gdmcode 0.1.5__tar.gz → 0.1.6__tar.gz

{gdmcode-0.1.5 → gdmcode-0.1.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gdmcode
-Version: 0.1.5
+Version: 0.1.6
 Summary: gdm: AI coding agent for professional developers
 Project-URL: Homepage, https://github.com/guidegdm/gdmcode
 Project-URL: Repository, https://github.com/guidegdm/gdmcode

gdmcode-0.1.6/gdmcode/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.1.6"

{gdmcode-0.1.5 → gdmcode-0.1.6}/gdmcode/cli.py

@@ -348,6 +348,34 @@ def cmd_health(
         except Exception as exc:  # noqa: BLE001
             return CheckResult(f"api ({cfg.provider})", "fail", detail=str(exc)[:60])
 
+    def _check_completion() -> CheckResult:
+        t0 = time.monotonic()
+        try:
+            from gdmcode.models.client import GdmClient
+
+            model_def = get_model(ModelTier.CODER, cfg.provider)
+            proxy_active = bool(
+                getattr(cfg, "proxy_enabled", False)
+                and getattr(cfg, "proxy_url", "")
+                and getattr(cfg, "proxy_token", "")
+            )
+            client = (
+                GdmClient.for_proxy(getattr(cfg, "proxy_url"), getattr(cfg, "proxy_token"))
+                if proxy_active
+                else GdmClient(cfg)
+            )
+            text = client.complete_text(
+                "Reply with exactly: OK",
+                model=model_def.id,
+                max_tokens=8,
+                system="You are a health check. Reply only with OK.",
+            )
+            status = "ok" if text.strip() else "fail"
+            detail = f"{model_def.id}: {text.strip()[:30] or 'empty response'}"
+            return CheckResult("llm completion", status, latency_ms=int((time.monotonic() - t0) * 1000), detail=detail, priority="P0")
+        except Exception as exc:  # noqa: BLE001
+            return CheckResult("llm completion", "fail", latency_ms=int((time.monotonic() - t0) * 1000), detail=str(exc)[:120], priority="P0")
+
     def _check_tools() -> CheckResult:
         t0 = time.monotonic()
         try:
@@ -401,10 +429,11 @@ def cmd_health(
         return results
 
     # Run scalar checks in parallel; sysdeps is fast so fine in its own slot
-    with ThreadPoolExecutor(max_workers=6) as executor:
+    with ThreadPoolExecutor(max_workers=7) as executor:
         f_db     = executor.submit(_check_db)
         f_daemon = executor.submit(_check_daemon)
         f_api    = executor.submit(_check_api)
+        f_completion = executor.submit(_check_completion)
         f_tools  = executor.submit(_check_tools)
         f_budget = executor.submit(_check_budget)
         f_sys    = executor.submit(_check_sysdeps)
@@ -413,6 +442,7 @@ def cmd_health(
         f_db.result(),
         f_daemon.result(),
         f_api.result(),
+        f_completion.result(),
         f_tools.result(),
         f_budget.result(),
         *f_sys.result(),
@@ -1201,13 +1231,17 @@ def _run_one_shot(cfg: object, prompt: str, *, yes: bool, model_override: str |
     """
     from rich.status import Status
     from gdmcode.agent.loop import EventType
-    from gdmcode.repl import _format_llm_error
+    from gdmcode.repl import _format_llm_error, _is_connection_probe
 
     if not _has_model_connection(cfg):
         console.print("[red]No model connection configured.[/red]")
         console.print("Run [bold]gdm login grok[/bold], [bold]gdm login gemini[/bold], or start [bold]gdm[/bold] and use [bold]/proxy token[/bold] then [bold]/proxy on[/bold].")
         raise typer.Exit(1)
 
+    if _is_connection_probe(prompt):
+        _run_connection_probe_once(cfg, prompt, model_override=model_override)
+        return
+
     try:
         loop, cost_tracker, _db = _setup_agent(cfg, yes=yes, model_override=model_override)
     except Exception as exc:  # noqa: BLE001
@@ -1271,7 +1305,12 @@ def _run_one_shot(cfg: object, prompt: str, *, yes: bool, model_override: str |
                 console.print(event.content)
             elif event.type == EventType.ERROR:
                 status.stop()
-                console.print(_format_llm_error(event.content))
+                content = str(event.content or "")
+                if content.lower().startswith("max turns"):
+                    console.print(f"[yellow]Agent stopped before completion:[/yellow] {content}")
+                    console.print("[dim]Increase --max-turns or simplify the prompt.[/dim]")
+                else:
+                    console.print(_format_llm_error(event.content))
                 had_error = True
             elif event.type == EventType.TOOL_CALL:
                 status.stop()
@@ -1301,6 +1340,45 @@ def _run_one_shot(cfg: object, prompt: str, *, yes: bool, model_override: str |
         raise typer.Exit(1)
 
