Feat/sandbox cli v10 (#138)

Author: innsd

agentkit/toolkit/cli/sandbox/README.md

@@ -214,7 +214,7 @@ Provider defaults:
 
 | Provider | Default model | Model base URL |
 | --- | --- | --- |
-| `model_square` | `deepseek-v4-flash-260425` | `https://ark.cn-beijing.volces.com/api/v3` |
+| `model_square` | `glm-5-2-260617` | `https://ark.cn-beijing.volces.com/api/v3` |
 | `coding_plan` | `deepseek-v4-flash` | `https://ark.cn-beijing.volces.com/api/coding/v3` |
 | `agent_plan` | `deepseek-v4-flash` | `https://ark.cn-beijing.volces.com/api/plan/v3` |
 | `byteplus_model_square` | `deepseek-v4-flash-260425` | `https://ark.ap-southeast.bytepluses.com/api/v3` |

agentkit/toolkit/cli/sandbox/model_config.py

@@ -83,8 +83,12 @@ class ModelProviderConfig:
     ModelProviderType.MODEL_SQUARE.value: ModelProviderConfig(
         model_base_url="https://ark.cn-beijing.volces.com/api/v3",
         anthropic_base_url="https://ark.cn-beijing.volces.com/api/compatible",
-        default_model_name="deepseek-v4-flash-260425",
+        default_model_name="glm-5-2-260617",
         models={
+            "glm-5-2-260617": ModelSpec(
+                supports_reasoning_summaries=True,
+                context_window=DEFAULT_MODEL_CONTEXT_WINDOW,
+            ),
             "doubao-seed-2-0-pro-260215": ModelSpec(
                 supports_reasoning_summaries=True,
                 context_window=LIMITED_MODEL_CONTEXT_WINDOW,
@@ -435,7 +439,11 @@ def build_codex_config_toml(
             'wire_api = "responses"',
             # OAuth (ChatGPT login) providers carry no API key - codex uses the injected
             # auth.json via `requires_openai_auth`; all others use the CODEX_API_KEY env.
-            ("requires_openai_auth = true" if requires_openai_auth else 'env_key = "CODEX_API_KEY"'),
+            (
+                "requires_openai_auth = true"
+                if requires_openai_auth
+                else 'env_key = "CODEX_API_KEY"'
+            ),
             "",
             "[tui]",
             "show_tooltips = false",
@@ -552,7 +560,7 @@ def build_codex_hot_update_command() -> str:
   CODEX_BASE_URL="$REQ_CODEX_BASE_URL"
 fi
 
-CODEX_MODEL="${CODEX_MODEL:-deepseek-v4-flash-260425}"
+CODEX_MODEL="${CODEX_MODEL:-glm-5-2-260617}"
 CODEX_BASE_URL="${CODEX_BASE_URL:-https://ark.cn-beijing.volces.com/api/v3}"
 
 : >"$ENV_FILE"
@@ -653,8 +661,10 @@ def _build_model_catalog_item(model_name: str, spec: ModelSpec) -> dict:
 
 def infer_model_spec(model_name: str) -> ModelSpec:
     normalized_model_name = model_name.strip().lower()
-    if normalized_model_name == "glm-5.2" or normalized_model_name.startswith(
-        "deepseek-v4"
+    if (
+        normalized_model_name == "glm-5.2"
+        or normalized_model_name.startswith("glm-5-")
+        or normalized_model_name.startswith("deepseek-v4")
     ):
         context_window = DEFAULT_MODEL_CONTEXT_WINDOW
     else:

agentkit/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-VERSION = "0.7.10"
+VERSION = "0.7.11"

