fips-agents-cli 0.6.2__tar.gz → 0.7.0__tar.gz

{fips_agents_cli-0.6.2 → fips_agents_cli-0.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: fips-agents-cli
-Version: 0.6.2
+Version: 0.7.0
 Summary: CLI tool for creating and managing FIPS-compliant AI agent projects
 Project-URL: Homepage, https://github.com/fips-agents/fips-agents-cli
 Project-URL: Repository, https://github.com/fips-agents/fips-agents-cli
@@ -141,6 +141,7 @@ fips-agents patch all --dry-run
 ```bash
 fips-agents --version
 fips-agents --help
+fips-agents add --help
 fips-agents create --help
 fips-agents generate --help
 fips-agents patch --help
@@ -338,7 +339,7 @@ fips-agents create model-car https://huggingface.co/ibm-granite/granite-3.1-2b-i
 
 ### Generate Commands
 
-The `generate` command group scaffolds MCP components (tools, resources, prompts, middleware) in existing MCP server projects created with `fips-agents create mcp-server`.
+The `generate` command group creates individual components (tools, resources, prompts, middleware) in existing projects.
 
 **Run these commands from within your MCP server project directory.** The CLI locates the project root by looking for `pyproject.toml` with a `fastmcp` dependency.
 
@@ -600,6 +601,40 @@ Update all patchable file categories at once. Prompts for confirmation before st
 
 All patch subcommands (except `check`) accept `--dry-run` to preview changes without modifying files.
 
+---
+
+### Add Commands
+
+The `add` command group wires optional capabilities into existing agent projects created with `fips-agents create agent`.
+
+**Run these commands from within your agent project directory.** The CLI locates the project root by looking for `agent.yaml`.
+
+#### `add code-executor`
+
+```bash
+fips-agents add code-executor
+```
+
+Adds the `code_executor` tool to your agent's `tools/` directory and enables the sandbox sidecar in `chart/values.yaml`. The tool sends Python code to a sandbox container for isolated execution.
+
+**What it does:**
+
+1. Writes `tools/code_executor.py` with the sandbox client tool
+2. Sets `sandbox.enabled: true` in `chart/values.yaml` (if the section exists)
+3. Prints deployment guidance for sidecar and remote service modes
+
+**Deployment modes:**
+
+- **Sidecar** (default): The sandbox runs as a container in the same pod. The tool connects to `http://localhost:8000`.
+- **Remote service**: The sandbox runs as a separate deployment. Set `SANDBOX_URL` to point to it.
+
+**Example:**
+
+```bash
+cd my-research-agent
+fips-agents add code-executor
+```
+
 ## Project Name Requirements
 
 Project names must follow these rules:
@@ -739,6 +774,7 @@ fips-agents-cli/
 │       ├── cli.py          # Main CLI application
 │       ├── version.py      # Version information
 │       ├── commands/       # CLI command implementations
+│       │   ├── add.py      # add code-executor (wire capabilities)
 │       │   ├── create.py   # create mcp-server, agent, gateway, ui
 │       │   ├── generate.py # generate tool/resource/prompt/middleware
 │       │   ├── model_car.py # create model-car
@@ -861,6 +897,31 @@ MIT License - see LICENSE file for details
 
 ## Changelog
 
+### Version 0.7.0
+
+- Feature: New `add` command group for composable agent capabilities
+- Feature: `fips-agents add code-executor` wires sandbox code execution into existing agent projects (tool + Helm chart sidecar config)
+- Improvement: Updated `generate` group description to reflect broader scope beyond MCP
+
+### Version 0.6.2
+
+- Fix: Go project scaffolding now updates import paths in all `.go` source files
+- Fix: Go project scaffolding now updates Helm template references (`chart/templates/*.yaml`)
+
+### Version 0.6.1
+
+- Chore: Updated all project URLs from `rdwj/` to `fips-agents/` org
+- Chore: Post-transfer PyPI publishing verification release
+
+### Version 0.6.0
+
+- Feature: Added `create sandbox` command for scaffolding code execution sandbox projects from fips-agents/code-sandbox template
+- Feature: All template URLs consolidated under `github.com/fips-agents/` organization
+
+### Version 0.5.1
+
+- Fix: SSE streaming tool call ID tracking by `call_id` not array index
+
 ### Version 0.5.0
 
 - Feature: Added `create gateway` command for scaffolding Go HTTP gateway projects from the gateway-template

{fips_agents_cli-0.6.2 → fips_agents_cli-0.7.0}/README.md

@@ -108,6 +108,7 @@ fips-agents patch all --dry-run
 ```bash
 fips-agents --version
 fips-agents --help
+fips-agents add --help
 fips-agents create --help
 fips-agents generate --help
 fips-agents patch --help
@@ -305,7 +306,7 @@ fips-agents create model-car https://huggingface.co/ibm-granite/granite-3.1-2b-i
 