 
+def _run_connection_probe_once(cfg: object, prompt: str, *, model_override: str | None) -> None:
+    """Run a one-shot no-tools LLM connection probe."""
+    from rich.status import Status
+    from gdmcode.models.client import GdmClient
+    from gdmcode.models.definitions import ModelTier, get_model
+    from gdmcode.repl import _format_llm_error
+
+    status = Status("[cyan]Checking LLM connection...[/cyan]", console=console, spinner="dots")
+    status.start()
+    try:
+        tier = model_override or ModelTier.CODER
+        model_def = get_model(tier, cfg.provider)  # type: ignore[union-attr]
+        proxy_active = bool(
+            getattr(cfg, "proxy_enabled", False)
+            and getattr(cfg, "proxy_url", "")
+            and getattr(cfg, "proxy_token", "")
+        )
+        client = (
+            GdmClient.for_proxy(getattr(cfg, "proxy_url"), getattr(cfg, "proxy_token"))
+            if proxy_active
+            else GdmClient(cfg)  # type: ignore[arg-type]
+        )
+        response = client.complete_text(
+            prompt,
+            model=model_def.id,
+            max_tokens=80,
+            system=(
+                "You are gdm code's CLI connection check. Reply in one short, friendly "
+                "sentence. Confirm the LLM is connected. Do not inspect files or call tools."
+            ),
+        ).strip()
+        status.stop()
+        console.print(response or "[green]LLM connected.[/green]")
+    except Exception as exc:  # noqa: BLE001
+        status.stop()
+        console.print(_format_llm_error(exc))
+        raise typer.Exit(1) from exc
+
+
 # ---------------------------------------------------------------------------
 # Helpers
 # ---------------------------------------------------------------------------

{gdmcode-0.1.5 → gdmcode-0.1.6}/gdmcode/integrations/mcp_server.py

@@ -51,7 +51,7 @@ _INTERNAL_ERROR   = -32603
 class MCPServer:
     """Minimal MCP server with stdio transport."""
 
-    def __init__(self, name: str = "gdmcode", version: str = "0.1.5") -> None:
+    def __init__(self, name: str = "gdmcode", version: str = "0.1.6") -> None:
         self._name = name
         self._version = version
         self._tools: dict[str, MCPTool] = {}

{gdmcode-0.1.5 → gdmcode-0.1.6}/gdmcode/models/definitions.py

@@ -201,14 +201,14 @@ GEMINI_SCOUT = ModelDef(
 )
 
 GEMINI_CODER = ModelDef(
-    id="gemini-2.5-flash-8b",
+    id="gemini-2.5-flash",
     tier=ModelTier.CODER,
     provider=Provider.GEMINI,
     input_per_m=0.075,
     output_per_m=0.30,
     cached_per_m=0.0,
     context_window=1_000_000,
-    notes="Ultra-cheap bulk: file indexing, convention extraction, boilerplate edits.",
+    notes="Cheap bulk: file indexing, convention extraction, boilerplate edits.",
 )
 
 GEMINI_THINKER = ModelDef(
@@ -456,4 +456,3 @@ def get_capability(tier: str, provider: str, capability: str) -> bool:
         return bool(getattr(m, capability, False))
     except KeyError:
         return False
-

{gdmcode-0.1.5 → gdmcode-0.1.6}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "gdmcode"
-version = "0.1.5"
+version = "0.1.6"
 description = "gdm: AI coding agent for professional developers"
 readme = "README.md"
 license = { text = "MIT" }

