare `yield db`. They're now one fixture in `conftest.py`, with a docstring explaining the silent-pass failure it prevents. **Refactor mutation check:** with the shared fixture made a no-op, 14 tests across 5 converted files went red; they're green again with it restored. The at-risk tests in `test_availability.py` guard this by construction: with no patch, the poller finds no session and `call_count == 1` fails. ## 2. `summarise` gets the resolved game-system context through `process_audio` `test_process_audio_system_context.py` drives `process_audio.run(...)` end to end, with only the provider calls stubbed (the seam from `test_hallucination_filter.py`). It intercepts `audio_service.summarise` and asserts on the argument it was actually called with. - **Free-form campaign:** receives `"Homebrew Steampunk System"` unchanged. - **Linked registry campaign:** receives the dnd5e `prompt_hint`, and not the raw column. Both cases run the **real** `resolve_system_prompt_context` against an in-memory `Campaign` (and, for the linked case, `GameSystem`); neither branch needs a database. Together they tell resolving the context apart from simply passing `campaign.game_system` through. **Mutation check:** passing `None` at the `summarise(...)` call site in `reminder_tasks.py` fails both tests (`assert None == 'Homebrew Steampunk System'`, `assert None == 'D&D 5th Edition characters are built from…'`); restored, `git diff webapp/backend/app` is empty. ## 3. The gap note is gone The "summarise() call site" note in `test_system_aware_prompts.py` is removed, and the module docstring now points to the new test. ## Verification - Full backend suite: **3476 passed, 13 skipped**. - The two follow-up files rerun: 83 passed. - `ruff format --check` / `ruff check` clean under CI's pinned 0.4.4. ## Against #457's acceptance criteria - [x] A shared fixture patches `task_session`; the hand-copied duplicates are removed. - [x] A test drives `process_audio` end to end with external seams stubbed. The `seam` fixture in `test_failed_windows.py` already did this; the new module adds a case aimed at `summarise`. - [x] `summarise` receives the resolved context, for linked and free-form campaigns. - [x] The note in `test_system_aware_prompts.py` is removed. Item 4 (sequencing guards for #421/#425/#431/#432/#397) was optional and isn't in this PR. Closes #457 🤖 Generated with [Claude Code](https://claude.com/claude-code)

claude-bot added 2 commits 2026-09-14 04:12:19 +00:00

test(backend): assert process_audio threads the game-system context into summarise; share same_task_session (#457) ... 8dbf8fa5b8

Adds test_process_audio_system_context.py, which drives process_audio.run
through a provider-stubbed seam (copied from test_hallucination_filter.py)
with audio_service.summarise intercepted instead of stubbed, and asserts the
game-system argument it receives for both a free-form campaign and one linked
to the registry (D&D 5e's prompt_hint). Both cases let the real
resolve_system_prompt_context run against an in-memory Campaign — no database
needed, since the free-form branch reads campaign.game_system directly and
the linked branch reads campaign.game_system_ref once it is set, satisfying
sqlalchemy.inspect's loaded check without a session. Verified with a
mutation check against reminder_tasks.py (restored after, no diff under
webapp/backend/app). Removes the corresponding gap note in
test_system_aware_prompts.py.

Also consolidates the ten near-identical module-level `same_task_session`
fixtures (all patched app.database.task_session identically) into one shared
fixture in conftest.py, and converts the equivalent inline closures in
test_reminder_tasks.py and test_previously_on_recap.py to use it.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

test(backend): fold the last two task_session copies into the shared fixture (#457) ... 7951cce5c2

test_availability.py and test_summary_recompose.py carried the same
patch under a different local name (_yield_db), so a search for
same_task_session missed them. Same behaviour, now the shared fixture.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

claude-bot merged commit f95c8d4e08 into main 2026-09-14 04:51:09 +00:00

claude-bot deleted branch test/457-process-audio-harness 2026-09-14 04:51:09 +00:00

claude-bot referenced this pull request from a commit 2026-09-14 04:51:10 +00:00

Merge PR #646: test(backend): assert process_audio threads the game-system context into summarise; share same_task_session (#457)

claude-bot referenced this pull request 2026-09-14 04:51:47 +00:00

[Tests] Nothing exercises process_audio's body — the pipeline's own orchestration is untested #457

claude-bot referenced this pull request 2026-09-14 05:09:59 +00:00

[Backend+Frontend] Campaign invites with expiry, usage limits and per-campaign join approval; fix the INVITE_CODE sign-up gate #643

claude-bot referenced this pull request 2026-09-14 05:55:39 +00:00

feat(webapp): campaign ownership — owner may be a player, owner-set GMs, transfer with acceptance (#644) #648

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

backend

Issue related to the backend

  

bot

Issue related to the bot

  

bug

Something is not working

  

duplicate

This issue or pull request already exists

  

enhancement

New feature

  

frontend

Issue related to the frontend

  

help wanted

Need some help

  

invalid

Something is wrong

  

question

More information is needed

  

recording

Issue related to the recording system

  

security

Security hardening / vulnerability work

  

wontfix

This won't be fixed

No labels

backend

bot

bug

duplicate

enhancement

frontend

help wanted

invalid

question

recording

security

wontfix

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

rbrooks/Quest-Board!646

No description provided.