server + ui: SSE Replay Buffer (#23226)

* server: SSE replay buffer, survives client disconnect

Opt in on POST /v1/chat/completions when the client sends
X-Stream-Resume: 1 and a non empty X-Conversation-Id. The conv id is
the session identity end to end, no extra opaque token. The drain
runs detached server side and buffers SSE bytes, the generation
survives HTTP disconnect, F5, or lets users switch from iOS Safari
to another app without losing the actively generated response.

Routes:
  GET    /v1/stream/<conv_id>?from=N       replay
  GET    /v1/streams[?conversation_id=X]   list, drives sidebar spinners
  DELETE /v1/stream/<conv_id>              Stop, idempotent

Router parent fans out to children for list and delete, probes on GET
to route to the owner, fans out DELETE on POST so "one session per
conv" holds across model swaps.

WebUI: the layout snapshots /v1/streams at mount and on
visibilitychange, the sidebar reflects live inferences across all
convs. The chat page reattaches on mount, append vs fresh is detected
from existing content so continue mid stream keeps its prefix.

update_slots: on llama_memory_seq_rm refusal at a deep position, full
clear of the seq and reprefill from zero instead of GGML_ABORT.

OAI strict path unchanged when the opt in headers are absent.

* server: create stream session only after post_tasks succeeds

* server, ui: drop X-Stream-Resume, X-Conversation-Id alone enables the replay buffer

* server: drop magic 17, derive the X-Conversation-Id header length from sizeof at build time

* refactor: address review feedback from ngxson

* server-context: cleaning

* server-stream: fix use-after-free on rd

Guard stop_producer with a shared alive flag, flipped by on_stream_end
before rd dies. Prevents a late cancel (session eviction by a later
POST on the same conv_id, or a DELETE arriving after the producer
ended) from touching a destroyed rd.

* ui: fix cross-conversation contamination

Scope streaming flags per conv so one finishing does not unflag the
others, guard discoverActiveStream against concurrent runs to avoid
duplicate attaches, and stop racing syncRemoteRunningStreams for the
sidebar set.

* server-http: keep request alive in detached SSE drain

The response next() lambda may reach into *request via &req long
after on_complete reset the request shared_ptr. Capture request in
the detached thread so it outlives the drain.

* ui: address review feedback from coder543

Forward Authorization to /v1/stream and /v1/streams fetches, the resumable routes
must obey --api-key like the rest of the API.

Wrap reader.read() in a try/catch, the underlying connection drop rejects with
TypeError instead of resolving done=true, treat it as a premature end of stream
so the existing resume loop kicks in.

Freeze the model at session start in chatStreamingStates.model and thread it
through cancel and resume, the dropdown selection may have changed since the
POST and the server side identity is fixed at that time.

* format

* ui: remove unused selectedModelName

* server-stream: poll session->is_cancelled() in stream_aware_should_stop

Address review feedback from coder543. The cancel propagation through
rd.stop() relies on the slot eventually processing the cancel task and
posting a result that notifies the recv condvar, remove_waiting_task_ids
does not notify directly. Add a defensive poll on session->is_cancelled()
so the producer-side next() loop exits on its next iteration after
cancel() without waiting for the cancel task to round trip through a slot.

* server-stream, ui: replace GET /v1/streams with POST /v1/streams/lookup

Address review feedback from coder543. Listing live sessions leaks the
conversation_id of every concurrent user, which defeats the random UUID
unguessability. The new route takes {conversation_ids: [...]} in the
body and returns matches only for the ids the caller already owns, so
foreign UUIDs stay private. The router fans out the same POST to every
child and aggregates, the WebUI passes the convs visible in its sidebar.

* ui: read conv ids from IndexedDB in syncRemoteRunningStreams

The conversations store is not hydrated yet at +layout onMount, so the
sidebar spinners stayed off for background convs until the user clicked
on them. Read straight from the DB to dodge the init race.