{gdmcode-0.1.5 → gdmcode-0.1.6}/tests/test_cli_smoke.py

@@ -89,8 +89,27 @@ class TestCliSmoke:
 
         response = SimpleNamespace(status_code=200)
         with patch("gdmcode.cli.httpx.get", return_value=response):
-            result = runner.invoke(app, ["health"])
+            with patch("gdmcode.models.client.GdmClient.complete_text", return_value="OK"):
+                result = runner.invoke(app, ["health"])
 
         assert "tools" in result.output
         assert "registered" in result.output
         assert "probe ok" in result.output
+        assert "llm completion" in result.output
+
+    def test_one_shot_greeting_uses_connection_probe(self, tmp_path, monkeypatch) -> None:
+        import gdmcode.cli as cli
+
+        cfg = SimpleNamespace(provider="gemini", api_key="key", project_root=tmp_path)
+        monkeypatch.setattr(cli, "load_config", lambda require_credentials=False: cfg)
+        monkeypatch.setattr(cli, "_print_header", lambda _cfg: None)
+        monkeypatch.setattr(cli, "_has_model_connection", lambda _cfg: True)
+        probe = MagicMock()
+        monkeypatch.setattr(cli, "_run_connection_probe_once", probe)
+
+        result = runner.invoke(app, ["code", "--prompt", "hey"])
+
+        assert result.exit_code == 0
+        probe.assert_called_once()
+        assert probe.call_args.args[1] == "hey"
+        assert probe.call_args.kwargs == {"model_override": None}

{gdmcode-0.1.5 → gdmcode-0.1.6}/tests/test_health.py

@@ -22,6 +22,9 @@ def _make_cfg(**kwargs):
     cfg.provider = kwargs.get("provider", "grok")
     cfg.api_key = kwargs.get("api_key", "sk-test")
     cfg.project_root = Path("/tmp/gdm-test")
+    cfg.proxy_enabled = kwargs.get("proxy_enabled", False)
+    cfg.proxy_url = kwargs.get("proxy_url", "")
+    cfg.proxy_token = kwargs.get("proxy_token", None)
     return cfg
 
 
@@ -63,6 +66,7 @@ def _run_health(
     daemon = _make_daemon(running=daemon_running)
     mock_registry = MagicMock()
     mock_registry.all_tools.return_value = [MagicMock()] * tools_count
+    mock_registry.call.return_value = MagicMock(ok=True, output="ok", error=None)
 
     mock_response = MagicMock()
     mock_response.status_code = api_status
@@ -75,6 +79,7 @@ def _run_health(
         patch("gdmcode.cli.shutil.which", side_effect=lambda cmd: which_map.get(cmd)),
         patch("gdmcode.cli.httpx.get",
               side_effect=api_exc if api_exc else lambda *a, **kw: mock_response),
+        patch("gdmcode.models.client.GdmClient.complete_text", return_value="OK"),
     ):
         runner = CliRunner()
         result = runner.invoke(app, ["health", *args])
@@ -138,6 +143,7 @@ def test_api_probe_sends_auth_header():
     daemon = _make_daemon()
     mock_registry = MagicMock()
     mock_registry.all_tools.return_value = [MagicMock()]
+    mock_registry.call.return_value = MagicMock(ok=True, output="ok", error=None)
 
     with (
         patch("gdmcode.cli.load_config", return_value=cfg),
@@ -146,6 +152,7 @@ def test_api_probe_sends_auth_header():
         patch("gdmcode.cli.REGISTRY", mock_registry),
         patch("gdmcode.cli.shutil.which", return_value="/usr/bin/git"),
         patch("gdmcode.cli.httpx.get", side_effect=fake_get),
+        patch("gdmcode.models.client.GdmClient.complete_text", return_value="OK"),
     ):
         runner = CliRunner()
         runner.invoke(app, ["health"])
@@ -280,6 +287,7 @@ def test_parallel_execution_completes():
     daemon = _make_daemon()
     mock_registry = MagicMock()
     mock_registry.all_tools.return_value = [MagicMock()]
+    mock_registry.call.return_value = MagicMock(ok=True, output="ok", error=None)
 
     def slow_api(*a, **kw):
         import time as _t
@@ -295,6 +303,7 @@ def test_parallel_execution_completes():
         patch("gdmcode.cli.REGISTRY", mock_registry),
         patch("gdmcode.cli.shutil.which", return_value="/bin/git"),
         patch("gdmcode.cli.httpx.get", side_effect=slow_api),
+        patch("gdmcode.models.client.GdmClient.complete_text", return_value="OK"),
     ):
         runner = CliRunner()
         t0 = time.monotonic()

gdmcode-0.1.5/gdmcode/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.1.5"