Commit graph

3 commits

Author SHA1 Message Date
jottakka
9c47f73602
[TOO-518] Enforce semver for MCPApp Versioning (#793)
Here's the PR summary:

---

## Enforce semver validation for `MCPApp` versioning

### Problem

`MCPApp.__init__` accepted any string as `version` with no validation.
Invalid versions like `"1.0.0dev"` or `"latest"` silently propagated to
the Engine, where `compareToolVersions` fell back to lexicographic
`strings.Compare` instead of `semver.Compare` — causing incorrect
ordering (e.g. `1.10.0 < 1.9.0`).

### Solution

Validate and normalize `version` at `MCPApp` instantiation time using
the same acceptance rules as Go's `golang.org/x/mod/semver v0.31.0` (the
exact version used by the Engine).

### Changes

**`arcade_mcp_server/_validation.py`** (new file)
- Shared regex constants: `SEMVER_PATTERN` (semver.org spec),
`SHORT_VERSION_PATTERN`, `MAJOR_ONLY_PATTERN`

**`arcade_mcp_server/mcp_app.py`**
- Added `_validate_version()` mirroring the existing `_validate_name()`
pattern
- Added `version` property + setter (validates on mutation too)
- `__init__` now stores `self._version` via `_validate_version()`

**`arcade_mcp_server/settings.py`**
- Added `@field_validator("version")` on `ServerSettings` — covers the
`MCP_SERVER_VERSION` env var path
- Fixed default from `"0.1.0dev"` → `"0.1.0"` (the old default was
itself invalid)

**`pyproject.toml`** — bumped `arcade-mcp-server` `1.17.4` → `1.17.5`

### Normalization pipeline

All inputs are normalized to canonical `MAJOR.MINOR.PATCH` before
storage:

| Input | Stored as |
|-------|-----------|
| `v1.0.0` | `1.0.0` |
| `1.0` / `v1.0` | `1.0.0` |
| `1` / `v1` | `1.0.0` |

### Verification

Validated against `golang.org/x/mod/semver v0.31.0` (Engine's exact
pinned version) — 40/40 accept/reject cases match. The Engine's own
`store_test.go` uses `"1.0"` and `"1.1"` as `ToolkitVersion` values,
confirming short forms are intentionally supported.

### Breaking change

Any user currently passing a non-semver version string (e.g.
`"1.0.0dev"`, `"latest"`) will get a `ValueError` on upgrade. This is
intentional — those versions were silently causing incorrect tool
ordering in the Engine.

Closes TOO-518

<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> **Medium Risk**
> Introduces stricter version validation/normalization that will raise
errors for previously-accepted non-semver inputs (including via env
vars), which may break existing consumers depending on lax version
strings.
> 
> **Overview**
> **Enforces semver for server versioning** across both `MCPApp` and
`ServerSettings`, rejecting invalid strings and normalizing accepted
inputs (e.g., stripping leading `v`, expanding `1`/`1.2` to
`1.0.0`/`1.2.0`).
> 
> Adds shared `normalize_version` logic in
`arcade_mcp_server/_validation.py`, updates `MCPApp` to validate on init
and via a new `version` property/setter, and adds a Pydantic `version`
validator so `MCP_SERVER_VERSION` is checked. Defaults are updated from
`0.1.0dev` to `0.1.0`, tests are expanded to cover accept/reject cases,
and the package version is bumped to `1.17.5`.
> 
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
2ceabacb25372e67eef9720b901c1ee2b214868f. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

---------

Co-authored-by: Eric Gustin <34000337+EricGustin@users.noreply.github.com>
2026-03-16 16:06:25 -07:00
jottakka
98fad93d21
Adding MCP Servers supports to Arcade Evals (#689)
# MCP Server Tool Evaluation Support

## Overview
Add support for evaluating tools from remote MCP servers without
requiring Python callables. Enables direct evaluation of any
MCP-compatible tool server.