 ### Generate Commands
 
-The `generate` command group scaffolds MCP components (tools, resources, prompts, middleware) in existing MCP server projects created with `fips-agents create mcp-server`.
+The `generate` command group creates individual components (tools, resources, prompts, middleware) in existing projects.
 
 **Run these commands from within your MCP server project directory.** The CLI locates the project root by looking for `pyproject.toml` with a `fastmcp` dependency.
 
@@ -567,6 +568,40 @@ Update all patchable file categories at once. Prompts for confirmation before st
 
 All patch subcommands (except `check`) accept `--dry-run` to preview changes without modifying files.
 
+---
+
+### Add Commands
+
+The `add` command group wires optional capabilities into existing agent projects created with `fips-agents create agent`.
+
+**Run these commands from within your agent project directory.** The CLI locates the project root by looking for `agent.yaml`.
+
+#### `add code-executor`
+
+```bash
+fips-agents add code-executor
+```
+
+Adds the `code_executor` tool to your agent's `tools/` directory and enables the sandbox sidecar in `chart/values.yaml`. The tool sends Python code to a sandbox container for isolated execution.
+
+**What it does:**
+
+1. Writes `tools/code_executor.py` with the sandbox client tool
+2. Sets `sandbox.enabled: true` in `chart/values.yaml` (if the section exists)
+3. Prints deployment guidance for sidecar and remote service modes
+
+**Deployment modes:**
+
+- **Sidecar** (default): The sandbox runs as a container in the same pod. The tool connects to `http://localhost:8000`.
+- **Remote service**: The sandbox runs as a separate deployment. Set `SANDBOX_URL` to point to it.
+
+**Example:**
+
+```bash
+cd my-research-agent
+fips-agents add code-executor
+```
+
 ## Project Name Requirements
 
 Project names must follow these rules:
@@ -706,6 +741,7 @@ fips-agents-cli/
 │       ├── cli.py          # Main CLI application
 │       ├── version.py      # Version information
 │       ├── commands/       # CLI command implementations
+│       │   ├── add.py      # add code-executor (wire capabilities)
 │       │   ├── create.py   # create mcp-server, agent, gateway, ui
 │       │   ├── generate.py # generate tool/resource/prompt/middleware
 │       │   ├── model_car.py # create model-car
@@ -828,6 +864,31 @@ MIT License - see LICENSE file for details
 
 ## Changelog
 
+### Version 0.7.0
+
+- Feature: New `add` command group for composable agent capabilities
+- Feature: `fips-agents add code-executor` wires sandbox code execution into existing agent projects (tool + Helm chart sidecar config)
+- Improvement: Updated `generate` group description to reflect broader scope beyond MCP
+
+### Version 0.6.2
+
+- Fix: Go project scaffolding now updates import paths in all `.go` source files
+- Fix: Go project scaffolding now updates Helm template references (`chart/templates/*.yaml`)
+
+### Version 0.6.1
+
+- Chore: Updated all project URLs from `rdwj/` to `fips-agents/` org
+- Chore: Post-transfer PyPI publishing verification release
+
+### Version 0.6.0
+
+- Feature: Added `create sandbox` command for scaffolding code execution sandbox projects from fips-agents/code-sandbox template
+- Feature: All template URLs consolidated under `github.com/fips-agents/` organization
+
+### Version 0.5.1
+
+- Fix: SSE streaming tool call ID tracking by `call_id` not array index
+
 ### Version 0.5.0
 
 - Feature: Added `create gateway` command for scaffolding Go HTTP gateway projects from the gateway-template

{fips_agents_cli-0.6.2 → fips_agents_cli-0.7.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "fips-agents-cli"
-version = "0.6.2"
+version = "0.7.0"
 description = "CLI tool for creating and managing FIPS-compliant AI agent projects"
 readme = "README.md"
 requires-python = ">=3.10"

