mirror of
https://gh.wpcy.net/https://github.com/discourse/discourse.git
synced 2026-05-03 22:38:10 +08:00
b2d73b346d
Introduce support for Model Context Protocol (MCP) servers in the discourse-ai plugin, allowing AI agents to connect to external tool servers via the MCP standard. Key additions: - AiMcpServer model with CRUD admin UI, health tracking, and tool caching (hourly refresh via scheduled job) - MCP client (Streamable HTTP transport) with session management and tool invocation - Full OAuth 2.1 flow support (discovery, dynamic registration, authorization code grant, token refresh, and disconnect) - MCP tool type for AI agents that proxies tool calls to remote MCP servers at runtime - Agent editor updated to show combined tool/token counts from both local tools and MCP servers - Agent import/export includes MCP server associations - Admin secrets UI updated to surface MCP server usage - Comprehensive specs for models, controllers, client, tool registry, and OAuth flow
23 lines
685 B
Ruby
23 lines
685 B
Ruby
# frozen_string_literal: true
|
|
|
|
RSpec.describe AiAgentMcpServer do
|
|
fab!(:ai_mcp_server)
|
|
|
|
before { enable_current_plugin }
|
|
|
|
it "memoizes serialized tools for count calculations" do
|
|
ai_mcp_server
|
|
.expects(:tools_for_serialization)
|
|
.once
|
|
.returns(
|
|
[{ name: "search_issues", token_count: 3 }, { name: "create_issue", token_count: 5 }],
|
|
)
|
|
|
|
assignment =
|
|
described_class.new(ai_mcp_server: ai_mcp_server, selected_tool_names: ["search_issues"])
|
|
|
|
expect(assignment.tools_for_serialization).to eq([{ name: "search_issues", token_count: 3 }])
|
|
expect(assignment.tool_count).to eq(1)
|
|
expect(assignment.token_count).to eq(3)
|
|
end
|
|
end
|
