Files
2026-07-13 12:58:18 +08:00

91 lines
4.3 KiB
C#

using Microsoft.Agents.AI;
using Microsoft.Extensions.AI;
using OpenAI;
// In-App HITL (frontend-tool + popup modal) agent.
//
// The agent is a support-ops copilot. Any action that materially affects
// a customer MUST be confirmed by the operator via the frontend-provided
// `request_user_approval` tool (registered via `useFrontendTool` on the
// page). The tool handler opens a modal OUTSIDE the chat surface and
// returns `{ approved: boolean, reason?: string }` back to the agent.
//
// This agent owns NO server-side tools — the approval tool lives on the
// frontend. The system prompt tells the model to invoke it whenever a
// customer-affecting action is requested.
//
// Harness column: the inner ChatClientAgent is built through the
// `chatClient.AsHarnessAgent(...)` wrapper (Microsoft Agent Harness over
// Microsoft Agent Framework) and the credential comes from the single shared
// `OpenAIClient` threaded in from Program.cs (built via the harness
// ApiKeyResolver) — no per-feature GitHubToken dance. See the W0 contract §1.
//
// Reference parity with:
// showcase/integrations/langgraph-python/src/agents/hitl_in_app.py
public sealed class HitlInAppAgentFactory
{
private const int HarnessMaxContextWindowTokens = 128_000;
private const int HarnessMaxOutputTokens = 8_192;
private const string SystemPrompt =
"You are a support operations copilot working alongside a human operator " +
"inside an internal support console. The operator can see a list of open " +
"support tickets on the left side of their screen and is chatting with " +
"you on the right.\n\n" +
"Whenever the operator asks you to take an action that affects a " +
"customer — for example: issuing a refund, updating a customer's plan, " +
"cancelling a subscription, escalating a ticket, or sending an apology " +
"credit — you MUST first call the frontend-provided " +
"`request_user_approval` tool to obtain the operator's explicit consent.\n\n" +
"How to use `request_user_approval`:\n" +
"- `message`: a short, plain-English summary of the exact action you " +
" are about to take, including concrete numbers (e.g. '$50 refund to " +
" customer #12345').\n" +
"- `context`: optional extra context the operator might want to review " +
" (the ticket ID, the policy rule you're applying, etc.). Keep it to " +
" one or two short sentences.\n\n" +
"The tool returns an object of the shape " +
"`{\"approved\": boolean, \"reason\": string | null}`.\n" +
"- If `approved` is `true`: confirm in one short sentence that you are " +
" processing the action. You do not actually need to call any other " +
" tool — this is a demo. Just acknowledge.\n" +
"- If `approved` is `false`: acknowledge the rejection in one short " +
" sentence and, if `reason` is non-empty, reflect the operator's " +
" reason back to them. Do NOT retry the action.\n\n" +
"Keep all chat replies to one or two short sentences. Never make up " +
"customer data — always use whatever the operator told you in the " +
"prompt.";
private readonly OpenAIClient _openAiClient;
private readonly ILogger _logger;
public HitlInAppAgentFactory(OpenAIClient openAiClient, ILoggerFactory loggerFactory)
{
ArgumentNullException.ThrowIfNull(openAiClient);
ArgumentNullException.ThrowIfNull(loggerFactory);
_openAiClient = openAiClient;
_logger = loggerFactory.CreateLogger<HitlInAppAgentFactory>();
}
public AIAgent CreateHitlInAppAgent()
{
var chatClient = _openAiClient.GetChatClient("gpt-4o-mini").AsIChatClient();
return chatClient.AsHarnessAgent(
HarnessMaxContextWindowTokens,
HarnessMaxOutputTokens,
new HarnessAgentOptions
{
Name = "HitlInAppAgent",
Description = "In-App HITL support-ops copilot powered by Microsoft Agent Harness over Microsoft Agent Framework.",
ChatOptions = new ChatOptions
{
Instructions = SystemPrompt,
MaxOutputTokens = HarnessMaxOutputTokens,
Tools = [],
},
});
}
}