google-agents-cli 1.2.1__tar.gz → 1.3.0__tar.gz

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: google-agents-cli
-Version: 1.2.1
+Version: 1.3.0
 Summary: Agents CLI — Agent Development Lifecycle toolchain for ADK
 Author-email: Google LLC <agents-cli-dev-pypi@google.com>
 Requires-Python: >=3.11
@@ -19,7 +19,7 @@ Requires-Dist: google-auth[pyopenssl]>=2.47.0
 Requires-Dist: rich>=13.7.0
 Requires-Dist: backoff>=2.2.1
 Requires-Dist: requests>=2.31.0
-Requires-Dist: a2a-sdk~=0.3.22
+Requires-Dist: a2a-sdk>=1.0,<2
 Requires-Dist: tomlkit>=0.13.0,<1
 Requires-Dist: psutil>=5.9.0
 Requires-Dist: python-dotenv>=1.0.0

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "google-agents-cli"
-version = "1.2.1"
+version = "1.3.0"
 description = "Agents CLI — Agent Development Lifecycle toolchain for ADK"
 readme = "README.md"
 license = "Apache-2.0"
@@ -22,7 +22,7 @@ dependencies = [
   "rich>=13.7.0",
   "backoff>=2.2.1",
   "requests>=2.31.0",
-  "a2a-sdk~=0.3.22",
+  "a2a-sdk>=1.0,<2",
   "tomlkit>=0.13.0,<1",
   "psutil>=5.9.0",
   "python-dotenv>=1.0.0",

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/skills/google-agents-cli-adk-code/SKILL.md

@@ -12,7 +12,7 @@ description: >
 metadata:
   author: Google
   license: Apache-2.0
-  version: 1.2.1
+  version: 1.3.0
   requires:
     bins:
       - agents-cli

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/skills/google-agents-cli-deploy/SKILL.md

@@ -12,7 +12,7 @@ description: >
 metadata:
   author: Google
   license: Apache-2.0
-  version: 1.2.1
+  version: 1.3.0
   requires:
     bins:
       - agents-cli

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/skills/google-agents-cli-eval/SKILL.md

@@ -11,7 +11,7 @@ description: >
 metadata:
   author: Google
   license: Apache-2.0
-  version: 1.2.1
+  version: 1.3.0
   requires:
     bins:
       - agents-cli
@@ -131,12 +131,17 @@ All `agents-cli eval` subcommands support `--help` for the authoritative flag li
 
 Runs an agent over an evaluation dataset and writes traces to disk.
 
+By default, runs the agent in a local HTTP server (launches the project's `fast_api_app.py` if it exists, or falls back to `adk api_server`) and sends each evaluation case over HTTP. You can generate traces from an already-running agent by passing its HTTP endpoint and app name to `--url` and `--app-name`.
+
 ```bash
 # Basic — uses tests/eval/datasets/, writes to artifacts/traces/
 agents-cli eval generate
 
 # Advanced — custom dataset and output dir
 agents-cli eval generate --dataset tests/eval/datasets/custom.json -o ./custom_traces/
+
+# Against a deployed agent (or one you started manually)
+agents-cli eval generate --url https://my-agent.run.app --app-name app
 ```
 
 ### `eval grade`
@@ -351,10 +356,6 @@ app = App(root_agent=root_agent, name="app")
 app = App(root_agent=root_agent, name="flight_booking_assistant")
 ```
 
-### Cross-session memory can't be tested in eval
-
-Each eval case runs in its own fresh in-memory session (`eval generate` creates a new `InMemorySessionService` and session id per case). Multi-turn *within* a case works via `agent_data.turns`, but behavior that depends on a *separate prior session* — e.g. Memory Bank recall across sessions — can't be exercised by eval. Validate cross-session continuity with pytest integration tests instead.
-
 ### Vertex eval region
 
 `eval grade` and `eval submit` **default to the `global` endpoint** — they don't inherit the manifest `region` (the eval services support only a subset of regions), and `eval analyze` is `global`-only. Override these per run with `--region <REGION>` (e.g. data residency); the service rejects an unsupported one:
@@ -363,7 +364,7 @@ Each eval case runs in its own fresh in-memory session (`eval generate` creates
 400 FAILED_PRECONDITION: Unsupported region for Vertex Evaluation Service: <region>
 ```
 