* server-models: deduplicate stream lookup timeouts behind one constant

* ui: extract visibility kick grace into a stream constant, bump to 1000 ms

* make it safer & more simple

* server-stream: survive client disconnect via stream_pipe::finish_producer

After the RAII rewrite the generation stopped the moment the client
disconnected. httplib bails its content provider on the is_peer_alive
check at the top of write_content_chunked, so returning true from the
provider never keeps it producing: the response resets, rd is destroyed
and its task gets cancelled.

Reinstate the disconnect survival inside the pipe. stream_pipe gains
finish_producer, which pumps the response next() into the ring buffer
until the generation ends, and mark_producer_done for the clean wire
end. server-http only triggers them: mark before sink.done on a clean
close, finish in on_complete when the peer left early. No detach, no
stream logic in server-http beyond the trigger, and the strict OAI path
is untouched when no pipe is attached.

Known limitation: finish_producer pumps synchronously on the http
worker, so a disconnected stream keeps its worker busy until the
generation ends. A follow-up will move the drain off the http worker so
no worker is held.

* server-stream: drain disconnected streams on a manager owned thread

The previous commit pumped the post disconnect drain synchronously in
on_complete, on the http worker, so a disconnected stream kept its
worker busy until the generation ended. Under a wave of reloads or tab
closes that pins workers from the pool.

Move the drain off the http worker. on_complete now hands the response
to stream_session_manager::adopt_orphan, which pumps it to completion on
a manager owned thread and releases the worker at once. One thread per
disconnected stream still generating, stored in a list, joined and
reaped on the next adopt, by the GC, and at shutdown. No detach, the
thread lifecycle is fully owned by the manager. needs_drain gates the
handoff so a cleanly finished stream never spawns a thread, and the
strict OAI path stays untouched when no pipe is attached.

stop_gc now cancels sessions before finalizing them, so an in flight
drain sees is_cancelled and exits instead of blocking the shutdown join
until the generation ends naturally.

* ui: add missing JSDoc

* server-stream: drain on the http worker, drop the manager thread

Address @ngxson review: httplib runs a large dynamic pool and a worker
blocked in next() sits on a condvar instead of burning cpu, so draining
the rest of the generation on that worker is fine and much simpler than
a dedicated thread.

on_complete calls finish_producer directly again. Removes adopt_orphan,
the orphan thread list and its reaping, the stop_gc session cancel that
only existed to unblock those threads, and the now dead drain_shutdown
flag.

* server-stream: split stream_pipe into producer and consumer classes

Address @ngxson review: one class covering both ends was messy. stream_pipe
is now a base holding the session and is_cancelled, with stream_pipe_producer
(write, mark_producer_done, finish_producer, cleanup, finalizes on destruct)
and stream_pipe_consumer (read only, no finalize) deriving from it.

Drops the is_producer_ discriminator and its runtime guards, the type now
encodes the role. res.spipe is retyped to shared_ptr<stream_pipe_producer>
since it is only ever a producer. No behavior change.

* server-stream: rename producer methods to unix pipe semantics

Address @ngxson review: mark_producer_done becomes done(), finish_producer
becomes close(), matching a unix pipe write end. The producer_done_ member
follows as done_. write() is unchanged. No behavior change.

* server, ui: route resumable streams via a conv map, persist resume identity

Address ngxson review: drop the polling probe, proxy_post records a conv_id ->
model map and the stream routes resolve the owning child with one lookup. The
map is the single source of truth, the ::model suffix stays for child session
uniqueness but the router never parses it.

UI: the server keys a session by the POST time identity (conv::model), but reload
probed with the bare conv id and missed model tagged sessions, so F5 stopped the
stream and sidebar spinners stayed off. Persist the model and rebuild the exact
identity on resume, single conv and bulk sidebar both send it.

Add unit coverage for the identity round trip.

* ui: resolve continue target by id to stop cross-conversation flash on switch

* ui: skip stream resume when the abort is intentional

* server: move the conv id to model map into a self contained tracker

Address review from ngxson: server_models held two mutexes side by side, the
global one and a bare conv_model_mu guarding a loose map, which made the locking
hard to follow. Wrap the map and its lock in a small conv_model_tracker struct
that owns its mutex, one mutex per struct. The remember, lookup and forget
methods move inline into the tracker, server_models exposes a single conv_models
member and the routes call models.conv_models.lookup and friends. No behavior
change, the map stays the single source of truth for routing resumable streams
to a child.

