@mseep/open-computer-use 1.0.0

package/tests/test-single-user-mode.sh

@@ -0,0 +1,269 @@
+#!/usr/bin/env bash
+# SPDX-License-Identifier: FSL-1.1-Apache-2.0
+# Copyright (c) 2025 Open Computer Use Contributors
+#
+# Integration tests for SINGLE_USER_MODE feature.
+# Builds computer-use-server image, starts it with different env vars,
+# sends MCP requests via curl, verifies responses.
+#
+# Usage: ./tests/test-single-user-mode.sh
+#
+# Requires: docker, curl, jq
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+PROJECT_DIR="$(dirname "$SCRIPT_DIR")"
+IMAGE_NAME="computer-use-server-test-sum"
+BASE_PORT=18081
+PASSED=0
+FAILED=0
+
+# Colors
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+NC='\033[0m'
+
+cleanup() {
+    echo ""
+    echo "Cleaning up test containers..."
+    containers=$(docker ps -aq --filter "label=test=single-user-mode")
+    if [ -n "$containers" ]; then
+        docker rm -f $containers 2>/dev/null || true
+    fi
+}
+trap cleanup EXIT
+
+pass() {
+    echo -e "  ${GREEN}PASS${NC}: $1"
+    PASSED=$((PASSED + 1))
+}
+
+fail() {
+    echo -e "  ${RED}FAIL${NC}: $1"
+    FAILED=$((FAILED + 1))
+}
+
+# Build the server image
+echo "=== Building computer-use-server image ==="
+docker build -q -t "$IMAGE_NAME" "$PROJECT_DIR/computer-use-server" > /dev/null
+echo "Image built: $IMAGE_NAME"
+echo ""
+
+# MCP initialize request body
+MCP_INIT='{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'
+
+# MCP tools/call bash_tool request body
+MCP_BASH='{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"bash_tool","arguments":{"command":"echo hello-from-test","description":"integration test"}}}'
+
+# Start a server container and wait for health
+# Args: container_name port [env_vars...]
+start_server() {
+    local name="$1"
+    local port="$2"
+    shift 2
+    local env_args=""
+    for e in "$@"; do
+        env_args="$env_args -e $e"
+    done
+
+    docker run -d \
+        --name "$name" \
+        --label "test=single-user-mode" \
+        -p "$port:8081" \
+        $env_args \
+        "$IMAGE_NAME" > /dev/null
+
+    # Wait for health (max 30s)
+    local attempts=0
+    while [ $attempts -lt 30 ]; do
+        if curl -sf "http://localhost:$port/health" > /dev/null 2>&1; then
+            return 0
+        fi
+        sleep 1
+        attempts=$((attempts + 1))
+    done
+    echo "  WARNING: Server did not become healthy in 30s"
+    docker logs "$name" 2>&1 | tail -5
+    return 1
+}
+
+# Initialize MCP session and call bash_tool, return the tool result text
+# Args: port [extra_header]
+call_bash_tool() {
+    local port="$1"
+    local extra_header="${2:-}"
+    local header_args=""
+    if [ -n "$extra_header" ]; then
+        header_args="-H \"$extra_header\""
+    fi
+
+    # Step 1: Initialize MCP session
+    eval curl -sf "http://localhost:$port/mcp" \
+        -H "'Content-Type: application/json'" \
+        -H "'Accept: application/json, text/event-stream'" \
+        $header_args \
+        -d "'$MCP_INIT'" > /dev/null 2>&1 || true
+
+    # Step 2: Call bash_tool
+    local response
+    response=$(eval curl -s "http://localhost:$port/mcp" \
+        -H "'Content-Type: application/json'" \
+        -H "'Accept: application/json, text/event-stream'" \
+        $header_args \
+        -d "'$MCP_BASH'" 2>/dev/null || echo "curl_failed")
+
+    echo "$response"
+}
+
+stop_server() {
+    docker rm -f "$1" > /dev/null 2>&1 || true
+}
+
+# ============================================================================
+# Test 1: Lenient mode (SINGLE_USER_MODE unset) + no X-Chat-Id
