aria-code 0.83.0__tar.gz → 0.85.0__tar.gz

{aria_code-0.83.0/src/aria_code.egg-info → aria_code-0.85.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aria-code
-Version: 0.83.0
+Version: 0.85.0
 Summary: Open-source AI agent for coding, financial analysis and logistics operations: numbers you can check, client data kept apart, actions a person approves
 License: Apache-2.0
 Project-URL: Homepage, https://github.com/artheras/aria-code

{aria_code-0.83.0 → aria_code-0.85.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "aria-code"
-version = "0.83.0"
+version = "0.85.0"
 description = "Open-source AI agent for coding, financial analysis and logistics operations: numbers you can check, client data kept apart, actions a person approves"
 readme = "README.md"
 license = { text = "Apache-2.0" }

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/_version.py

@@ -17,4 +17,4 @@ from __future__ import annotations
 
 __all__ = ["__version__"]
 
-__version__ = "0.83.0"
+__version__ = "0.85.0"

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/apps/cli/prompts/coding.py

@@ -26,6 +26,12 @@ CODING_SYSTEM_PROMPT = (
     "For multi-file projects: emit one <tool_call> per file in the SAME response (up to 5 write_file calls). "
     "Then run/verify in the NEXT response. Never mix write_file and run_command in the same response.\n\n"
 
+    "## FINAL SUMMARY\n"
+    "The user has already seen every step: each file written, with a preview, and each command's "
+    "output. End with 1-4 sentences: which files changed, what was verified and the result "
+    "(\"3 tests passed\"), and anything left open. Do NOT paste file contents, diffs or command "
+    "output back; quote a line or two only when it matters.\n\n"
+
     "## ABSOLUTELY FORBIDDEN\n"
     "1. NEVER pass slash-commands (/config, /model, /note, /apikey, etc.) to run_command — "
     "   they are NOT shell commands. To change policy tell the user to type the slash command directly.\n"

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/apps/cli/prompts/system_prompts.py

@@ -46,6 +46,8 @@ def build_response_style_rule(lang: str) -> str:
             "- For analysis: Conclusion -> Evidence -> Risk/Limitations -> Next.\n"
             "- Keep output compact: avoid repeated disclaimers, generic education, and duplicated sections.\n"
             "- Keep the interactive answer near one terminal screen (about 24 lines) unless the user explicitly requests detail.\n"
+            "- After using tools, do not paste back what the user already saw: name the files you wrote or changed and "
+            "quote only the lines that matter, and summarise command output (\"4 tests passed\") instead of reprinting it.\n"
             "- Never invent data. If a field is unavailable, omit it or note 'N/A' and continue with available data.\n\n"
         )
     return (
@@ -59,6 +61,7 @@ def build_response_style_rule(lang: str) -> str:
         "- 分析类回答按：结论 -> 依据 -> 风险/限制 -> 下一步。\n"
         "- 输出要紧凑：避免重复免责声明、泛泛教学和重复章节。\n"
         "- 交互回答尽量控制在一屏（约24行）；用户明确要求详细内容时除外。\n"
+        "- 用过工具后不要重复用户已看到的内容：写入或修改的文件只说文件名并只引用关键几行，命令输出用一句话概括（如“4 个测试全部通过”），不要整段重贴。\n"
         "- 绝不编造数据。字段不可用时说明数据源未返回，并继续使用已有数据。\n\n"
     )
 

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/apps/cli/runtime_consumer.py