-`eval generate` and `eval dataset synthesize` run your agent locally, so they honor the agent's own `.env` — notably `GOOGLE_CLOUD_LOCATION`, which selects the model endpoint **when the agent uses Vertex AI** (`GOOGLE_GENAI_USE_VERTEXAI=true`); it's unused with a `GEMINI_API_KEY` (AI Studio). They take **no** `--region` and never override your `.env` with the manifest `region`; change the model region by editing `.env`. One caveat for `synthesize`: its scenario-generation step is a **server-side** eval call at `GOOGLE_CLOUD_LOCATION`, so keep that an eval-supported region (`global` by default) even though the agent itself could run elsewhere.
+`eval generate` (without the `--url` flag) and `eval dataset synthesize` run your agent locally, so they honor the agent's own `.env` — notably `GOOGLE_CLOUD_LOCATION`, which selects the model endpoint **when the agent uses Vertex AI** (`GOOGLE_GENAI_USE_VERTEXAI=true`); it's unused with a `GEMINI_API_KEY` (AI Studio). They take **no** `--region` and never override your `.env` with the manifest `region`; change the model region by editing `.env`. One caveat for `synthesize`: its scenario-generation step is a **server-side** eval call at `GOOGLE_CLOUD_LOCATION`, so keep that an eval-supported region (`global` by default) even though the agent itself could run elsewhere.
 
 **No eval region fits your data-residency rules?** Fall back to **local custom metrics** — a `custom_metrics` entry with a `custom_function` (`execution: local`, the default) grades in-process with no GCP region required. You lose the managed built-in metrics, but your `custom_function` can still call an LLM judge in a compliant region itself — so LLM-as-judge grading stays available anywhere.
 

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/skills/google-agents-cli-observability/SKILL.md

@@ -13,7 +13,7 @@ description: >
 metadata:
   author: Google
   license: Apache-2.0
-  version: 1.2.1
+  version: 1.3.0
   requires:
     bins:
       - agents-cli

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/skills/google-agents-cli-publish/SKILL.md

@@ -15,7 +15,7 @@ description: >
 metadata:
   author: Google
   license: Apache-2.0
-  version: 1.2.1
+  version: 1.3.0
   requires:
     bins:
       - agents-cli

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/skills/google-agents-cli-scaffold/SKILL.md

@@ -12,7 +12,7 @@ description: >
 metadata:
   author: Google
   license: Apache-2.0
-  version: 1.2.1
+  version: 1.3.0
   requires:
     bins:
       - agents-cli

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/skills/google-agents-cli-workflow/SKILL.md

@@ -12,7 +12,7 @@ description: >
 metadata:
   author: Google
   license: Apache-2.0
-  version: 1.2.1
+  version: 1.3.0
   requires:
     bins:
       - agents-cli
@@ -26,8 +26,8 @@ metadata:
 > **Before writing agent code, make sure a scaffolded project exists (see Phase 2).** Skipping scaffolding loses eval boilerplate, CI/CD config, and project conventions.
 
 
-> Requires: google-agents-cli ~= 1.2.1
-> If version is behind, run: uv tool install "google-agents-cli~=1.2.1"
+> Requires: google-agents-cli ~= 1.3.0
+> If version is behind, run: uv tool install "google-agents-cli~=1.3.0"
 
 > Check version: agents-cli info
 > [Install uv](https://docs.astral.sh/uv/getting-started/installation/index.md) first if needed.

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/src/google/agents/cli/_tools.py

@@ -18,8 +18,6 @@ import os
 import re
 import shlex
 import shutil
-import subprocess
-import threading
 from functools import cache
 from pathlib import Path
 
@@ -149,125 +147,22 @@ def require_tool(name: str, install_hint: str = "") -> str:
     return path
 
 
-def run_npx_skills(args: list[str], spinner_msg: str) -> list[str]:
+def run_npx_skills(args: list[str], spinner_msg: str):
     """Run an npx skills command, streaming output in real-time.
 
     Always starts with ``["npx", "-y", SKILLS_NPX_PACKAGE]`` and appends
     the additional ``args`` provided.
-    Streams stdout/stderr line-by-line, filtering npm/npx boilerplate.
-    All non-noise lines are printed immediately. Only concise summary
-    lines (e.g. "Installed 6 skills", "Found 6 skills") are collected
-    for the end summary.
-
-    Failure lines (e.g. "✗ Failed to update <skill>") are re-colored red and
-    counted; if any are observed the command raises even when npx itself
-    exits 0, since `npx skills update` reports per-skill failures via stdout
-    but still exits cleanly.
-
-    Returns:
-        A list of summary-worthy lines (short, no decorative content).
+
     Raises:
-        click.ClickException: If the npx process exits non-zero, or if any
-            per-skill failures were observed in the streamed output.
+        click.ClickException: If the npx process exits non-zero.
     """
-    from google.agents.cli._runner import popen_resolved
+    from google.agents.cli._runner import run_resolved
     from google.agents.cli._skills_check import SKILLS_NPX_PACKAGE
 
     full_args = ["npx", "-y", SKILLS_NPX_PACKAGE, *args]
     click.secho(f"  \u25b8 {shlex.join(full_args)}", fg="cyan", dim=True)
 