+# Expected: tool response contains warning about SINGLE_USER_MODE
+# ============================================================================
+echo "[1/6] Lenient mode + no X-Chat-Id → works (no required error)"
+CONTAINER="test-sum-1"
+PORT=$((BASE_PORT))
+stop_server "$CONTAINER"
+if start_server "$CONTAINER" "$PORT"; then
+    RESPONSE=$(call_bash_tool "$PORT")
+    if echo "$RESPONSE" | grep -qi "required.*SINGLE_USER_MODE=false\|X-Chat-Id header is required"; then
+        fail "Got chat_id required error — should be lenient without SINGLE_USER_MODE"
+    elif echo "$RESPONSE" | grep -qi "SINGLE_USER_MODE"; then
+        pass "Response contains SINGLE_USER_MODE warning (with Docker output or alone)"
+    else
+        # Docker socket not available — but chat_id validation PASSED (no "required" error)
+        pass "Chat-id validation passed (no 'required' error)"
+    fi
+else
+    fail "Server failed to start"
+fi
+stop_server "$CONTAINER"
+
+# ============================================================================
+# Test 2: Lenient mode + X-Chat-Id present
+# Expected: normal response, NO warning
+# ============================================================================
+echo "[2/6] Lenient mode + X-Chat-Id → works without warning"
+CONTAINER="test-sum-2"
+PORT=$((BASE_PORT + 1))
+stop_server "$CONTAINER"
+if start_server "$CONTAINER" "$PORT"; then
+    RESPONSE=$(call_bash_tool "$PORT" "X-Chat-Id: test-session-123")
+    if echo "$RESPONSE" | grep -qi "SINGLE_USER_MODE"; then
+        fail "Should NOT contain SINGLE_USER_MODE warning when chat_id provided"
+    elif echo "$RESPONSE" | grep -qi "X-Chat-Id.*required"; then
+        fail "Should not require chat_id — it was provided"
+    else
+        pass "No warning when chat_id provided"
+    fi
+else
+    fail "Server failed to start"
+fi
+stop_server "$CONTAINER"
+
+# ============================================================================
+# Test 3: Single-user mode + no X-Chat-Id
+# Expected: works, NO warning
+# ============================================================================
+echo "[3/6] SINGLE_USER_MODE=true + no X-Chat-Id → works without warning"
+CONTAINER="test-sum-3"
+PORT=$((BASE_PORT + 2))
+stop_server "$CONTAINER"
+if start_server "$CONTAINER" "$PORT" "SINGLE_USER_MODE=true"; then
+    RESPONSE=$(call_bash_tool "$PORT")
+    if echo "$RESPONSE" | grep -qi "SINGLE_USER_MODE"; then
+        fail "Should NOT contain warning in single-user mode"
+    elif echo "$RESPONSE" | grep -qi "X-Chat-Id.*required"; then
+        fail "Should not require chat_id in single-user mode"
+    else
+        pass "No warning in single-user mode"
+    fi
+else
+    fail "Server failed to start"
+fi
+stop_server "$CONTAINER"
+
+# ============================================================================
+# Test 4: Single-user mode + X-Chat-Id present (should be ignored)
+# Expected: works, NO warning, uses default container
+# ============================================================================
+echo "[4/6] SINGLE_USER_MODE=true + X-Chat-Id → works (header ignored)"
+CONTAINER="test-sum-4"
+PORT=$((BASE_PORT + 3))
+stop_server "$CONTAINER"
+if start_server "$CONTAINER" "$PORT" "SINGLE_USER_MODE=true"; then
+    RESPONSE=$(call_bash_tool "$PORT" "X-Chat-Id: should-be-ignored")
+    if echo "$RESPONSE" | grep -qi "SINGLE_USER_MODE"; then
+        fail "Should NOT contain warning in single-user mode"
+    elif echo "$RESPONSE" | grep -qi "X-Chat-Id.*required"; then
