revert(providers): remove mandatory LiteLLM/SSO enforcement gate from setup

.ai-run/guides/development/development-practices.md

@@ -337,9 +337,8 @@ codemie doctor
 | Dev watch | `npm run dev` | Watch mode (tsc --watch) |
 | Lint | `npm run lint` | ESLint check (zero warnings) |
 | Lint fix | `npm run lint:fix` | Auto-fix issues |
-| Test | `npm test` | ONLY if user requests |
-| Test unit | `npm run test:unit` | Unit tests only |
-| Test integration | `npm run test:integration` | Integration tests only |
+| Test | `npm test` | Full local suite (unit + cli + agent), ONLY if user requests |
+| Test one project | `npx vitest run --project unit\|cli\|agent` | Scoped run while iterating |
 | CI | `npm run ci` | Full CI pipeline |
 | Link global | `npm link` | Link for local testing |
 

.ai-run/guides/quality-gates.md

@@ -36,21 +36,21 @@ Run order is fastest-to-slowest. Each gate is a real `npm run` script in `packag
 
 ### Unit tests
 
-**Run**: `npm run test:unit` (`vitest run src`)
+**Run**: `npx vitest run --project unit` (part of `npm test`)
 **Pass**: all tests under `src/**/__tests__/` and `src/**/*.test.ts` pass.
 **Fail**: Vitest prints failing specs with stack traces.
 **Auto-fix**: none.
 **Skip if**: never, unless the change is `.ai-run/guides/` or doc-only.
 
 ### Cross-platform CI (Windows)
 
-CI runs a separate `test-windows` job (`.github/workflows/ci.yml`) using the same `npm run test:unit`/`test:integration` commands on `windows-latest`. GitHub's Windows runners default `core.autocrlf=true`, so any text file is checked out with CRLF unless `.gitattributes` forces LF. The repo's `.gitattributes` (`* text=auto eol=lf`) exists specifically to prevent this — without it, a `.mjs`/`.js` file starting with a shebang line (`#!/usr/bin/env node`) breaks Vite/Vitest's module transform with `SyntaxError: Invalid or unexpected token` when checked out with CRLF. See `src/agents/plugins/claude/plugin/statusline.mjs:1`.
+CI runs a separate `test-windows` job (`.github/workflows/ci.yml`) using the same `npm run ci` test commands (`vitest run --project unit` / `--project cli`) on `windows-latest`. GitHub's Windows runners default `core.autocrlf=true`, so any text file is checked out with CRLF unless `.gitattributes` forces LF. The repo's `.gitattributes` (`* text=auto eol=lf`) exists specifically to prevent this — without it, a `.mjs`/`.js` file starting with a shebang line (`#!/usr/bin/env node`) breaks Vite/Vitest's module transform with `SyntaxError: Invalid or unexpected token` when checked out with CRLF. See `src/agents/plugins/claude/plugin/statusline.mjs:1`.
 
 **Local repro**: convert a file to CRLF (`perl -pi -e 's/\n/\r\n/ unless /\r\n$/' <file>`) and re-run `npx vitest run <its-test>` — this reproduces Windows-only CI failures without needing a Windows machine.
 
 ### Integration tests
 
-**Run**: `npm run test:integration` (`vitest run tests/integration`)
+**Run**: `npx vitest run --project cli` (part of `npm test`; `tests/integration/**` minus `agent-*.test.ts`)
 **Pass**: all specs under `tests/integration/` pass.
 **Fail**: Vitest output identifies the failing scenario; check `tests/integration/session/fixtures/` for snapshot drift.
 **Auto-fix**: none.
@@ -80,7 +80,7 @@ CI runs a separate `test-windows` job (`.github/workflows/ci.yml`) using the sam
 
 ### Full CI
 
-**Run**: `npm run ci` (`license-check && lint && build && test:unit && test:integration`)
+**Run**: `npm run ci` (`license-check && lint && build && vitest run --project unit && vitest run --project cli`)
 **Pass**: every above-listed gate passes in order.
 **Fail**: stops at the first failing gate.
 **Skip if**: never before merge.