-    # Disable color in the child process. We re-emit each captured line via
-    # click.echo, so any ANSI sequences from the child would either render as
-    # literal escape codes or, on Windows PowerShell, bleed across line
-    # boundaries — causing error lines to inherit the previous line's color
-    # (e.g. green progress markers making "Failed to update" appear green).
-    # See b/525049570.
-    child_env = {**os.environ, "NO_COLOR": "1", "FORCE_COLOR": "0"}
-
-    summary_lines: list[str] = []
-    failure_lines: list[str] = []
-    # Pipe stderr and drain it concurrently in a background thread to avoid
-    # pipe-buffer deadlock.
-    proc = popen_resolved(
-        full_args,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.PIPE,
-        text=True,
-        encoding="utf-8",
-        errors="replace",
-        env=child_env,
-    )
-
-    stderr_capture = {"data": ""}
-
-    def _drain_stderr() -> None:
-        try:
-            stderr_capture["data"] = proc.stderr.read() if proc.stderr else ""
-        except Exception:
-            stderr_capture["data"] = ""
-
-    stderr_thread = threading.Thread(target=_drain_stderr, daemon=True)
-    stderr_thread.start()
-
-    # Read stdout line-by-line
-    assert proc.stdout is not None
-    for line in proc.stdout:
-        # Defensively strip any leaked ANSI escape sequences so we control how
-        # the line renders.
-        stripped = _ANSI_RE.sub("", line).strip()
-        if not stripped:
-            continue
-        # Skip npx download/cache noise
-        if stripped.startswith("npm ") or stripped.startswith("npx:"):
-            continue
-        # Skip ASCII art banners (block characters)
-        if any(ch in stripped for ch in "█╗╔║╚╝"):
-            continue
-        # Strip leading box-drawing / bullet / status-glyph prefixes to extract
-        # text. We intentionally also strip "✗" so we can re-emit the line with
-        # our own (correct) red coloring below.
-        clean = stripped.lstrip("┌┐└┘├┤│◇●◆✓✗─╮╯ ")
-        if not clean:
-            continue
-        # Skip per-agent detail lines
-        if clean.startswith(("universal:", "symlink", "overwrites:")):
-            continue
-        # Skip purely decorative headers (e.g. "Installation Summary ────")
-        if "──" in clean:
-            continue
-
-        # Detect per-skill failures. npx skills reports these as
-        # "✗ Failed to update <skill>" followed by a "Failed to update N
-        # skill(s)" summary line. Either form is a hard failure for us.
-        is_failure = clean.startswith("Failed ") or "Failed to update" in clean
-        if is_failure:
-            failure_lines.append(clean)
-            click.secho(f"  ✗ {clean}", fg="red")
-        else:
-            click.echo(f"  {clean}")
-
-        # Collect concise summary-worthy lines for the recap
-        if len(clean) < 80 and clean.startswith(
-            ("Installed", "Found", "Done", "Removed", "Updated")
-        ):
-            summary_lines.append(clean)
-
-    proc.wait()
-    stderr_thread.join()
-
-    if proc.returncode != 0:
-        stderr = stderr_capture["data"]
-        click.secho("  Error running npx skills:", fg="red")
-        if stderr.strip():
-            for line in stderr.strip().splitlines():
-                click.echo(f"  {line}")
-        raise click.ClickException("npx skills failed")
-
-    if failure_lines:
-        # npx skills exits 0 even when individual skill operations fail, so we
-        # surface those failures as a non-zero exit ourselves.
-        raise click.ClickException(
-            f"npx skills reported {len(failure_lines)} failure(s); see output above."
-        )
-
-    return summary_lines
+    try:
+        run_resolved(full_args, check=True, encoding="utf-8", errors="replace")
+    except Exception as e:
+        raise click.ClickException("Error running npx skills") from e

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/src/google/agents/cli/deploy/agent_runtime.py

@@ -24,6 +24,7 @@ import datetime
 import json
 import logging
 import os
+import urllib.parse
 import warnings
 from pathlib import Path
 from typing import Any
@@ -172,6 +173,16 @@ def _get_resource_name_from_operation(operation_name: str) -> str:
     return resource_name
 
 
+def build_agent_engine_logs_url(operation_name: str, project: str) -> str:
+    """Build a Logs Explorer URL for the engine's logs, keyed off the LRO name."""
+    engine_id = _get_resource_name_from_operation(operation_name).split("/")[-1]
+    query = f'resource.labels.reasoning_engine_id="{engine_id}"'
+    encoded = urllib.parse.quote(query, safe="")
+    return (
+        f"https://console.cloud.google.com/logs/query;query={encoded}?project={project}"
+    )
+
+
 def write_deployment_metadata(
     remote_agent: Any,
     cfg: ProjectConfig,
@@ -602,21 +613,18 @@ def deploy_agent_runtime(
     # Deploy (create or update)
     action = "Updating" if matching_agents else "Creating"
 