+        fail "Should not require chat_id in single-user mode"
+    else
+        pass "Header ignored in single-user mode"
+    fi
+else
+    fail "Server failed to start"
+fi
+stop_server "$CONTAINER"
+
+# ============================================================================
+# Test 5: Strict multi-user mode + no X-Chat-Id
+# Expected: ERROR — chat_id required
+# ============================================================================
+echo "[5/6] SINGLE_USER_MODE=false + no X-Chat-Id → error"
+CONTAINER="test-sum-5"
+PORT=$((BASE_PORT + 4))
+stop_server "$CONTAINER"
+if start_server "$CONTAINER" "$PORT" "SINGLE_USER_MODE=false"; then
+    RESPONSE=$(call_bash_tool "$PORT")
+    if echo "$RESPONSE" | grep -qi "required"; then
+        pass "Correctly rejected — chat_id required"
+    else
+        fail "Expected 'required' error, got: $(echo "$RESPONSE" | head -c 200)"
+    fi
+else
+    fail "Server failed to start"
+fi
+stop_server "$CONTAINER"
+
+# ============================================================================
+# Test 6: Strict multi-user mode + X-Chat-Id present
+# Expected: works normally, no warning
+# ============================================================================
+echo "[6/6] SINGLE_USER_MODE=false + X-Chat-Id → works"
+CONTAINER="test-sum-6"
+PORT=$((BASE_PORT + 5))
+stop_server "$CONTAINER"
+if start_server "$CONTAINER" "$PORT" "SINGLE_USER_MODE=false"; then
+    RESPONSE=$(call_bash_tool "$PORT" "X-Chat-Id: test-session-456")
+    if echo "$RESPONSE" | grep -qi "required"; then
+        fail "Should NOT get 'required' error when chat_id provided"
+    elif echo "$RESPONSE" | grep -qi "SINGLE_USER_MODE"; then
+        fail "Should NOT contain warning in strict mode with chat_id"
+    else
+        pass "Works with chat_id in strict mode"
+    fi
+else
+    fail "Server failed to start"
+fi
+stop_server "$CONTAINER"
+
+# ============================================================================
+# Summary
+# ============================================================================
+echo ""
+echo "==============================="
+echo "  PASSED: $PASSED"
+echo "  FAILED: $FAILED"
+echo ""
+if [ "$FAILED" -eq 0 ]; then
+    echo -e "  RESULT: ${GREEN}ALL PASSED${NC}"
+else
+    echo -e "  RESULT: ${RED}$FAILED FAILED${NC}"
+fi
+echo "==============================="
+
+exit "$FAILED"

package/tests/test-skill-no-hardcoded-models.sh

@@ -0,0 +1,65 @@
+#!/bin/bash
+# SPDX-License-Identifier: FSL-1.1-Apache-2.0
+# Copyright (c) 2025 Open Computer Use Contributors
+#
+# Phase 2 D-16: enforces no-hardcoded-model-names invariant for sub-agent skill files.
+# Greps for forbidden tokens; lines wrapped in <!-- canonical-example -->...<!-- /canonical-example -->
+# markers (or otherwise containing "canonical-example") are excluded — those are
+# narrative examples that legitimately need to name a model (e.g., claude's `sonnet` baseline).
+# Lines that mention "list-subagent-models" are also excluded (the script name itself
+# contains "models" but does not encode a hardcoded model id).
+
+set -uo pipefail
+
+PASSED=0; FAILED=0; FAILURES=""
+
+pass() { PASSED=$((PASSED + 1)); echo "  PASS: $1"; }
+fail() { FAILED=$((FAILED + 1)); FAILURES="${FAILURES}\n  - $1"; echo "  FAIL: $1"; }
+
+FORBIDDEN_PATTERN='sonnet|opus|haiku|gpt-5-codex|claude-sonnet-4-6|qwen-3-coder|claude-opus|claude-haiku'
+FILES=(
+    "skills/public/sub-agent/SKILL.md"
+    "skills/public/sub-agent/references/usage.md"
+)
+
+# Resolve repo root: the script may be invoked from any directory.