@@ -33,15 +33,44 @@ _REPETITION_NOTICE = (
 
 
 def _redact_activity_text(value: Any, limit: int = 90) -> str:
+    """One line, secrets masked — for anything echoed to the screen or a log.
+
+    Shell commands are now shown on their step line, so this also covers
+    "Authorization: Bearer …", credentials in URLs and the usual key shapes.
+    """
     text = re.sub(r"\s+", " ", str(value or "")).strip()
     text = re.sub(
-        r"(?i)(api[_-]?key|token|password|secret)(\s*[=:]\s*|\s+)[^\s]+",
+        r"(?i)(api[_-]?key|token|password|passwd|secret)(\s*[=:]\s*|\s+)[^\s]+",
         r"\1\2***",
         text,
     )
+    text = re.sub(r"(?i)\b(bearer|basic)\s+[A-Za-z0-9._~+/=-]+", r"\1 ***", text)
+    text = re.sub(r"://[^/\s:@]+:[^@\s]+@", "://***@", text)
+    text = re.sub(r"\b(sk|pk|rk)-[A-Za-z0-9_-]{8,}|\bgh[pousr]_[A-Za-z0-9]{12,}|\bAIza[0-9A-Za-z_-]{20,}"
+                  r"|\bxox[abposr]-[A-Za-z0-9-]{10,}", "***", text)
     return text[:limit]
 
 
+# How long the runtime measured a tool call to take, carried with its result.
+MEASURED_ELAPSED_KEY = "_elapsed_s"
+
+
+def with_measured_elapsed(result: Any, elapsed: Any) -> Any:
+    if isinstance(result, dict) and isinstance(elapsed, (int, float)):
+        return {**result, MEASURED_ELAPSED_KEY: float(elapsed)}
+    return result
+
+
+def approval_subject(tool_name: str, params: dict[str, Any]) -> str:
+    """What an answered approval leaves behind: the file, or the command with secrets masked."""
+    if tool_name == "run_command":
+        return _redact_activity_text(params.get("command", ""), limit=80)
+    for key in ("path", "file_path"):
+        if params.get(key):
+            return Path(str(params[key])).name
+    return tool_name.replace("_", " ")
+
+
 def _tool_activity_hint(tool: str, params: dict[str, Any]) -> str:
     """Return one safe, useful argument for a compact activity row."""
     if not isinstance(params, dict):
@@ -506,6 +535,11 @@ class TerminalRuntimeEventConsumer:
         if not starts:
             self.tool_start_times.pop(tool, None)
         elapsed_ms = int((time.time() - started_at) * 1000)
+        # The runtime times each call itself. Measuring from the announcement
+        # counted the person reading an approval, and results of a batch
+        # arrive together — "✓ writing file test_fx.py 13.9s" for a 9ms write.
+        if isinstance(summary, dict) and isinstance(summary.get(MEASURED_ELAPSED_KEY), (int, float)):
+            elapsed_ms = int(summary[MEASURED_ELAPSED_KEY] * 1000)
         params_queue = self.tool_params.get(tool) or []
         params = params_queue.pop(0) if params_queue else {}
         if not params_queue:
@@ -519,6 +553,10 @@ class TerminalRuntimeEventConsumer:
             detail = ""
             if not ok and isinstance(summary, dict):
                 detail = str(summary.get("error") or "")[:90]
+            elif ok:
+                # Name what was done: two bare "✓ writing file" lines in a row
+                # did not say which file each was.
+                detail = _tool_activity_hint(tool, params)
             self.print_tool_done(tool, elapsed_ms, success=ok, summary=detail)
 
         ts = time.strftime("%H:%M:%S")
@@ -588,7 +626,7 @@ class TerminalRuntimeEventConsumer:
         elif isinstance(event, AgentEventToolCall):
             self.on_tool_call(event.tool, event.params)
         elif isinstance(event, AgentEventToolResult):
-            self.on_tool_result(event.tool, event.result)
+            self.on_tool_result(event.tool, with_measured_elapsed(event.result, getattr(event, "elapsed", None)))
         elif isinstance(event, AgentEventStatus):
             self.on_status(event.state, event.message)
 
@@ -638,7 +676,9 @@ class TerminalApprovalEventConsumer:
         if not approval.approved:
             from aria_code.ui.render.output import print_tool_blocked
 
