inject a mid-turn message instead of replacing the turn

Phase C server half, and a real flaw in phase B.

On the subprocess path a second message could only supersede — kill the process, start again,
lose the turn — because opencode run has no input channel. The serve takes another prompt
into the running turn, so a message arriving mid-turn is handed over with delivery steer and
the existing turn is left exactly as it is.

Keeping the same turn object is the load-bearing part. Phase B retired it and registered a
replacement, which stops officer routing events the serve is still producing while the serve
carries on regardless: output goes nowhere and the turn looks hung.

Verified end to end through the chat socket — sent a count to 50, injected a change of plan
eight seconds in, and BANANA INJECTED came back inside the same turn with deltas streaming
throughout.

No client change was needed. Officer composer already sends while generating; the difference
is only what the sidecar does with it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

docs/opencode-serve-migration-plan.md

@@ -58,6 +58,19 @@ Each of these cost time to find. None is in the API docs.
 5. **The event names are `session.next.*`** — `step.started`, `text.started`, `text.ended`,
    `tool.called`, `tool.success`, `step.ended`, `step.failed`. Not the shapes `mapRunLine` handles.
 
+## Status, 2026-08-10
+
+- **Phase A — done.** `serve-events.ts` + tests, fixtures captured from real turns.
+- **Phase B — done, behind `OPENCODE_TURNS=serve` (default: subprocess).** `serve-runner.ts`. Verified
+  end to end through the chat socket: tool call, tool result, **3 streaming deltas**, text, cost. Stop is
+  an interrupt and the session survives it.
+- **Phase C — server half done.** A message sent while a turn runs is injected with `delivery: "steer"`
+  into the RUNNING turn, verified end to end. No client change was needed: officer's composer already
+  sends mid-turn, and the subprocess path was superseding where the serve steers.
+- **Phase D — not started, deliberately.**
+
+Not yet lived with. Nothing here has run a real conversation with a person at the other end.
+
 ## Shape of the work
 
 **Phase A — read the stream without depending on it.** Add a serve-based reader alongside the existing

src/servers/sidecar/opencode/serve-runner.ts

@@ -168,11 +168,39 @@ export async function runOpenCodeTurnOnServe(
 
   ensureEventStream(config);
 
-  // Supersede any turn still registered under this key. Unlike the subprocess path there is no process
-  // to kill — the serve owns execution — so this is bookkeeping only, and the old turn is retired
-  // silently rather than reporting an error against a key that now belongs to its replacement.
-  const stale = bySessionKey.get(sessionKey);
-  if (stale) retire(stale, null);
+  // ── A message arriving while a turn is running is an INJECTION, not a new turn ──
+  //
+  // This is where the subprocess and the serve genuinely part company. `opencode run` had no input
+  // channel, so a second message could only supersede: kill the process, start again, lose the turn.
+  // The serve takes another prompt into the RUNNING turn, so the right move is to hand it over and keep
+  // the existing turn exactly as it is.
+  //
+  // Keeping the same turn object is the load-bearing part. Retiring it and registering a replacement —
+  // which is what this did at first — stops officer routing the events the serve is still producing,
+  // while the serve carries on regardless. The output goes nowhere and the turn looks hung.
+  //
+  // `steer` because the user typed it during the turn and means it now; officer's composer already
+  // treats a send-while-generating as "add this to what you are doing". A prompt sent when nothing is
+  // running takes `queue`, which is a no-op with an empty queue but never accidentally merges two
+  // messages into one turn.
+  const live = bySessionKey.get(sessionKey);
+  if (live && !live.done) {
+    try {
+      await serveJson(config, `/api/session/${live.openCodeSessionId}/prompt`, {
+        method: 'POST',
+        body: JSON.stringify({ prompt: { text: params.prompt }, delivery: 'steer' }),
+        cwd,
+      });
+    } catch (err) {
+      // The turn itself is unharmed — only the injection failed — so say so and leave it running.
+      emit({
+        type: 'opencode:event',
+        sessionKey,
+        event: { type: 'error', message: `OpenCode would not take that mid-turn: ${errText(err)}` },
+      });
+    }
+    return;
+  }
 
   let openCodeSessionId = params.resumeSessionId ?? '';