refactor: cursor-agent uses runtime.extract for workspace detection

- Remove llmProvider and workspace from CursorAgentConfig (now just command/model/timeout) - extractWorkspacePath uses runtime.extract + runtime.cas instead of standalone reactor - TextProducerFn signature gains runtime parameter: (ctx, prompt, runtime) - develop-entry.ts hardcodes cursor-agent path, no more env var dependency - Drop @uncaged/workflow-reactor dep from workflow-agent-cursor - Update tests for simplified config 小橘 <xiaoju@shazhou.work>
2026-05-13 15:51:09 +00:00
parent e14643a50b
commit 94f725c50b
16 changed files with 259 additions and 164 deletions
@@ -0,0 +1,101 @@
+/**
+ * greet workflow — smoke test entry
+ * Single role: greeter takes a prompt and returns a structured greeting.
+ * 小橘 🍊
+ */
+
+import type {
+  AdapterFn,
+  ModeratorTable,
+  RoleFn,
+  RoleResult,
+  ThreadContext,
+  WorkflowDefinition,
+  WorkflowRuntime,
+} from "@uncaged/workflow-runtime";
+import { createWorkflow, END, START } from "@uncaged/workflow-runtime";
+import * as z from "zod/v4";
+
+type GreetMeta = {
+  greeter: { greeting: string; language: string };
+};
+
+const greeterSchema = z.object({
+  greeting: z.string().describe("A friendly greeting message"),
+  language: z.string().describe("The language of the greeting"),
+});
+
+const roles: WorkflowDefinition<GreetMeta>["roles"] = {
+  greeter: {
+    description: "Generates a friendly greeting",
+    systemPrompt:
+      "You are a friendly greeter. Given a user prompt, produce a warm greeting. Respond in valid JSON with keys: greeting (string), language (string).",
+    schema: greeterSchema,
+    extractRefs: null,
+  },
+};
+
+const table: ModeratorTable<GreetMeta> = {
+  [START]: [{ condition: "FALLBACK", role: "greeter" }],
+  greeter: [{ condition: "FALLBACK", role: END }],
+};
+
+export const descriptor = {
+  name: "greet",
+  description: "A simple greeting workflow for smoke testing",
+  graph: { [START]: ["greeter"], greeter: [END] },
+  roles: { greeter: { description: "Generates a friendly greeting" } },
+};
+
+function createLazyAdapter(): AdapterFn {
+  let cached: { baseUrl: string; apiKey: string; model: string } | null = null;
+  function getProvider() {
+    if (cached !== null) return cached;
+    const apiKey = process.env.DASHSCOPE_API_KEY;
+    if (!apiKey) throw new Error("missing env: DASHSCOPE_API_KEY");
+    cached = {
+      baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1",
+      apiKey,
+      model: process.env.WORKFLOW_MODEL ?? "qwen-plus",
+    };
+    return cached;
+  }
+
+  return (<T>(prompt: string, schema: z.ZodType<T>): RoleFn<T> => {
+    return async (ctx: ThreadContext, _runtime: WorkflowRuntime): Promise<RoleResult<T>> => {
+      const provider = getProvider();
+      const response = await fetch(`${provider.baseUrl}/chat/completions`, {
+        method: "POST",
+        headers: {
+          Authorization: `Bearer ${provider.apiKey}`,
+          "Content-Type": "application/json",
+        },
+        body: JSON.stringify({
+          model: provider.model,
+          messages: [
+            { role: "system", content: prompt },
+            {
+              role: "user",
+              content: `${ctx.start.content}\n\nRespond with JSON: ${JSON.stringify(z.toJSONSchema(schema))}`,
+            },
+          ],
+          response_format: { type: "json_object" },
+        }),
+      });
+      if (!response.ok) {
+        const body = await response.text();
+        throw new Error(`LLM error ${response.status}: ${body.slice(0, 500)}`);
+      }
+      const data = (await response.json()) as { choices: Array<{ message: { content: string } }> };
+      const text = data.choices[0]?.message?.content;
+      if (!text) throw new Error("Empty LLM response");
+      const parsed = schema.parse(JSON.parse(text));
+      return { meta: parsed, childThread: null };
+    };
+  }) as AdapterFn;
+}
+
+export const run = createWorkflow<GreetMeta>(
+  { roles, table },
+  { adapter: createLazyAdapter(), overrides: null },
+);