## What's New

### Core Features
- **`MCPToolRegistry`**: Evaluate tools from a single MCP server
- **`CompositeMCPRegistry`**: Evaluate tools from multiple MCP servers
simultaneously
- **Automatic loaders**: `load_from_stdio()` and `load_from_http()` to
fetch tools from running servers
- **Automatic namespacing**: Tools prefixed with server name (e.g.,
`server_tool_name`)
- **Smart name resolution**: Use short names if unique, full names if
ambiguous
- **OpenAI strict mode**: Automatic schema conversion prevents parameter
hallucinations

### Usage

**Automatic Loading:**
```python
from arcade_evals import load_from_stdio, MCPToolRegistry

# Load tools automatically from MCP server
tools = load_from_stdio(["npx", "-y", "@modelcontextprotocol/server-github"])
registry = MCPToolRegistry(tools)
```

**Single MCP Server:**
```python
from arcade_evals import MCPToolRegistry, ExpectedToolCall

registry = MCPToolRegistry(mcp_tools)
suite = EvalSuite(catalog=registry)

suite.add_case(
    expected_tool_calls=[
        ExpectedToolCall(tool_name="tool_name", args={...})
    ]
)
```

**Multiple MCP Servers:**
```python
from arcade_evals import CompositeMCPRegistry, load_from_stdio

# Load from multiple servers
github_tools = load_from_stdio(["npx", "-y", "@modelcontextprotocol/server-github"])
slack_tools = load_from_stdio(["npx", "-y", "@modelcontextprotocol/server-slack"])

composite = CompositeMCPRegistry(
    tool_lists={
        "github": github_tools,
        "slack": slack_tools,
    }
)

suite = EvalSuite(catalog=composite)

suite.add_case(
    expected_tool_calls=[
        ExpectedToolCall(tool_name="github_list_issues", args={...})
    ]
)
```

## Implementation

### Files Changed
- **`libs/arcade-evals/arcade_evals/registry.py`** (NEW): Registry
abstractions and implementations
- **`libs/arcade-evals/arcade_evals/loaders.py`** (NEW): Automatic tool
loading from MCP servers
- **`libs/arcade-evals/arcade_evals/eval.py`** (MODIFIED): Enhanced
`ExpectedToolCall` and evaluation logic
- **`libs/arcade-evals/arcade_evals/__init__.py`** (MODIFIED): Exported
new registries and loaders

### Key Technical Details
- Added `BaseToolRegistry` interface for abstraction
- `MCPToolRegistry` handles single server tools
- `CompositeMCPRegistry` manages multiple servers with collision
detection
- `load_from_stdio()` and `load_from_http()` for automatic tool
discovery
- Fixed name normalization bug: MCP tools use underscores (not dots)
- Optimized tool copying: 2.5x faster via shallow copy

## Testing
-  41 tests passing (25 new tests added)
-  `test_eval_mcp_registry.py`: MCPToolRegistry functionality
-  `test_eval_composite_mcp.py`: CompositeMCPRegistry with multiple
servers
-  Verified backward compatibility with Python tools

## Backward Compatibility
 **100% backward compatible** - No breaking changes


## Breaking Changes
**None**


<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> Adds end-to-end eval UX: examples, a robust CLI runner, and rich
outputs.
> 
> - **New examples**: `eval_arcade_gateway.py`,
`eval_stdio_mcp_server.py`, `eval_http_mcp_server.py`,
`eval_comprehensive_comparison.py` with timeouts, error handling, and
track-based comparisons; detailed `README.md`
> - **CLI runner**: `arcade_cli/evals_runner.py` to execute
evals/capture in parallel with progress, error isolation, failed-only
filtering, context inclusion, and multi-provider/model support
> - **Output formatters**: `arcade_cli/formatters/` (txt, md, html,
json) for evals and capture; comparative and multi-model HTML with tabs
and context rendering
> - **Display refactor**: `display.py` now supports writing multiple
formats, failed-only disclaimers, include-context, and improved console
summaries
> 
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
ff8acf9c34a6b61462a019a1ee9df081006517d0. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

---------

Co-authored-by: Francisco Liberal <francisco@arcade.dev>
Co-authored-by: Mateo Torres <torresmateo@gmail.com>
2026-01-07 20:26:23 -03:00
Eric Gustin
83c0eeab2b
Fix server info bug (#614)
Name, title, version, etc. for an `MCPApp` were being overwritten by its
internal `MCPServer`.

⚠️ this is blocking `arcade deploy` from working
2025-10-13 13:04:18 -07:00