-            print_tool_blocked(tool_name, "用户取消", console=self.console, has_rich=self.has_rich)
+            zh = str(self.terminal.config.get("ui_lang", "en")).lower().startswith("zh")
+            print_tool_blocked(tool_name, "用户取消" if zh else "declined", console=self.console,
+                               has_rich=self.has_rich)
         return approval
 
     def apply(self, tool_params: dict, approval: ApprovalDecision) -> dict:
@@ -649,7 +689,9 @@ class TerminalApprovalEventConsumer:
             try:
                 self.save_config(self.terminal.config)
                 if self.has_rich:
-                    self.console.print("  [dim]策略已升级为 balanced 并保存[/dim]")
+                    zh = str(self.terminal.config.get("ui_lang", "en")).lower().startswith("zh")
+                    self.console.print("  [dim]策略已升级为 balanced 并保存[/dim]" if zh else
+                                       "  [dim]Policy set to balanced and saved[/dim]")
             except Exception:
                 pass
         return tool_params

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/apps/cli/tool_executor.py

@@ -57,7 +57,7 @@ def _command_approval_prefix(command: str) -> tuple[str, ...]:
     return ()
 
 
-def _arrow_select(options, selected: int = 0, title: str = "") -> int:
+def _arrow_select(options, selected: int = 0, title: str = "", **_kwargs) -> int:
     return len(options) - 1
 
 
@@ -472,7 +472,10 @@ def _confirm_tool_execution_decision(tool_name: str, params: dict,
                 ("Allow & set balanced", "允许并升级策略（本会话有效）" if zh else "and use balanced for this session"),
                 ("No", "拒绝执行" if zh else "do not run it"),
             ]
-            choice = _arrow_select(options, selected=0, title="")
+            # Imported here: aria_cli rebinds this function to its own globals.
+            from aria_code.apps.cli.runtime_consumer import approval_subject as _subject
+            choice = _arrow_select(options, selected=0, title="",
+                                   collapse_to=_subject(tool_name, params))
             if choice == 0:
                 return ApprovalDecision.allow(policy="balanced", user_approved=True)
             if choice == 1:
@@ -518,7 +521,10 @@ def _confirm_tool_execution_decision(tool_name: str, params: dict,
         ("Yes, allow all tools",             "本会话内所有工具自动允许" if zh else "every tool, this session"),
         ("No",                               ""),
     ]
-    choice = _arrow_select(options, selected=0, title="")
+    # Imported here: aria_cli rebinds this function to its own globals.
+    from aria_code.apps.cli.runtime_consumer import approval_subject as _subject
+    choice = _arrow_select(options, selected=0, title="",
+                           collapse_to=_subject(tool_name, params))
 
     if choice == 0:
         if tool_name == "run_command":

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/apps/cli/tools/system_tools.py

@@ -77,6 +77,41 @@ def _cprint(msg: str, *, console, has_rich: bool) -> None:
         print(plain)
 
 
+OUTPUT_PREVIEW_LINES = 5
+
+
+def print_command_outcome(console, returncode: int, stdout: str, stderr: str, full_output_path: str = "") -> None:
+    """One step line and the last few lines of what the command printed.
+
+    This printed "Command completed (exit 0)", up to six stdout lines and
+    "full output saved" — and nothing at all for `python3 -m unittest -v`,
+    whose report goes to stderr, so the model pasted the test results into
+    its answer instead. Lines are plain text: program output such as
+    "[/dim]" is not Rich markup. As Codex does, the tail is kept — that is
+    where "OK" or the failure is — and the rest is counted.
+    """
+    from rich.text import Text
+
+    lines = stdout.rstrip().splitlines() + (stderr.rstrip().splitlines() if stderr.strip() else [])
+    ok = returncode == 0
+    head = Text("  ⎿  ", style="dim")
+    head.append(f"exit {returncode}", style="green" if ok else "red")
+    if lines:
+        head.append(f" · {len(lines)} line{'s' if len(lines) != 1 else ''}", style="dim")
+    console.print(head)
+    notes = []
+    if len(lines) > OUTPUT_PREVIEW_LINES:
+        notes.append(f"… +{len(lines) - OUTPUT_PREVIEW_LINES} lines")
+    if full_output_path:
+        from aria_code.ui.render.output import display_path
+
+        notes.append(f"full output saved: {display_path(full_output_path)}")
+    if notes:
+        console.print(Text("     " + " · ".join(notes), style="dim"))
+    for line in lines[-OUTPUT_PREVIEW_LINES:]:
+        console.print(Text("     " + line[:160], style="dim" if ok else "red"), overflow="ellipsis", no_wrap=True)
+
+
 def tool_run_command(
     params: dict,
     *,
@@ -374,20 +409,8 @@ def tool_run_command(
         # ── End auto-fix ─────────────────────────────────────────────────────
 