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "agentkit-sdk-python"
-version = "0.7.10"
+version = "0.7.11"
 description = "Python SDK for transforming any AI agent into a production-ready application. Framework-agnostic primitives for runtime, memory, authentication, and tools with volcengine-managed infrastructure."
 readme = "README.md"
 requires-python = ">=3.10"

tests/toolkit/cli/test_cli_create_tool.py

@@ -912,8 +912,7 @@ def test_create_command_uses_sandbox_yaml_when_image_options_are_omitted(
     sandbox_yaml_path = cli_create._get_sandbox_yaml_path()
     sandbox_yaml_path.parent.mkdir(parents=True, exist_ok=True)
     sandbox_yaml_path.write_text(
-        "tool_type: Private\n"
-        "image_url: registry.example.com/custom-image:from-yaml\n",
+        "tool_type: Private\nimage_url: registry.example.com/custom-image:from-yaml\n",
         encoding="utf-8",
     )
 
@@ -939,8 +938,7 @@ def test_create_command_cli_tool_type_disables_sandbox_yaml_defaults(
     sandbox_yaml_path = cli_create._get_sandbox_yaml_path()
     sandbox_yaml_path.parent.mkdir(parents=True, exist_ok=True)
     sandbox_yaml_path.write_text(
-        "tool_type: Private\n"
-        "image_url: registry.example.com/custom-image:from-yaml\n",
+        "tool_type: Private\nimage_url: registry.example.com/custom-image:from-yaml\n",
         encoding="utf-8",
     )
 
@@ -969,8 +967,7 @@ def test_create_command_cli_image_url_disables_sandbox_yaml_defaults(
     sandbox_yaml_path = cli_create._get_sandbox_yaml_path()
     sandbox_yaml_path.parent.mkdir(parents=True, exist_ok=True)
     sandbox_yaml_path.write_text(
-        "tool_type: Private\n"
-        "image_url: registry.example.com/custom-image:from-yaml\n",
+        "tool_type: Private\nimage_url: registry.example.com/custom-image:from-yaml\n",
         encoding="utf-8",
     )
 
@@ -1085,7 +1082,9 @@ def test_build_create_tool_request_adds_code_env_config_envs(monkeypatch):
     assert models[0]["slug"] == "deepseek-v4-pro-260425"
     assert models[0]["display_name"] == "deepseek-v4-pro-260425"
     assert "deepseek-v4-flash-260425" in [model["slug"] for model in models]
+    assert "glm-5-2-260617" in [model["slug"] for model in models]
     models_by_slug = {model["slug"]: model for model in models}
+    assert models_by_slug["glm-5-2-260617"]["max_context_window"] == 1000000
     assert "doubao-seed-2-0-pro-260215" in models_by_slug
     assert models_by_slug["doubao-seed-2-0-pro-260215"]["max_context_window"] == 200000
     assert models[0]["truncation_policy"] == {"mode": "tokens", "limit": 10000}
@@ -1126,6 +1125,7 @@ def test_build_create_tool_request_uses_model_provider(monkeypatch):
     models = {model["slug"]: model for model in catalog["models"]}
     assert "deepseek-v4-flash" in models
     assert "deepseek-v4-flash-260425" not in models
+    assert "glm-5-2-260617" not in models
     assert "glm-5.2" not in models
 
 
@@ -1319,6 +1319,7 @@ def test_build_codex_model_catalog_infers_custom_model_context_window():
 
     expected_context_windows = {
         "deepseek-v4-flash-260428": 1000000,
+        "glm-5-2-260617": 1000000,
         "glm-5.2": 1000000,
         "doubao-seed-2-0-code-preview-260215": 200000,
         "glm-4-7-251222": 200000,
@@ -1405,9 +1406,9 @@ def test_create_command_accepts_model_base_url_without_model_name(monkeypatch):
     envs = {item.key: item.value for item in _FakeToolsClient.last_request.envs}
     assert envs["AGENTKIT_SANDBOX_MODEL_PROVIDER"] == "custom_provider"
     assert envs["CODEX_BASE_URL"] == "https://models.example.com"
-    assert envs["CODEX_MODEL"] == "deepseek-v4-flash-260425"
+    assert envs["CODEX_MODEL"] == "glm-5-2-260617"
     assert 'model_provider = "custom_provider"' in envs["CODEX_CONFIG_TOML"]
-    assert 'model = "deepseek-v4-flash-260425"' in envs["CODEX_CONFIG_TOML"]
+    assert 'model = "glm-5-2-260617"' in envs["CODEX_CONFIG_TOML"]
     assert 'base_url = "https://models.example.com"' in envs["CODEX_CONFIG_TOML"]
     assert "model_catalog_json" not in envs["CODEX_CONFIG_TOML"]
     assert "CODEX_MODEL_CATALOG_JSON" not in envs
@@ -1525,9 +1526,9 @@ def test_build_create_tool_request_adds_default_model_base_url(monkeypatch):
 
