All 1652 Tests Pass

All 1652 Tests Pass | Generated by AI

Home 2026.07

All 1652 tests pass with 3 skipped (same as before - the skipped ones are for optional dependencies). Zero errors, zero failures.

Summary of Fixes

Here’s what was wrong and what I fixed:

1. `ww/env/init.py` — `load_env()` only loaded from `~/.config/ww/.env`

Problem: The load_env() function only loaded from ~/.config/ww/.env, but the AGENTS.md docs and tests expected it to also load from BASE_PATH, CWD, and project root.

Fix: Updated load_env() to load .env from 4 locations in order (later overrides earlier):

Current working directory
Project root (parent of ww/)
BASE_PATH env var (if set)
~/.config/ww/.env

2. `tests/social/test_x_bulk_unfollow.py` — 11 errors + 1 failure

The source ww/social/x_bulk_unfollow.py was significantly refactored (batch-based LLM picking, retry loop in launch_chrome) but the tests weren’t updated.

Fixes:

test_launch_chrome: Fixed mock target from ww.social.x_bulk_unfollow.socket.create_connection (fails since socket is imported inside the function) to socket.create_connection. Updated sleep assertion since the function now has a connection retry loop that exits immediately when mocked.
TestAskLlmShouldUnfollow → TestAskLlmPickUnfollow: Rewrote all 5 tests to match the new ask_llm_pick_unfollow(profiles) function that returns (index, reason) instead of the old ask_llm_should_unfollow(profile) that returned (decision, reason).
TestUnfollowWithLlm: Completely rewrote all 5 tests. The old tests used dry_run (removed from API), mocked ask_llm_should_unfollow (renamed), and had complex page mock logic that didn’t match the new code’s control flow. New tests mock collect_batch and do_unfollow at a high level with proper side_effect functions to handle the double-call-per-iteration pattern.
_make_page_with_cells helper: Simplified to return consistent cell counts (the old version decremented count after first call).

Back Donate