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+REPO_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+
+echo "=== Skill audit: no hardcoded model names ==="
+for f in "${FILES[@]}"; do
+    fpath="$REPO_ROOT/$f"
+    if [ ! -f "$fpath" ]; then
+        fail "$f does not exist"
+        continue
+    fi
+    # Lines containing forbidden tokens, EXCLUDING marker-tagged lines.
+    # No additional exemptions — FORBIDDEN_PATTERN does not match the
+    # discovery command literal "list-subagent-models", so it never causes
+    # false positives. (The previous "grep -v list-subagent-models" filter
+    # was a defensive overreach that hid real violations on any line that
+    # mentioned the discovery command.)
+    hits=$(grep -nE "$FORBIDDEN_PATTERN" "$fpath" \
+            | grep -v 'canonical-example' \
+            || true)
+    if [ -z "$hits" ]; then
+        pass "$f has no unmarked hardcoded model names"
+    else
+        fail "$f contains unmarked hardcoded model names:"
+        echo "$hits" | sed 's/^/      /'
+    fi
+done
+
+echo ""
+echo "==============================="
+echo "  PASSED: $PASSED  FAILED: $FAILED"
+if [ "$FAILED" -gt 0 ]; then
+    echo ""
+    echo "  Failures:"
+    echo -e "$FAILURES"
+    echo "  RESULT: FAIL"
+    exit 1
+else
+    echo "  RESULT: ALL TESTS PASSED"
+    exit 0
+fi

package/tests/test-subagent-cli-surface.py

@@ -0,0 +1,137 @@
+# SPDX-License-Identifier: FSL-1.1-Apache-2.0
+# Copyright (c) 2025 Open Computer Use Contributors
+"""Test that sub_agent FastMCP Tool.description is correct per SUBAGENT_CLI.
+
+Verifies Plan 01-04 behaviour (D-05):
+- claude/unset: all three Claude aliases (sonnet, opus, haiku) appear in doc
+- opencode: no bare Claude alias tokens; doc contains list-subagent-models redirect
+- codex: no Claude alias tokens at all; doc contains list-subagent-models redirect
+- unknown CLI falls back to the claude variant
+
+Uses snippet extraction (same pattern as test_subagent_docstring.py) to avoid
+importing the full mcp_tools module which requires docker/mcp runtime.
+"""
+
+import re
+import sys
+from pathlib import Path
+
+import pytest
+
+ROOT = Path(__file__).resolve().parent.parent
+SRC = ROOT / "computer-use-server" / "mcp_tools.py"
+
+CLAUDE_ALIASES = ("sonnet", "opus", "haiku")
+
+
+def _load_docstring_helpers():
+    """Extract and exec only the docstring-related portion of mcp_tools.py.
+
+    Avoids importing the full module (which requires mcp, docker, etc.).
+    Returns a namespace dict with the three constants and the helper function.
+    """
+    src = SRC.read_text()
+    lines = src.splitlines()
+    start = None
+    end = None
+    for i, line in enumerate(lines):
+        if start is None and line.startswith("_SUBAGENT_DOC_CLAUDE = "):
+            start = i
+        if start is not None and line.startswith("    return _SUBAGENT_DOC_CLAUDE"):
+            end = i + 1
+            break
+
+    assert start is not None, "Could not find _SUBAGENT_DOC_CLAUDE in mcp_tools.py"
+    assert end is not None, "Could not find end of _subagent_docstring_for_cli"
+
+    snippet = "\n".join(lines[start:end])
+    ns: dict = {}
+    exec(compile(snippet, str(SRC), "exec"), ns)  # noqa: S102
+    return ns
+
+
+_NS = _load_docstring_helpers()
+
+
+def _get_subagent_doc(cli: str) -> str:
+    """Return the sub_agent docstring for the given CLI value."""