.ai-run/guides/testing/testing-patterns.md

@@ -192,7 +192,7 @@ Reference: `tests/integration/*.test.ts`
 
 ## Test Commands
 
-See `.ai-run/guides/quality-gates.md` for full command definitions (`npm test`, `test:unit`, `test:integration`, `test:coverage`, `test:watch`).
+See `.ai-run/guides/quality-gates.md` for full command definitions (`npm test` runs the full local suite; `npx vitest run --project unit|cli|agent` for a scoped run; `test:coverage`, `test:watch`).
 
 Run a specific file:
 ```bash

.claude/agents/qa-lead.md

@@ -71,9 +71,9 @@ npm run validate:secrets
 The repository policy says tests are run only on explicit user request. If the user explicitly requested tests or coverage, run the requested scope:
 
 ```bash
-npm test
-npm run test:unit
-npm run test:integration
+npm test                              # full local suite: unit + cli + agent
+npx vitest run --project unit         # scoped: unit only
+npx vitest run --project cli          # scoped: CLI integration only
 npm run test:coverage
 ```
 

.claude/skills/codemie-release/SKILL.md

@@ -106,7 +106,7 @@ Confirm release version 0.0.36?
 **After user confirms the version, run the full test suite before any commits.**
 
 ```bash
-npm run test:all
+npm test
 ```
 
 This runs unit tests, CLI integration tests, and agent tests (`unit` + `cli` + `agent` projects).

.github/workflows/ci.yml

@@ -183,10 +183,10 @@ jobs:
           path: dist/
 
       - name: Run unit tests
-        run: npm run test:unit
+        run: npx vitest run --project unit
 
       - name: Run integration tests
-        run: npm run test:integration
+        run: npx vitest run --project cli
 
   test-windows:
     name: Test (Windows)
@@ -216,7 +216,7 @@ jobs:
           path: dist/
 
       - name: Run unit tests
-        run: npm run test:unit
+        run: npx vitest run --project unit
 
       - name: Run integration tests
-        run: npm run test:integration
+        run: npx vitest run --project cli

CONTRIBUTING.md

@@ -199,15 +199,13 @@ To get the project running locally, follow these steps:
 
 ### Run All Tests
 ```bash
-npm test                        # Run tests in watch mode
-npm run test:run                # Run tests once
-npm run test:unit               # Run unit tests only
-npm run test:integration        # Run integration tests only
-npm run test:integration:agent  # Run agent integration tests only
-npm run test:all                # Run unit + CLI + agent tests in sequence
+npm test                        # Run unit + CLI + agent tests once, in sequence — the full local suite
+npm run test:watch              # Run tests in interactive watch mode
 ```
 
-> **Note:** Agent integration tests (`test:integration:agent` and the agent stage of `test:all`) only execute if you have a working CodeMie SSO setup. If your active profile provider is not `ai-run-sso`, the agent tests are automatically skipped and a message is printed — no credentials error will occur.
+`npm test` is the same test suite `npm run ci` runs (minus the `agent` project, which CI can't run without live SSO credentials) — it's what the release process uses too. To run just one project while iterating, use vitest directly, e.g. `npx vitest run --project unit` or `npx vitest run --project cli -- <file>`.
+
+> **Note:** The agent stage of `npm test` only executes if you have a working CodeMie SSO setup. If your active profile provider is not `ai-run-sso`, the agent tests are automatically skipped and a message is printed — no credentials error will occur.
 
 ### Run Validation Checks
 
@@ -252,8 +250,7 @@ Before committing, ensure:
 1. ✅ Commit message follows Conventional Commits format
 2. ✅ Code passes ESLint with zero warnings: `npm run lint`
 3. ✅ TypeScript compiles: `npm run build`