     assert [(item.key, item.value) for item in request.envs] == [
         ("AGENTKIT_SANDBOX_MODEL_PROVIDER", "model_square"),
-        ("OPENCODE_MODEL", "deepseek-v4-flash-260425"),
-        ("CODEX_MODEL", "deepseek-v4-flash-260425"),
-        ("ANTHROPIC_MODEL", "deepseek-v4-flash-260425"),
+        ("OPENCODE_MODEL", "glm-5-2-260617"),
+        ("CODEX_MODEL", "glm-5-2-260617"),
+        ("ANTHROPIC_MODEL", "glm-5-2-260617"),
         ("OPENCODE_BASE_URL", "https://ark.cn-beijing.volces.com/api/v3"),
         ("CODEX_BASE_URL", "https://ark.cn-beijing.volces.com/api/v3"),
         ("MODEL_BASE_URL", "https://ark.cn-beijing.volces.com/api/v3"),

tests/toolkit/cli/test_cli_sandbox.py

@@ -1138,6 +1138,15 @@ def test_build_codex_hot_update_env_allows_explicit_empty_api_key() -> None:
     }
 
 
+def test_build_codex_hot_update_command_uses_model_square_default() -> None:
+    from agentkit.toolkit.cli.sandbox.model_config import build_codex_hot_update_command
+
+    command = build_codex_hot_update_command()
+
+    assert 'CODEX_MODEL="${CODEX_MODEL:-glm-5-2-260617}"' in command
+    assert 'CODEX_MODEL="${CODEX_MODEL:-deepseek-v4-flash-260425}"' not in command
+
+
 def test_build_model_envs_uses_model_base_url_and_emits_codex_config(
     monkeypatch,
 ) -> None:
@@ -1299,9 +1308,9 @@ def test_build_model_envs_allows_model_base_url_without_model_name(monkeypatch)
 
     assert [(item.key, item.value) for item in envs] == [
         ("AGENTKIT_SANDBOX_MODEL_PROVIDER", "custom_provider"),
-        ("OPENCODE_MODEL", "deepseek-v4-flash-260425"),
-        ("CODEX_MODEL", "deepseek-v4-flash-260425"),
-        ("ANTHROPIC_MODEL", "deepseek-v4-flash-260425"),
+        ("OPENCODE_MODEL", "glm-5-2-260617"),
+        ("CODEX_MODEL", "glm-5-2-260617"),
+        ("ANTHROPIC_MODEL", "glm-5-2-260617"),
         ("OPENCODE_BASE_URL", "https://models.example.com/v1"),
         ("CODEX_BASE_URL", "https://models.example.com/v1"),
         ("MODEL_BASE_URL", "https://models.example.com/v1"),
@@ -1312,7 +1321,7 @@ def test_build_model_envs_allows_model_base_url_without_model_name(monkeypatch)
         ),
     ]
     assert 'model_provider = "custom_provider"' in envs[8].value
-    assert 'model = "deepseek-v4-flash-260425"' in envs[8].value
+    assert 'model = "glm-5-2-260617"' in envs[8].value
     assert 'base_url = "https://models.example.com/v1"' in envs[8].value
     assert "model_catalog_json" not in envs[8].value
 
@@ -5872,6 +5881,7 @@ def test_cli_exec_model_provider_sets_default_model_and_codex_config(
     models = {model["slug"]: model for model in catalog["models"]}
     assert "deepseek-v4-flash" in models
     assert "deepseek-v4-flash-260425" not in models
+    assert "glm-5-2-260617" not in models
     assert "glm-5.2" not in models