+    return _NS["_subagent_docstring_for_cli"](cli)
+
+
+# ---------------------------------------------------------------------------
+# Claude CLI — all three aliases must appear
+# ---------------------------------------------------------------------------
+
+def test_claude_doc_contains_aliases():
+    doc = _get_subagent_doc("claude")
+    for alias in CLAUDE_ALIASES:
+        assert alias in doc, f"Expected alias '{alias}' in claude sub_agent doc"
+
+
+def test_unset_cli_doc_contains_aliases():
+    """Empty string (unset SUBAGENT_CLI) falls back to claude variant."""
+    doc = _get_subagent_doc("")
+    for alias in CLAUDE_ALIASES:
+        assert alias in doc, f"Expected alias '{alias}' in unset-CLI sub_agent doc"
+
+
+# ---------------------------------------------------------------------------
+# opencode CLI — no bare Claude alias tokens; list-subagent-models present
+# ---------------------------------------------------------------------------
+
+def test_opencode_doc_contains_list_subagent_models():
+    doc = _get_subagent_doc("opencode")
+    assert "list-subagent-models" in doc, (
+        "opencode sub_agent doc must mention 'list-subagent-models'"
+    )
+
+
+def test_opencode_doc_no_bare_aliases():
+    """opencode doc must not expose bare sonnet/opus/haiku tokens.
+
+    The doc may legitimately reference anthropic/claude-sonnet-4-6 in an
+    example; that is acceptable because the alias is embedded in a
+    provider/model id (preceded by 'claude-' or '/'). The assertion here
+    only rejects bare occurrences where no such prefix is present.
+    """
+    doc = _get_subagent_doc("opencode")
+    for alias in CLAUDE_ALIASES:
+        matches = list(re.finditer(rf"\b{alias}\b", doc))
+        for m in matches:
+            pre = doc[max(0, m.start() - 10) : m.start()]
+            assert "/" in pre or "claude-" in pre, (
+                f"Bare alias '{alias}' found at offset {m.start()} in opencode doc: "
+                f"...{doc[max(0, m.start()-20):m.end()+20]}..."
+            )
+
+
+# ---------------------------------------------------------------------------
+# codex CLI — zero Claude alias tokens; list-subagent-models present
+# ---------------------------------------------------------------------------
+
+def test_codex_doc_contains_list_subagent_models():
+    doc = _get_subagent_doc("codex")
+    assert "list-subagent-models" in doc, (
+        "codex sub_agent doc must mention 'list-subagent-models'"
+    )
+
+
+def test_codex_doc_no_aliases():
+    """codex doc must have zero occurrences of any Claude alias."""
+    doc = _get_subagent_doc("codex")
+    for alias in CLAUDE_ALIASES:
+        assert alias.lower() not in doc.lower(), (
+            f"Unexpected alias '{alias}' found in codex sub_agent doc"
+        )
+
+
+# ---------------------------------------------------------------------------
+# Unknown CLI falls back to claude variant
+# ---------------------------------------------------------------------------
+
+def test_unknown_cli_falls_back_to_claude():
+    claude_doc = _get_subagent_doc("claude")
+    unknown_doc = _get_subagent_doc("banana")
+    assert unknown_doc == claude_doc, (
+        "Unknown CLI 'banana' must fall back to claude sub_agent doc"
+    )

package/tests/test-subagent-runtime.sh

@@ -0,0 +1,109 @@
+#!/bin/bash
+# SPDX-License-Identifier: FSL-1.1-Apache-2.0
+# Copyright (c) 2025 Open Computer Use Contributors
+#
+# Umbrella runner for all sub-agent runtime tests (Phase 1 — Plans 01-03..01-06,
+# Phase 2 — Plan 02-02 removes the deprecation step, Plan 02-06 adds audit step).