-4. ✅ All tests pass: `npm run test:run`
-5. ✅ Agent tests pass: `npm run test:integration:agent` or `npm run test:all` (only if CodeMie SSO is configured)
+4. ✅ All tests pass: `npm test` (agent stage only if CodeMie SSO is configured)
 6. ✅ No secrets exposed: `npm run validate:secrets` (optional, requires Docker)
 7. ✅ Dependencies have approved licenses: `npm run license-check`
 

docs/COMMANDS.md

@@ -185,7 +185,7 @@ reduced hidden-reasoning continuity. Prefer a current VS Code release (1.122 or
 stateless flag and marker suppression are honored.
 
 Check the daemon context with `codemie proxy status`. Automated VS Code BYOK configuration
-and routing coverage runs as part of `npm run test:all`.
+and routing coverage runs as part of `npm test`.
 
 #### Troubleshooting VS Code BYOK
 

package.json

@@ -35,14 +35,7 @@
     "copy-plugin": "node scripts/copy-plugins.js",
     "prepare:install-artifacts": "node scripts/prepare-install-artifacts.mjs",
     "dev": "tsc --watch",
-    "test": "vitest",
-    "test:unit": "vitest run --project unit",
-    "test:integration": "vitest run --project cli",
-    "test:integration:cli": "vitest run --project cli",
-    "test:integration:vscode-models": "vitest run --project cli tests/integration/vscode-byok.test.ts tests/integration/vscode-models.live.test.ts",
-    "test:integration:agent": "vitest run --project agent",
-    "test:run": "vitest run --project unit --project cli",
-    "test:all": "vitest run --project unit && vitest run --project cli && vitest run --project agent",
+    "test": "vitest run --project unit && vitest run --project cli && vitest run --project agent",
     "test:coverage": "vitest run --project unit --coverage",
     "test:watch": "vitest --watch",
     "test:ui": "vitest --ui",
@@ -55,7 +48,7 @@
     "commitlint:last": "commitlint --from HEAD~1 --to HEAD --verbose",
     "validate:secrets": "node scripts/validate-secrets.js",
     "license-check": "node scripts/license-check.js",
-    "ci": "npm run license-check && npm run lint && npm run build && npm run test:unit && npm run test:integration",
+    "ci": "npm run license-check && npm run lint && npm run build && vitest run --project unit && vitest run --project cli",
     "ci:full": "npm run commitlint:last && npm run ci",
     "prepare": "husky",
     "prepublishOnly": "npm run build",

scripts/release.sh

@@ -6,7 +6,7 @@
 # Designed to be resumable - can continue from failed steps
 #
 # Release flow: version bump → commit → agent tests gate → tag → push → GitHub release
-# Agent tests gate: runs `npm run test:integration:agent` before tagging.
+# Agent tests gate: runs `npx vitest run --project agent` before tagging.
 #   - Tests pass → continue automatically
 #   - Tests fail → release blocked (fix tests first)
 #   - Tests cannot run (missing SSO/JWT credentials) → manual confirmation required
@@ -207,7 +207,7 @@ echo ""
 echo "🧪 Running agent tests..."
 AGENT_TEST_JSON=$(mktemp /tmp/agent-test-XXXXX.json) || { echo "ERROR: mktemp failed, cannot capture agent test results"; exit 1; }
 trap 'rm -f "$AGENT_TEST_JSON"' EXIT INT TERM
-npm run test:integration:agent -- --reporter=verbose --reporter=json --outputFile="$AGENT_TEST_JSON"
+npx vitest run --project agent --reporter=verbose --reporter=json --outputFile="$AGENT_TEST_JSON"
 AGENT_EXIT_CODE=$?
 
 AGENT_PASSED=0
@@ -237,7 +237,7 @@ else
     echo "     (check: cat ~/.codemie/codemie-cli.config.json)"
     echo "   • CI:    set CI_IS_LOCAL_RUN=false and provide tests/.env.test.local"
     echo ""
-    echo "   To run manually: npm run test:integration:agent"
+    echo "   To run manually: npx vitest run --project agent"
     echo ""
     read -p "❓ Have you manually run agent tests and confirmed they pass? (y/N): " -n 1 -r
     echo