* ui: replace stream magic values with enums and shared constants

Address review from allozaur: lift the inline literals around the resumable
stream code into named symbols so the intent is explicit and reusable.

* ui: fold the stream resume and discovery helpers into ChatService

Address review from allozaur: drop the two standalone stream-*.service files.
They were used only by the chat service and store, carried no shared state, and
did not follow the static class pattern the other services use, so a separate
abstraction was not warranted. Move the helpers onto ChatService as static
methods. No behavior change, tests now exercise them through ChatService.

* docs: document the SSE replay buffer in server README-dev

Add the resumable streaming section, list stream_session_manager in the
backend component inventory, and link PR 23226 in the related PRs.

* ui: align attachServerStream call with onCompletionId param in handleStreamResponse

* server-http: rename del_ to del to match get and post

* ui: address review feedback from allozaur

* ui: drop duplicate SSE constants, keep sse.ts canonical

* ui: use svelte:document for the visibilitychange listener

address review from allozaur: replace the manual document.addEventListener
in onMount with a declarative <svelte:document onvisibilitychange>. svelte
handles attach, detach and SSR, so the typeof document guard and the onMount
cleanup go away. onMount keeps only the first load snapshot.

* server: trim redundant stream drain comments

Address review from ngxson

* server: balance and clean up stream comments

remove redundant comments and tighten the verbose ones across the resumable
stream code, keeping the concurrency and lifetime rationale that is not obvious
from the code. also fix two stale comments in server.cpp and server-models.h
that still described the old ::model suffix probe and fan out routing, now
replaced by the conv_id -> model map

Address review from ngxson

* ui: balance and clean up stream comments

dedup repeated rationale (frozen conv::model identity, the lookup privacy note,
the abort patterns) down to one canonical spot, tighten the verbose blocks, and
keep the concurrency and resume-offset reasoning. fix stale comments in
stream-identity.ts and chat.service.ts that still described the old loopback
probe and fan out routing, now the conv_id -> model map.

---------

Co-authored-by: Xuan Son Nguyen <son@huggingface.co>

tools/ui/src/lib/services/chat.service.ts

@@ -1,6 +1,7 @@
-import { getJsonHeaders } from '$lib/utils/api-headers';
+import { getAuthHeaders, getJsonHeaders } from '$lib/utils/api-headers';
 import { formatAttachmentText } from '$lib/utils/formatters';
 import { isAbortError } from '$lib/utils/abort';
+import { streamIdentity } from '$lib/utils/stream-identity';
 import {
 	ATTACHMENT_LABEL_PDF_FILE,
 	ATTACHMENT_LABEL_MCP_PROMPT,
@@ -13,7 +14,10 @@ import {
 	CONTROL_ACTION,
 	SSE_LINE_SEPARATOR,
 	SSE_DATA_PREFIX,
-	SSE_DONE_MARKER
+	SSE_DONE_MARKER,
+	STREAM_VISIBILITY_KICK_MS,
+	STREAM_RESUME_LOCALSTORAGE_KEY_PREFIX,
+	API_STREAM
 } from '$lib/constants';
 import {
 	AttachmentType,
@@ -21,12 +25,14 @@ import {
 	FileTypeAudio,
 	MessageRole,
 	MimeTypeAudio,
-	ReasoningFormat
+	ReasoningFormat,
+	StreamConnectionState
 } from '$lib/enums';
 import type {
 	ApiChatMessageContentPart,
 	ApiChatMessageData,
-	ApiChatCompletionToolCall
+	ApiChatCompletionToolCall,
+	ApiStreamSession
 } from '$lib/types/api';
 import type {
 	AudioInputFormat,
@@ -54,6 +60,19 @@ function getAudioInputFormat(mimeType: string): AudioInputFormat {
 	return FileTypeAudio.MP3;
 }
 
+interface ResumableStreamState {
+	bytesReceived: number;
+	updatedAt: number;
+
+	// model frozen at POST time, lets a reload rebuild the exact conv::model identity the
+	// server keyed the session under. null when the POST carried no explicit model
+	model?: string | null;
+}
+
+function streamStorageKey(conversationId: string): string {
+	return STREAM_RESUME_LOCALSTORAGE_KEY_PREFIX + conversationId;
+}
+
 export class ChatService {
 	/**
 	 *
@@ -128,6 +147,7 @@ export class ChatService {
 			onChunk,
 			onComplete,
 			onError,
+			onConnectionState,
 			onReasoningChunk,
 			onToolCallChunk,
 			onModel,
@@ -312,9 +332,16 @@ export class ChatService {
 		}
 
 		try {
+			const headers: Record<string, string> = { ...getJsonHeaders() };
+			// tag streaming requests with the conversation id, this single header is the opt in for the
+			// server side replay buffer and powers discoverActiveStream on tab reopen. with an explicit
+			// model the ::model suffix keeps the per model session distinct
+			if (stream && conversationId) {
+				headers['X-Conversation-Id'] = streamIdentity(conversationId, options.model);
+			}
 			const response = await fetch(API_CHAT.COMPLETIONS, {
 				method: 'POST',
-				headers: getJsonHeaders(),
+				headers,
 				body: JSON.stringify(requestBody),
 				signal
 			});
@@ -341,7 +368,9 @@ export class ChatService {
 					onCompletionId,
 					onTimings,
 					conversationId,
-					signal
+					signal,
+					onConnectionState,
+					options.model
 				);
 
