fix(core): restore session request headers (#43188)

packages/core/src/session/compaction.ts

@@ -202,6 +202,7 @@ export const make = (dependencies: Dependencies) => {
       .stream(
         LLM.request({
           model: input.model,
+          http: input.request.http,
           messages: [Message.user(summaryPrompt)],
           tools: [],
           generation: { maxTokens: summaryOutput },

packages/core/src/session/runner/llm.ts

@@ -204,6 +204,13 @@ const layer = Layer.effect(
       const promptCacheKey = /^ses_[0-9a-f]{64}$/.test(session.id) ? session.id.slice(4) : session.id
       const request = LLM.request({
         model,
+        http: {
+          headers: {
+            "x-session-affinity": session.id,
+            "X-Session-Id": session.id,
+            ...(session.parentID ? { "x-parent-session-id": session.parentID } : {}),
+          },
+        },
         providerOptions: { openai: { promptCacheKey } },
         system: [agent.info?.system, system.baseline]
           .filter((part): part is string => part !== undefined && part.length > 0)

packages/core/test/session-runner.test.ts

@@ -1101,6 +1101,16 @@ describe("SessionRunnerLLM", () => {
       yield* session.resume(sessionID)
 
       expect(requests).toHaveLength(2)
+      expect(requests.map((request) => request.http?.headers)).toEqual([
+        {
+          "x-session-affinity": sessionID,
+          "X-Session-Id": sessionID,
+        },
+        {
+          "x-session-affinity": sessionID,
+          "X-Session-Id": sessionID,
+        },
+      ])
       expect(userTexts(requests[0])[0]).toContain("## Objective")
       expect(userTexts(requests[1])).toHaveLength(1)
       expect(userTexts(requests[1])[0]).toContain("<summary>\n## Objective\n- Preserve the task\n</summary>")
@@ -2508,6 +2518,43 @@ describe("SessionRunnerLLM", () => {
     }),
   )
 
+  it.effect("adds session correlation headers to model requests", () =>
+    Effect.gen(function* () {
+      yield* setup
+      const session = yield* SessionV2.Service
+      yield* session.prompt({ sessionID, prompt: Prompt.make({ text: "Run correlated request" }), resume: false })
+
+      requests.length = 0
+      yield* session.resume(sessionID)
+
+      expect(requests[0]?.http?.headers).toEqual({
+        "x-session-affinity": sessionID,
+        "X-Session-Id": sessionID,
+      })
+    }),
+  )
+
+  it.effect("adds the parent session header to child model requests", () =>
+    Effect.gen(function* () {
+      yield* setup
+      const session = yield* SessionV2.Service
+      const parentID = SessionV2.ID.make("ses_runner_parent")
+      const { db } = yield* Database.Service
+      yield* db
+        .update(SessionTable)
+        .set({ parent_id: parentID })
+        .where(eq(SessionTable.id, sessionID))
+        .run()
+        .pipe(Effect.orDie)
+      yield* session.prompt({ sessionID, prompt: Prompt.make({ text: "Run child request" }), resume: false })
+
+      requests.length = 0
+      yield* session.resume(sessionID)
+
+      expect(requests[0]?.http?.headers?.["x-parent-session-id"]).toBe(parentID)
+    }),
+  )
+
   it.effect("bounds 64-character session prompt cache keys", () =>
     Effect.gen(function* () {
       yield* setup