-    if no_wait:
-        click.echo(f"\n🚀 {action} agent: {display_name} (returning immediately)...")
-        operation = _start_and_record_operation(
-            client, config, matching_agents, project, location
-        )
-        click.echo(f"\n📋 Operation started: {operation.name}")
-        click.echo("   Check status with: agents-cli deploy --status")
-        return None
-
-    click.echo(f"\n🚀 {action} agent: {display_name} (this can take 5-10 minutes)...")
+    wait_note = "not waiting for completion" if no_wait else "this can take a few minutes"
+    click.echo(f"\n🚀 {action} agent: {display_name} ({wait_note})...")
 
     operation = _start_and_record_operation(
         client, config, matching_agents, project, location
     )
-    click.echo(f"   Operation: {operation.name}")
+    logs_url = build_agent_engine_logs_url(operation.name, project)
+
+    click.echo(f"   Operation: {operation.name}\n   Monitor deploy logs: {logs_url}")
+    if no_wait:
+        click.echo("   Check status with: agents-cli deploy --status")
+        return None
     click.echo(
         "   If this command is interrupted, run 'agents-cli deploy --status' to check progress."
     )
@@ -779,6 +787,9 @@ def check_agent_runtime_operation(
             seconds = int(delta.total_seconds() % 60)
             elapsed = f" ({minutes}m {seconds}s elapsed)"
 
-        click.echo(f"⏳ Deployment still in progress{elapsed}")
-        click.echo(f"   Operation: {operation_name}")
-        click.echo("   Run 'agents-cli deploy --status' again to check.")
+        click.echo(
+            f"⏳ Deployment still in progress{elapsed}\n"
+            f"   Operation: {operation_name}\n"
+            f"   Monitor deploy logs: {build_agent_engine_logs_url(operation_name, project)}\n"
+            "   Run 'agents-cli deploy --status' again to check."
+        )

{google_agents_cli-1.2.1 → google_agents_cli-1.3.0}/src/google/agents/cli/deploy/cmd_deploy.py

@@ -195,10 +195,14 @@ _CLOUD_RUN_TRANSIENT_DEPLOY_SIGNATURES = (
     "artifactregistry.repositories.downloadArtifacts",
 )
 
-# Retry budget for transient Cloud Run deploy failures. Propagation can take a
-# few minutes, so we allow several attempts with exponential backoff + jitter.
-_CLOUD_RUN_DEPLOY_MAX_TRIES = 5
-_CLOUD_RUN_DEPLOY_MAX_TIME = 300
+# Retry budget for transient Cloud Run deploy failures. IAM propagation can take
+# several minutes, so we retry until max_time with a capped exponential backoff:
+# waits ramp 5s, 10s, 20s then hold at 30s (each full-jittered), i.e. steady
+# ~30s polling until the deadline. max_tries is unset so max_time is the sole
+# stop condition.
+_CLOUD_RUN_DEPLOY_MAX_TIME = 600
+_CLOUD_RUN_DEPLOY_BACKOFF_FACTOR = 5
+_CLOUD_RUN_DEPLOY_BACKOFF_MAX_VALUE = 30
 
 
 class _TransientCloudRunDeployError(click.ClickException):
@@ -212,15 +216,18 @@ class _TransientCloudRunDeployError(click.ClickException):
 @backoff.on_exception(
     backoff.expo,
     _TransientCloudRunDeployError,
-    max_tries=_CLOUD_RUN_DEPLOY_MAX_TRIES,
+    factor=_CLOUD_RUN_DEPLOY_BACKOFF_FACTOR,
+    max_value=_CLOUD_RUN_DEPLOY_BACKOFF_MAX_VALUE,
+    max_tries=None,
     max_time=_CLOUD_RUN_DEPLOY_MAX_TIME,
     jitter=backoff.full_jitter,
     on_backoff=lambda details: logging.warning(
         "Cloud Run deploy hit a transient IAM-propagation error; retrying in "
-        "%.0fs (attempt %d/%d)...",
+        "%.0fs (attempt %d, %.0fs/%ds elapsed)...",
         details["wait"],
         details["tries"] + 1,  # the upcoming attempt; details['tries'] = ones done
-        _CLOUD_RUN_DEPLOY_MAX_TRIES,
+        details["elapsed"],
+        _CLOUD_RUN_DEPLOY_MAX_TIME,
     ),
 )
 def _run_cloud_run_deploy_with_retry(args: list[str], *, project: str | None) -> None: