sdk, agent: review fixes (T18R)

- The serve loop ends the process after EOF or exit_after_response,
  also with background tasks or listeners.
- One schema generator per direction for the whole document, so types
  with the same name get distinct names.
- The agent gives the complete error (code, message, data) to the
  model, writes SKIP only after a successful request, caps tool output
  at 64 KiB, writes each history line in one write, and skips a damaged
  line.
- Endpoints skip blank and non-RPC lines. client_rs does not panic.
- #[must_use] on builders and the lock guard; #[schemars] attributes
  and duplicate hooks are compile errors; raw identifiers; -32000 for
  local SDK errors; x-sausage-params only for methods with params.

agent/src/llm.rs

@@ -122,6 +122,9 @@ pub enum Entry {
     Opaque {
         /// The API shape that owns this entry.
         api: String,
+        /// The model that produced this item. Legacy items without a model are not replayed.
+        #[serde(default)]
+        model: String,
         /// The original provider object.
         item: Value,
     },
@@ -244,7 +247,7 @@ async fn openai(
             Entry::ToolResult { id, output } => {
                 input.push(json!({"type": "function_call_output", "call_id": id, "output": output}))
             }
-            Entry::Opaque { api, item } if api == "openai-responses" => input.push(item.clone()),
+            Entry::Opaque { api, model, item } if api == "openai-responses" && model == &cfg.model => input.push(item.clone()),
             Entry::Opaque { .. } => {}
         }
     }
@@ -299,12 +302,17 @@ async fn openai(
             }),
             Some("reasoning") => out.push(Entry::Opaque {
                 api: "openai-responses".into(),
+                model: cfg.model.clone(),
                 item,
             }),
             _ => {}
         }
     }
-    if resp["status"] == "incomplete" && out.is_empty() {
+    if resp["status"] == "incomplete"
+        && !out
+            .iter()
+            .any(|entry| matches!(entry, Entry::Assistant { text, .. } if !text.is_empty()))
+    {
         return Err(format!(
             "incomplete response: {}",
             resp["incomplete_details"]
@@ -350,7 +358,9 @@ async fn anthropic(
                 json!({"type": "tool_result", "tool_use_id": id, "content": output}),
             ),
             // Thinking blocks go back unchanged (with their signature): required next to tool use.
-            Entry::Opaque { api, item } if api == "anthropic-messages" && thinking => {
+            Entry::Opaque { api, model, item }
+                if api == "anthropic-messages" && model == &cfg.model && thinking =>
+            {
                 push("assistant", item.clone())
             }
             Entry::Opaque { .. } => {}
@@ -405,6 +415,7 @@ async fn anthropic(
             }),
             Some("thinking") | Some("redacted_thinking") => out.push(Entry::Opaque {
                 api: "anthropic-messages".into(),
+                model: cfg.model.clone(),
                 item: block,
             }),
             _ => {}