{fips_agents_cli-0.6.2 → fips_agents_cli-0.7.0}/src/fips_agents_cli/cli.py

@@ -3,6 +3,7 @@
 import click
 from rich.console import Console
 
+from fips_agents_cli.commands.add import add
 from fips_agents_cli.commands.create import create
 from fips_agents_cli.commands.generate import generate
 from fips_agents_cli.commands.patch import patch
@@ -24,6 +25,7 @@ def cli(ctx):
 
 
 # Register commands
+cli.add_command(add)
 cli.add_command(create)
 cli.add_command(generate)
 cli.add_command(patch)

fips_agents_cli-0.7.0/src/fips_agents_cli/commands/add.py

@@ -0,0 +1,211 @@
+"""Add command group for wiring capabilities into existing agent projects."""
+
+import sys
+from pathlib import Path
+
+import click
+from rich.console import Console
+from rich.panel import Panel
+
+console = Console()
+
+# The code_executor tool source, embedded directly to avoid network dependencies.
+# Source: agent-template/examples/code-sandbox-agent/tools/code_executor.py
+CODE_EXECUTOR_TOOL_SOURCE = '''\
+"""Code execution tool — sends Python code to the sandbox sidecar."""
+
+import os
+
+import httpx
+
+from fipsagents.baseagent.tools import tool
+
+SANDBOX_URL = os.environ.get("SANDBOX_URL", "http://localhost:8000")
+
+
+@tool(
+    description="Execute Python code in an isolated sandbox and return the output. "
+    "Use this for any computation, math, data processing, or logic that "
+    "benefits from exact results. The code runs in a restricted environment "
+    "with access to: math, statistics, itertools, functools, re, datetime, "
+    "collections, json, csv, string, textwrap, decimal, fractions, random, "
+    "operator, typing. Use print() to produce output.",
+    visibility="llm_only",
+)
+async def code_executor(code: str, timeout: float = 10.0) -> str:
+    """Execute Python code in the sandbox sidecar.
+
+    Args:
+        code: Python source code to execute. Must use print() for output.
+        timeout: Maximum execution time in seconds (1-30).
+    """
+    timeout = max(1.0, min(timeout, 30.0))
+
+    async with httpx.AsyncClient(timeout=timeout + 5) as client:
+        try:
+            resp = await client.post(
+                f"{SANDBOX_URL}/execute",
+                json={"code": code, "timeout": timeout},
+            )
+        except httpx.ConnectError:
+            return (
+                "ERROR: Cannot connect to sandbox sidecar at "
+                f"{SANDBOX_URL}. Is it running?"
+            )
+        except httpx.TimeoutException:
+            return "ERROR: Request to sandbox timed out."
+
+    data = resp.json()
+
+    if resp.status_code == 400:
+        if "violations" in data:
+            violations = "\\n".join(f"  - {v}" for v in data["violations"])
+            return f"CODE BLOCKED by sandbox guardrails:\\n{violations}"
+        return f"ERROR: {data.get(\'error\', \'Unknown error\')}"
+
+    stdout = data.get("stdout", "").strip()
+    stderr = data.get("stderr", "").strip()
+    exit_code = data.get("exit_code", -1)
+    timed_out = data.get("timed_out", False)
+
+    if timed_out:
+        return f"TIMEOUT: Code exceeded {timeout}s limit.\\nPartial output:\\n{stdout}"
+
+    parts = []
+    if stdout:
+        parts.append(stdout)
+    if stderr:
+        parts.append(f"STDERR:\\n{stderr}")
+    if exit_code != 0:
+        parts.append(f"(exit code {exit_code})")
+
+    return "\\n".join(parts) if parts else "(no output — did you forget print()?)"
+'''
+
+
+def _find_agent_project_root() -> Path | None:
+    """Find the agent project root by looking for agent.yaml."""
+    current = Path.cwd()
+    for parent in [current] + list(current.parents):
+        if (parent / "agent.yaml").exists():
+            return parent
+    return None
+
+
+@click.group()
+def add():
+    """Add capabilities to an existing agent project."""
+    pass
+
+
+@add.command("code-executor")
+def code_executor_cmd():
+    """Wire sandbox code execution into the current agent project.
+
+    Adds the code_executor tool to tools/ and enables the sandbox sidecar
+    in chart/values.yaml. Run from your agent project root directory.
+
+    Example:
+
+        cd my-research-agent
+
+        fips-agents add code-executor
+    """
+    try:
+        # Step 1: Detect project root
+        project_root = _find_agent_project_root()
+        if project_root is None:
+            console.print(
+                "[red]Error:[/red] Not in an agent project directory.\n"