         if has_rich and console is not None:
-            if result.returncode == 0:
-                console.print(f"  [green]Command completed[/green] [dim](exit {result.returncode})[/dim]")
-            else:
-                console.print(f"  [dim]Command exited {result.returncode}[/dim]")
-            out_preview = output.strip().splitlines()[:6]
-            for ol in out_preview:
-                console.print(f"    [dim]{ol[:120]}[/dim]")
-            if len(output.strip().splitlines()) > 6:
-                console.print("    [dim]...truncated[/dim]")
-            if output_artifact.get("full_output_path"):
-                console.print("    [dim]full output saved[/dim]")
-            if stderr.strip() and result.returncode != 0:
-                for el in stderr.strip().splitlines()[:3]:
-                    console.print(f"    [red]{el[:120]}[/red]")
+            print_command_outcome(console, result.returncode, output, stderr,
+                                  output_artifact.get("full_output_path", ""))
         else:
             print(f"  Command exit: {result.returncode}")
         data = {

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/aria_cli.py

@@ -2623,7 +2623,10 @@ def _format_tool_params(tool_name: str, params: dict) -> str:
         # Basename only — informative without leaking full workspace paths.
         return _short(pathlib.Path(str(params.get("path", ""))).name or "file tool", 40)
     if tool_name == "run_command":
-        return "shell tool"
+        # The command itself, secrets masked: "shell tool" said nothing, while
+        # the approval menu under it named the command anyway.
+        from apps.cli.runtime_consumer import _redact_activity_text
+        return _short(_redact_activity_text(params.get("command", ""), limit=200) or "shell", 72)
     if tool_name == "list_files":
         return _short(params.get("pattern") or params.get("path") or "file tool", 40)
     if tool_name == "search_code":
@@ -2782,7 +2785,7 @@ def _print_tool_done(tool_name: str, elapsed_ms: int, success: bool = True, summ
         return
     action = _TOOL_ACTION_LABELS.get(tool_name, tool_name.replace("_", " "))
     icon   = "[green]✓[/green]" if success else "[red]✗[/red]"
-    t_txt  = f"({elapsed_ms}ms)" if elapsed_ms > 0 else ""
+    t_txt  = ("" if elapsed_ms <= 0 else f"{elapsed_ms}ms" if elapsed_ms < 1000 else f"{elapsed_ms / 1000:.1f}s")
     if summary:
         # 单行预算:summary 按显示宽度(CJK 记 2 格)截断,保证图标+动作+摘要+时长
         # 排在同一行——否则 Rich 整行回卷,时长 chip 顶格孤立在下一行,树形缩进被破坏。

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/runtime/gateway.py

@@ -126,7 +126,9 @@ async def run_turn(
                 on_tool_call(ev.tool, dict(ev.params))
         elif isinstance(ev, AgentEventToolResult):
             if on_tool_result is not None:
-                on_tool_result(ev.tool, dict(ev.result))
+                from aria_code.apps.cli.runtime_consumer import with_measured_elapsed
+
+                on_tool_result(ev.tool, with_measured_elapsed(dict(ev.result), getattr(ev, "elapsed", None)))
         elif isinstance(ev, AgentEventStatus):
             if on_status is not None:
                 phase = getattr(ev, "phase", "") or getattr(ev, "state", "") or ""

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/ui/console.py

@@ -139,10 +139,14 @@ if HAS_RICH:
         from ui.render.output import adapt_markdown_for_width
 
         markup = adapt_markdown_for_width(markup, getattr(console, "width", 80))
+        # Rich's "bw" theme paints its own white background, so every code
+        # block was a white slab in a dark terminal. The ANSI themes keep the
+        # terminal's background and colour with its palette.
+        code_theme = "ansi_light" if ARIA_RICH_THEME_NAME == "light" else "ansi_dark"
         return Markdown(
             markup,
-            code_theme="bw",
-            inline_code_theme="bw",
+            code_theme=code_theme,
+            inline_code_theme=code_theme,
         )
 else:
     class _FallbackConsole:

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code/ui/picker.py

@@ -19,7 +19,8 @@ if _HAS_TERMIOS:
 
 def arrow_select(options: list, selected: int = 0, title: str = "",
                  max_visible: int = 10,
-                 controls_hint: str = "↑↓  Enter  Esc/q Cancel") -> int:
+                 controls_hint: str = "↑↓  Enter  Esc/q Cancel",
+                 collapse_to: str | None = None) -> int:
     """Interactive arrow-key selector with scrolling.
 
     Args:
@@ -27,6 +28,10 @@ def arrow_select(options: list, selected: int = 0, title: str = "",
         selected:    initially highlighted index
         title:       optional header line
         max_visible: max rows shown at once; scrolls when list is larger
+        collapse_to: when given, the menu is erased once answered and one
+                     line is left in its place — "✓ Yes  fx.py" — instead of
+                     the whole menu, as Codex leaves an answered approval.
+                     The text names what was being approved.
     Returns:
         index of chosen option, or -1 if cancelled
     """
@@ -130,21 +135,13 @@ def arrow_select(options: list, selected: int = 0, title: str = "",
 
     _render.drawn = False
     _render.last_phys_height = visible + 1
+    header_text = f"  {title}  {controls_hint}" if title else f"  {controls_hint}"
+    # blank line + rule, then the title/hint line(s) + blank line
+    header_height = 2 + _physical_lines(header_text) + 1
+    result = -1
 
-    try:
-        _esc_watcher.pause()
-        tty.setcbreak(fd)
-        sys.stdout.flush()
-
-        # Top boundary
-        _raw(f"\n  \033[2m{_rule}\033[0m\n")
-        if title:
-            _raw(f"  \033[1m{title}\033[0m  \033[2m{controls_hint}\033[0m\n\n")
-        else:
-            _raw(f"  \033[2m{controls_hint}\033[0m\n\n")
-
-        _render()
-
+    def _choose() -> int:
+        nonlocal selected
         while True:
             ch = os.read(fd, 1)
             if ch == b'\x1b':
@@ -175,12 +172,40 @@ def arrow_select(options: list, selected: int = 0, title: str = "",
                 selected = n - 1; _render()
             elif ch in (b'\x03', b'\x04'):
                 return -1
+
+    try:
+        _esc_watcher.pause()
+        tty.setcbreak(fd)
+        sys.stdout.flush()
+
+        # Top boundary
+        _raw(f"\n  \033[2m{_rule}\033[0m\n")
+        if title:
+            _raw(f"  \033[1m{title}\033[0m  \033[2m{controls_hint}\033[0m\n\n")
+        else:
+            _raw(f"  \033[2m{controls_hint}\033[0m\n\n")
+
+        _render()
+        result = _choose()
+        return result
     except (EOFError, KeyboardInterrupt, OSError):
         return -1
     finally:
         termios.tcsetattr(fd, termios.TCSADRAIN, old_settings)
         _esc_watcher.resume()
-        _raw(f"  \033[2m{_rule}\033[0m\n\n")
+        if collapse_to is not None and _render.drawn:
+            label = ""
+            if 0 <= result < n:
+                opt = options[result]
+                label = opt[0] if isinstance(opt, tuple) else opt
+            declined = result < 0 or label.strip().lower() in ("no", "否", "取消", "cancel")
+            mark = "\033[31m✗\033[0m" if declined else "\033[32m✓\033[0m"
+            shown = label or "Cancelled"
+            subject = f"  \033[2m{collapse_to}\033[0m" if collapse_to else ""
+            # Up over the header and the menu, clear to the end, one line instead.
+            _raw(f"\033[{header_height + _render.last_phys_height}A\r\033[J  {mark} {shown}{subject}\n")
+        else:
+            _raw(f"  \033[2m{_rule}\033[0m\n\n")
 