+#
+# Invokes in sequence:
+#   1. pytest tests/test-subagent-cli-surface.py          (per-CLI docstring assertions)
+#   2. pytest tests/test-opencode-error-mapping.py        (opencode error event mapping)
+#   3. pytest tests/test_subagent_docstring.py            (docstring helper unit tests)
+#   4. pytest tests/test-default-model-resolution.py      (D-08/D-09 resolution order)
+#   5. bash  tests/test-list-subagent-models.sh           (script invocation per CLI)
+#   6. bash  tests/test-skill-no-hardcoded-models.sh      (D-16 skill audit)
+#
+# Exit code: 0 = all sub-tests passed, 1 = one or more failed.
+# Usage: bash tests/test-subagent-runtime.sh
+
+set -uo pipefail
+
+TEST_DIR="$(cd "$(dirname "$0")" && pwd)"
+REPO_ROOT="$(cd "$TEST_DIR/.." && pwd)"
+
+PASSED=0
+FAILED=0
+FAILURES=""
+
+pass() {
+    PASSED=$((PASSED + 1))
+    echo "  PASS: $1"
+}
+
+fail() {
+    FAILED=$((FAILED + 1))
+    FAILURES="${FAILURES}\n  - $1"
+    echo "  FAIL: $1"
+}
+
+run_pytest() {
+    local label="$1"
+    local testfile="$2"
+    set +e
+    python3 -m pytest "$testfile" -x --tb=short -q 2>&1
+    local rc=$?
+    set -e
+    if [ "$rc" -eq 0 ]; then
+        pass "$label"
+    else
+        fail "$label (pytest exit $rc)"
+    fi
+}
+
+run_bash() {
+    local label="$1"
+    local script="$2"
+    set +e
+    bash "$script" 2>&1
+    local rc=$?
+    set -e
+    if [ "$rc" -eq 0 ]; then
+        pass "$label"
+    else
+        fail "$label (exit $rc)"
+    fi
+}
+
+echo "=== Sub-agent runtime tests ==="
+echo ""
+
+echo "[1/6] Per-CLI docstring surface (test-subagent-cli-surface.py)"
+run_pytest "test-subagent-cli-surface.py" "$TEST_DIR/test-subagent-cli-surface.py"
+
+echo ""
+echo "[2/6] opencode error event mapping (test-opencode-error-mapping.py)"
+run_pytest "test-opencode-error-mapping.py" "$TEST_DIR/test-opencode-error-mapping.py"
+
+echo ""
+echo "[3/6] Sub-agent docstring helper unit tests (test_subagent_docstring.py)"
+run_pytest "test_subagent_docstring.py" "$TEST_DIR/test_subagent_docstring.py"
+
+echo ""
+echo "[4/6] Default model resolution order (test-default-model-resolution.py)"
+run_pytest "test-default-model-resolution.py" "$TEST_DIR/test-default-model-resolution.py"
+
+echo ""
+echo "[5/6] list-subagent-models script invocation (test-list-subagent-models.sh)"
+run_bash "test-list-subagent-models.sh" "$TEST_DIR/test-list-subagent-models.sh"
+
+echo ""
+echo "[6/6] Skill no-hardcoded-models audit (test-skill-no-hardcoded-models.sh)"
+run_bash "test-skill-no-hardcoded-models.sh" "$TEST_DIR/test-skill-no-hardcoded-models.sh"
+
+# Summary
+echo ""
+echo "==============================="
+echo "  PASSED: $PASSED"
+echo "  FAILED: $FAILED"
+if [ "$FAILED" -gt 0 ]; then
+    echo ""
+    echo "  Failures:"
+    echo -e "$FAILURES"
+    echo ""
+    echo "  RESULT: FAIL"
+    exit 1
+else
+    echo ""
+    echo "  RESULT: ALL TESTS PASSED"
+    exit 0
+fi