+                "Could not find agent.yaml in this directory or any parent.\n\n"
+                "[yellow]Hint:[/yellow] Run this command from an agent project "
+                "created with [dim]fips-agents create agent[/dim]."
+            )
+            sys.exit(1)
+
+        console.print(f"[green]Found project root:[/green] {project_root}")
+
+        # Step 2: Check tools/ directory exists
+        tools_dir = project_root / "tools"
+        if not tools_dir.exists():
+            console.print(
+                "[red]Error:[/red] No tools/ directory found in project root.\n\n"
+                "[yellow]Hint:[/yellow] Create it with: [dim]mkdir tools[/dim]"
+            )
+            sys.exit(1)
+
+        # Step 3: Write the tool file
+        tool_file = tools_dir / "code_executor.py"
+        if tool_file.exists():
+            console.print(
+                "[yellow]Warning:[/yellow] tools/code_executor.py already exists. Skipping."
+            )
+        else:
+            tool_file.write_text(CODE_EXECUTOR_TOOL_SOURCE)
+            console.print("[green]+[/green] Created tools/code_executor.py")
+
+        # Step 4: Update chart/values.yaml if present
+        values_file = project_root / "chart" / "values.yaml"
+        if values_file.exists():
+            values_content = values_file.read_text()
+            if "sandbox:" in values_content:
+                if "enabled: false" in values_content:
+                    values_content = values_content.replace("enabled: false", "enabled: true", 1)
+                    values_file.write_text(values_content)
+                    console.print("[green]+[/green] Set sandbox.enabled: true in chart/values.yaml")
+                elif "enabled: true" in values_content:
+                    console.print("[dim]sandbox.enabled already true in chart/values.yaml[/dim]")
+                else:
+                    console.print(
+                        "[yellow]Warning:[/yellow] Found sandbox: section in "
+                        "chart/values.yaml but could not locate enabled field. "
+                        "Please set sandbox.enabled: true manually."
+                    )
+            else:
+                console.print(
+                    "[yellow]Warning:[/yellow] No sandbox: section in chart/values.yaml.\n"
+                    "  Add the following to your values.yaml:\n"
+                    "  [dim]sandbox:\n"
+                    "    enabled: true\n"
+                    "    image: <your-sandbox-image>[/dim]"
+                )
+        else:
+            console.print("[dim]No chart/values.yaml found (not using Helm chart). Skipping.[/dim]")
+
+        # Step 5: Success panel
+        next_steps = """
+[bold cyan]Next Steps:[/bold cyan]
+
+  1. Build or deploy the sandbox sidecar:
+     [dim]fips-agents create sandbox my-sandbox[/dim]
+     or use the pre-built image from fips-agents/code-sandbox
+
+  2. Configure the sandbox URL for your agent:
+
+     [bold]Sidecar mode[/bold] (same pod, default):
+       The tool defaults to http://localhost:8000
+       No extra config needed if sandbox runs as a sidecar container.
+
+     [bold]Remote service mode[/bold] (separate deployment):
+       Set the SANDBOX_URL environment variable:
+       [dim]export SANDBOX_URL=http://sandbox-service:8000[/dim]
+
+  3. The agent will auto-discover tools/code_executor.py.
+     Use it in your agent's step() by calling:
+       [dim]await self.use_tool("code_executor", code="print(1+1)")[/dim]
+"""
+
+        console.print(
+            Panel(
+                f"[bold green]Code executor tool added.[/bold green]\n{next_steps}",
+                border_style="green",
+                padding=(1, 2),
+            )
+        )
+
+    except KeyboardInterrupt:
+        console.print("\n[yellow]Operation cancelled.[/yellow]")
+        sys.exit(130)
+    except Exception as e:
+        console.print(f"\n[red]Error:[/red] {e}")
+        sys.exit(1)

{fips_agents_cli-0.6.2 → fips_agents_cli-0.7.0}/src/fips_agents_cli/commands/generate.py

@@ -301,7 +301,7 @@ def generate_component_workflow(
 
 @click.group()
 def generate():
-    """Generate new MCP components in existing projects."""
+    """Generate individual components (tools, prompts, resources, middleware) in existing projects."""
     pass
 
 

{fips_agents_cli-0.6.2 → fips_agents_cli-0.7.0}/src/fips_agents_cli/version.py

@@ -1,3 +1,3 @@
 """Version information for fips-agents-cli."""
 
-__version__ = "0.6.2"
+__version__ = "0.7.0"