 
 async def run_picker_in_thread(options: list, current_idx: int,

{aria_code-0.83.0 → aria_code-0.85.0/src/aria_code.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aria-code
-Version: 0.83.0
+Version: 0.85.0
 Summary: Open-source AI agent for coding, financial analysis and logistics operations: numbers you can check, client data kept apart, actions a person approves
 License: Apache-2.0
 Project-URL: Homepage, https://github.com/artheras/aria-code

{aria_code-0.83.0 → aria_code-0.85.0}/src/aria_code.egg-info/SOURCES.txt

@@ -711,6 +711,7 @@ tests/test_tls_trust_store.py
 tests/test_tool_policy.py
 tests/test_tool_rounds_on_openai_apis.py
 tests/test_tool_scope.py
+tests/test_tool_step_output.py
 tests/test_trading_service.py
 tests/test_tradingview_bridge.py
 tests/test_turn_planning.py

{aria_code-0.83.0 → aria_code-0.85.0}/tests/test_release_chain.py

@@ -163,7 +163,7 @@ class CalledWorkflowsAcceptATag(unittest.TestCase):
 class NpmDispatcherGate(unittest.TestCase):
     """Exercise the actual release step without contacting npm."""
 
-    def _run_wait(self, npm_available: bool):
+    def _run_wait(self, npm_available: bool, status: str = ""):
         steps = _load("publish")["jobs"]["publish-npm"]["steps"]
         script = next(step["run"] for step in steps
                       if step.get("name") == "Wait for this version's platform packages")
@@ -182,7 +182,16 @@ class NpmDispatcherGate(unittest.TestCase):
             # The registry is asked directly (curl), not through npm's cache.
             calls = bin_dir / "calls"
             curl = bin_dir / "curl"
-            curl.write_text(f"#!/bin/sh\necho \"$@\" >> {calls}\nexit {0 if npm_available else 22}\n")
+            # Like the registry: a status for the version path, the package
+            # document (with its version list) for the fallback.
+            code = status or ("200" if npm_available else "404")
+            listed = '{"versions":{"0.52.0":{}}}' if npm_available else '{"versions":{"0.51.0":{}}}'
+            curl.write_text(
+                "#!/bin/sh\n"
+                f"echo \"$@\" >> {calls}\n"
+                'case "$*" in *http_code*) printf %s ' + code + "; exit 0;; esac\n"
+                f"echo '{listed}'\n"
+            )
             node.chmod(0o755)
             curl.chmod(0o755)
             env = dict(os.environ, PATH=f"{tmp}:{os.environ.get('PATH', '')}",
@@ -200,6 +209,17 @@ class NpmDispatcherGate(unittest.TestCase):
         missing = self._run_wait(False)
         self.assertNotEqual(missing.returncode, 0)
         self.assertIn("Refusing to publish a dispatcher", missing.stdout)
+        self.assertIn("linux-x64@0.52.0 (HTTP 404)", missing.stdout)
+
+    def test_a_package_listed_in_its_document_counts_as_published(self):
+        """A 404 on the version path alone is not proof it is missing."""
+        listed = self._run_wait(True, status="404")
+        self.assertEqual(listed.returncode, 0, listed.stdout + listed.stderr)
+
+    def test_throttling_is_reported_as_throttling(self):
+        throttled = self._run_wait(False, status="429")
+        self.assertNotEqual(throttled.returncode, 0)
+        self.assertIn("(HTTP 429)", throttled.stdout)
 
 
 if __name__ == "__main__":

aria_code-0.85.0/tests/test_tool_step_output.py

@@ -0,0 +1,147 @@
+"""Each tool step reads as one block, as Codex shows it.
+
+Recorded on a real coding turn: the command header said "executing command
+shell tool"; two "✓ writing file (13910ms)" lines named no file and counted
+the person reading the diff; answered approval menus stayed on screen in
+full; `python3 -m unittest -v` showed no output (its report is on stderr),
+so the model pasted the results into its answer; code blocks were white
+slabs in a dark terminal.
+"""
+
+from __future__ import annotations
+
+import io
+import sys
+
+import pytest
+from rich.console import Console
+
+
+def _console(width=100):
+    return Console(file=io.StringIO(), width=width, color_system=None, highlight=False)
+
+
+def test_command_outcome_shows_the_tail_of_stdout_and_stderr():
+    from aria_code.apps.cli.tools.system_tools import print_command_outcome
+
+    console = _console()
+    report = "\n".join(f"test_{i} ... ok" for i in range(8)) + "\n" + "-" * 20 + "\nRan 8 tests in 0.001s\n\nOK"
+    print_command_outcome(console, 0, "", report, "/Users/someone/.aria/out/abc.log")
+    lines = console.file.getvalue().splitlines()
+    assert lines[0].strip() == "⎿  exit 0 · 12 lines"
+    assert "… +7 lines · full output saved: abc.log" in lines[1]
+    assert lines[-1].strip() == "OK" and len(lines) == 2 + 5
+    assert "/Users/someone" not in console.file.getvalue()
+
+
+def test_program_output_is_not_markup():
+    from aria_code.apps.cli.tools.system_tools import print_command_outcome
+
+    console = _console()
+    print_command_outcome(console, 1, "[bold]not bold[/bold] and a stray [/dim]", "")
+    text = console.file.getvalue()
+    assert "exit 1" in text and "[bold]not bold[/bold] and a stray [/dim]" in text
+
+
+@pytest.mark.parametrize("raw, hidden", [
+    ("curl -H 'Authorization: Bearer abc.def.ghi' https://x", "abc.def.ghi"),
+    ("git clone https://user:hunter2@github.com/o/r", "hunter2"),
+    ("OPENAI_API_KEY=sk-live1234567890abcd python run.py", "sk-live1234567890abcd"),
+    ("export GITHUB_TOKEN=ghp_abcdefghijklmnop", "ghp_abcdefghijklmnop"),
+    ("gemini --key AIzaSyA1234567890abcdefghijk", "AIzaSyA1234567890abcdefghijk"),
+])
+def test_commands_are_shown_with_secrets_masked(raw, hidden):
+    from aria_code.apps.cli.runtime_consumer import _redact_activity_text
+
+    shown = _redact_activity_text(raw, limit=200)
+    assert hidden not in shown and "***" in shown
+
+
+def test_the_step_header_names_the_command():
+    sys.argv = ["aria-code"]
+    import aria_code.aria_cli as cli
+
+    assert cli._format_tool_params("run_command", {"command": "python3 -m unittest -v"}) == "python3 -m unittest -v"
+    assert "s3cr3t" not in cli._format_tool_params("run_command", {"command": "mysql --password s3cr3t"})
+
+
+def _consumer(calls):
+    from aria_code.apps.cli.runtime_consumer import TerminalRuntimeEventConsumer
+
+    done = []
+    consumer = TerminalRuntimeEventConsumer.__new__(TerminalRuntimeEventConsumer)
+    consumer.tool_start_times = {"write_file": [100.0] * len(calls)}
+    consumer.tool_params = {"write_file": [{"path": path} for path in calls]}
+    consumer.print_tool_done = lambda tool, ms, success=True, summary="": done.append((summary, ms))
+    consumer.terminal = type("T", (), {"_transcript_log": []})()
+    consumer.tool_spinner = None
+    return consumer, done
+
+
+def test_the_done_line_names_its_target_and_shows_the_measured_time(monkeypatch):
+    """Two writes announced together, approved one after the other, reported
+    together: each line shows its own file and the time the runtime measured,
+    not the time since it was announced."""
+    from aria_code.apps.cli.runtime_consumer import with_measured_elapsed
+
+    monkeypatch.setattr("aria_code.apps.cli.runtime_consumer.time.time", lambda: 113.9)
+    consumer, done = _consumer(["src/fx.py", "test_fx.py"])
+    consumer.on_tool_result("write_file", with_measured_elapsed({"success": True}, 0.009))
+    consumer.on_tool_result("write_file", with_measured_elapsed({"success": True}, 0.012))
+    assert done == [("fx.py", 9), ("test_fx.py", 12)]
+
+
+def test_without_a_measurement_the_announcement_time_is_used(monkeypatch):
+    monkeypatch.setattr("aria_code.apps.cli.runtime_consumer.time.time", lambda: 100.25)
+    consumer, done = _consumer(["fx.py"])
+    consumer.on_tool_result("write_file", {"success": True})
+    assert done == [("fx.py", 250)]
+
+
+def test_the_gateway_passes_the_measured_time_on():
+    import inspect
+
+    from aria_code.runtime import gateway
+
+    assert "with_measured_elapsed" in inspect.getsource(gateway)
+
+
+def test_an_approval_collapses_to_what_was_approved(monkeypatch):
+    """Through aria_cli, which rebinds the approval function to its own
+    globals: a helper defined in tool_executor was not visible there, and the
+    first recorded run crashed every approval with a NameError."""
+    sys.argv = ["aria-code"]
+    import aria_code.aria_cli as cli
+
+    seen = {}
+
+    def fake_select(options, selected=0, title="", **kwargs):
+        seen.update(kwargs)
+        return 0
+
+    monkeypatch.setattr(cli, "_arrow_select", fake_select)
+    monkeypatch.setattr(cli, "HAS_RICH", False)
+    decision = cli._confirm_tool_execution_decision(
+        "write_file", {"path": "/tmp/w/fx.py", "content": "x = 1\n"}, config_policy="safe")
+    assert decision.approved and seen["collapse_to"] == "fx.py"
+    seen.clear()
+    cli._confirm_tool_execution_decision(
+        "run_command", {"command": "TOKEN=abc123 python3 -m unittest -v"}, config_policy="safe")
+    assert "abc123" not in seen.get("collapse_to", "") and "unittest" in seen.get("collapse_to", "")
+
+
+def test_the_answer_rules_say_not_to_paste_back_tool_output():
+    from aria_code.apps.cli.prompts.system_prompts import build_response_style_rule
+
+    assert "do not paste back" in build_response_style_rule("en")
+    assert "不要整段重贴" in build_response_style_rule("zh")
+
+
+def test_the_coding_prompt_asks_for_a_short_summary():
+    """A coding turn gets CODING_SYSTEM_PROMPT, not the answer-style rule: the
+    recorded turn pasted both files and the whole test run back."""
+    from aria_code.apps.cli.prompts.coding import CODING_SYSTEM_PROMPT
+
+    assert "## FINAL SUMMARY" in CODING_SYSTEM_PROMPT
+    assert "Do NOT paste file contents" in CODING_SYSTEM_PROMPT
+

{aria_code-0.83.0 → aria_code-0.85.0}/tests/test_ui_console_theme.py

@@ -30,10 +30,15 @@ def test_light_markdown_theme_uses_high_contrast_text():
 
 
 def test_make_markdown_uses_neutral_code_theme():
+    """Code keeps the terminal's background. "bw" was meant to be neutral, but
+    it paints #ffffff behind every block: white slabs in a dark terminal."""
+    from rich.syntax import Syntax
+
     md = make_markdown("`AAPL` and ```python\nprint('ok')\n```")
 
-    assert getattr(md, "code_theme", "") == "bw"
-    assert getattr(md, "inline_code_theme", "") == "bw"
+    assert md.code_theme in ("ansi_dark", "ansi_light")
+    assert md.inline_code_theme == md.code_theme
+    assert not Syntax.get_theme(md.code_theme).get_background_style()
 
 
 def test_rich_theme_defaults_apple_terminal_to_light_without_color_metadata():