Tests, licence, and the tooling to keep the repo consistent

A test suite with no dependencies beyond bash and the Python standard
library, run by one command:

  scripts/test.sh                    (or: make test)
  scripts/test.sh --windows [host]   adds the PowerShell suite over SSH

- tests/static.sh: script syntax, JSON validity, VERSION against the widget
  manifest and CHANGELOG, providers.json against the kinds, probes and checks
  the code implements (and ids that would shadow a command), qmllint and
  shellcheck when installed.
- tests/python: unit tests for cm-json.py - providers and the column contract
  bash depends on, the original three scaffolds byte for byte, every
  catalogue parser and the cache, session scan/dismiss/repair, rename,
  defaults, health, the cost guard.
- tests/cli: the CLI in a sandbox with its own HOME, CM_ROOT,
  CLAUDE_CONFIG_DIR and a file-only vault, against fake Ollama, LM Studio,
  keyed-gateway and proxy servers - preflight, presets, models, doctor, a real
  switch, sessions, and the key helper under a rename race.
- tests/windows: the same idea on a Windows host, with USERPROFILE pointed at
  a temp folder so the real install is never touched.

Also: MIT licence, CHANGELOG.md reconstructed from history, .editorconfig and
.gitattributes pinning LF everywhere (what the tree already is), and
scripts/bump-version.sh, since the version lives in two files and drifted
once before.

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

tests/cli/test_models.sh

@@ -0,0 +1,46 @@
+#!/usr/bin/env bash
+# Model catalogues: fetched by each provider's kind, cached per server.
+. "$(dirname "$0")/../lib.sh"
+start_fake
+C="$CM_ROOT/models-cache.json"
+
+ready_preset ollama "$OLLAMA_URL"
+run_cm models --preset ollama
+expect_out 'qwen3-coder:latest' 'ollama models are listed'
+run_cm models --preset ollama --refresh
+expect_out 'cached 2 ollama model(s)' 'and cached'
+expect_eq "$(python3 -c 'import json,sys; print(json.load(open(sys.argv[1]))["providers"]["ollama"]["baseUrl"])' "$C")" \
+    "$OLLAMA_URL" 'against the server they came from'
+
+ready_preset lmstudio "$LMSTUDIO_URL"
+run_cm models --preset lmstudio
+expect_out 'qwen3-coder-30b' 'lm studio models are listed'
+
+python3 - "$P/custom.json" "$KEYED_URL" <<'PY'
+import json, sys
+p = json.load(open(sys.argv[1]))
+p.update(baseUrl=sys.argv[2], auth={"mode": "literal", "token": "sk-real"}, configured=True)
+json.dump(p, open(sys.argv[1], "w"), indent=2)
+PY
+run_cm models --preset custom
+expect_out 'gw/model-b' 'a keyed endpoint lists its models with the key'
+
+cp "$P/custom.json" "$P/proxy.json"
+python3 "$J" set-url "$P/proxy.json" "$PROXY_URL" >/dev/null
+run_cm models --preset proxy --refresh
+expect_rc 1 'a proxy has no list to fetch'
+expect_out 'model ids can still be typed by hand' 'and that is explained'
+
+run_cm models --preset zai
+expect_out 'glm-5.3' 'the static zai list'
+run_cm models --preset zai --json
+expect_out '"glm-4.7"' '--json prints the cached node'
+
+run_cm models --preset nope --refresh
+expect_rc 1 'an unknown preset is refused'
+
+for word in keyRef token auth; do
+    case "$(cat "$C")" in *"\"$word\""*) fail "the cache holds a $word" ;; *) pass ;; esac
+done
+
+finish