Refactor AI Assist to background-task + buffer architecture

The assist flow previously tied the entire LLM generation to a single
POST request with no keepalives, causing NS_ERROR_NET_PARTIAL_TRANSFER
in Firefox when Hypercorn closed the connection during gaps between
chunks. This refactor decouples generation into a background task with
a buffer and a separate SSE stream — the same pattern used by chat.

- generation_buffer.py: Widen _buffers to support string keys, add
  create/get/remove_assist_buffer() using "assist:{user_id}" keys,
  fix cleanup log format for string keys
- generation_task.py: Add run_assist_generation() — lightweight
  background task with no DB persistence or title generation
- notes.py: Replace single POST SSE route with POST /api/notes/assist
  (returns 202) + GET /api/notes/assist/stream (SSE with 15s keepalives
  and Last-Event-ID reconnection); 409 if already running
- useAssist.ts: Switch from apiStreamPost to apiPost + apiSSEStream
  two-step pattern with named event mapping and stream handle cleanup

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

frontend/src/composables/useAssist.ts

@@ -1,5 +1,5 @@
 import { ref, computed, watch, type Ref } from "vue";
-import { apiStreamPost } from "@/api/client";
+import { apiPost, apiSSEStream, type SSEStreamHandle } from "@/api/client";
 import { useChatStore } from "@/stores/chat";
 import {
   parseMarkdownSections,
@@ -28,6 +28,7 @@ export function useAssist(body: Ref<string>) {
 
   // Snapshot of body at the time a section/selection was chosen
   let bodySnapshot = "";
+  let streamHandle: SSEStreamHandle | null = null;
 
   const target = computed<AssistTarget | null>(() => {
     if (customSelection.value) {
@@ -79,6 +80,10 @@ export function useAssist(body: Ref<string>) {
   }
 
   function clearSelection() {
+    if (streamHandle) {
+      streamHandle.close();
+      streamHandle = null;
+    }
     selectedSection.value = null;
     customSelection.value = null;
     instruction.value = "";
@@ -97,28 +102,33 @@ export function useAssist(body: Ref<string>) {
     error.value = "";
 
     try {
-      await apiStreamPost(
-        "/api/notes/assist",
-        {
-          body: body.value,
-          target_section: target.value.text,
-          instruction: instruction.value,
-        },
-        (data) => {
-          if (data.chunk) {
-            streamingText.value += data.chunk as string;
-          }
-          if (data.done) {
-            proposedText.value = (data.full_text as string) || streamingText.value;
-            state.value = "review";
-          }
-          if (data.error) {
-            error.value = data.error as string;
-            state.value = "idle";
-          }
+      // Step 1: Launch background generation
+      await apiPost("/api/notes/assist", {
+        body: body.value,
+        target_section: target.value.text,
+        instruction: instruction.value,
+      });
+
+      // Step 2: Tail the SSE buffer
+      streamHandle = apiSSEStream("/api/notes/assist/stream", (evt) => {
+        if (evt.event === "chunk") {
+          streamingText.value += evt.data.chunk as string;
         }
-      );
-      // If stream ended without a done event, use accumulated text
+        if (evt.event === "done") {
+          proposedText.value = (evt.data.full_text as string) || streamingText.value;
+          state.value = "review";
+          streamHandle = null;
+        }
+        if (evt.event === "error") {
+          error.value = evt.data.error as string;
+          state.value = "idle";
+          streamHandle = null;
+        }
+      });
+
+      await streamHandle.done;
+
+      // Fallback: if stream ended without done/error, use accumulated text
       if (state.value === "streaming") {
         if (streamingText.value) {
           proposedText.value = streamingText.value;
@@ -126,10 +136,12 @@ export function useAssist(body: Ref<string>) {
         } else {
           state.value = "idle";
         }
+        streamHandle = null;
       }
     } catch (e) {
       error.value = e instanceof Error ? e.message : "Request failed";
       state.value = "idle";
+      streamHandle = null;
     }
   }