 				return;
@@ -473,6 +502,116 @@ export class ChatService {
 	 * @param excludeReasoning - Whether to strip reasoning content (should match excludeReasoningFromContext setting)
 	 * @param signal - Optional AbortSignal to cancel the pre-encode request
 	 */
+	static async cancelServerStream(conversationId: string, model?: string | null): Promise<void> {
+		if (!conversationId) return;
+		try {
+			const id = streamIdentity(conversationId, model);
+			await fetch(`${API_STREAM.BASE}/${encodeURIComponent(id)}`, {
+				method: 'DELETE',
+				headers: getAuthHeaders()
+			});
+		} catch (e) {
+			console.warn('cancelServerStream failed:', e);
+		}
+	}
+
+	/**
+	 * Pick the running session to splice into when discoverActiveStream lists candidates for a
+	 * conversation. Finalized sessions are not candidates: their final content was already written
+	 * to the DB by the original onComplete handler, so attaching to them would replay a buffer that
+	 * may not match what the DB holds. A continue session's buffer holds only the appended deltas,
+	 * not the pre continue prefix, so replaying it as a fresh generation would erase the original.
+	 *
+	 * Among running sessions we tie break on the most recent started_at, which covers the case of
+	 * multiple inferences left running on the same conversation.
+	 */
+	static selectActiveStream(
+		sessions: ApiStreamSession[] | null | undefined
+	): ApiStreamSession | null {
+		if (!Array.isArray(sessions) || sessions.length === 0) {
+			return null;
+		}
+		const running = sessions.filter((s) => !s.is_done);
+		if (running.length === 0) {
+			return null;
+		}
+		return running.reduce((best, cur) => (cur.started_at > best.started_at ? cur : best));
+	}
+
+	// persist the running byte count and the frozen model for a conversation, a later visit
+	// resumes the SSE replay at the right offset under the same conv::model identity
+	static saveStreamState(
+		conversationId: string,
+		bytesReceived: number,
+		model?: string | null
+	): void {
+		if (!conversationId) return;
+		try {
+			const state: ResumableStreamState = {
+				bytesReceived,
+				updatedAt: Date.now(),
+				model: model ?? null
+			};
+			localStorage.setItem(streamStorageKey(conversationId), JSON.stringify(state));
+		} catch {
+			// localStorage may be full or disabled, silently ignore
+		}
+	}
+
+	static getStreamState(conversationId: string): ResumableStreamState | null {
+		if (!conversationId) return null;
+		try {
+			const raw = localStorage.getItem(streamStorageKey(conversationId));
+			if (!raw) return null;
+			const parsed = JSON.parse(raw) as ResumableStreamState;
+			if (!parsed || typeof parsed.bytesReceived !== 'number') return null;
+			return parsed;
+		} catch {
+			return null;
+		}
+	}
+
+	static clearStreamState(conversationId: string): void {
+		if (!conversationId) return;
+		try {
+			localStorage.removeItem(streamStorageKey(conversationId));
+		} catch {
+			// nothing to do
+		}
+	}
+
+	/**
+	 * Rebuild the stream identity for a resume. The model persisted at POST time wins, including a
+	 * stored null which means the POST carried no explicit model so the identity stays the bare conv
+	 * id. Only fall back to the caller supplied current model when nothing was persisted.
+	 */
+	static resumeStreamIdentity(
+		conversationId: string,
+		state: ResumableStreamState | null,
+		fallbackModel: string | null
+	): string {
+		const model = state && state.model !== undefined ? state.model : fallbackModel;
+		return streamIdentity(conversationId, model);
+	}
+
+	/**
+	 * Reconnect to an interrupted stream for this conversation. Returns the fetch Response so the
+	 * existing SSE parser drains it like a fresh stream. The server returns 200 on success, 404 if
+	 * no session exists for the conv_id, and 400 if the offset is below the dropped prefix.
+	 */
+	static async resumeStream(
+		conversationId: string,
+		signal?: AbortSignal,
+		model?: string | null
+	): Promise<Response | null> {
+		if (!conversationId) return null;
+		const state = ChatService.getStreamState(conversationId);
+		const from = state?.bytesReceived ?? 0;
+		const id = streamIdentity(conversationId, model);
+		const url = `${API_STREAM.BASE}/${encodeURIComponent(id)}?from=${from}`;
+		return await fetch(url, { method: 'GET', signal, headers: getAuthHeaders() });
+	}
+
 	static async preEncode(
 		messages: ApiChatMessageData[] | (DatabaseMessage & { extra?: DatabaseMessageExtra[] })[],
 		model?: string | null,
@@ -557,7 +696,7 @@ export class ChatService {
 	 * @returns {Promise<void>} Promise that resolves when streaming is complete
 	 * @throws {Error} if the stream cannot be read or parsed
 	 */
-	private static async handleStreamResponse(
+	static async handleStreamResponse(
 		response: Response,
 		onChunk?: (chunk: string) => void,
 		onComplete?: (
@@ -573,15 +712,34 @@ export class ChatService {
 		onCompletionId?: (id: string) => void,
 		onTimings?: (timings?: ChatMessageTimings, promptProgress?: ChatMessagePromptProgress) => void,
 		conversationId?: string,
-		abortSignal?: AbortSignal
+		abortSignal?: AbortSignal,
+		onConnectionState?: (state: StreamConnectionState) => void,
+		streamModel?: string | null
 	): Promise<void> {
-		const reader = response.body?.getReader();
+		let reader = response.body?.getReader();
 
 		if (!reader) {
 			throw new Error('No response body');
 		}
 
-		const decoder = new TextDecoder();
+		// bytesParsed is the absolute server side buffer offset of the next byte to parse
+		// segmentStartOffset is the absolute offset where the current reader started, reset on resume
+		// segmentBytesRead is wire bytes read by the current reader
+		let bytesParsed = 0;
+		let segmentStartOffset = 0;
+		let segmentBytesRead = 0;
+		let lastByteAt = Date.now();
+		// each resume must produce at least one byte to be retried again
+		// if a resume returns 200 but yields nothing, we abandon
+		// since the session has a bounded size, the total number of retries is bounded by construction
+		let madeProgress = true;
+		const encoder = new TextEncoder();
+		if (conversationId) {
+			ChatService.saveStreamState(conversationId, 0, streamModel);
+		}
+		onConnectionState?.(StreamConnectionState.STREAMING);
+
+		let decoder = new TextDecoder();
 		let aggregatedContent = '';
 		let fullReasoningContent = '';
 		let aggregatedToolCalls: ApiChatCompletionToolCall[] = [];
@@ -633,84 +791,180 @@ export class ChatService {
 			}
 		};
 
+		const onVisibilityChange = () => {
+			if (typeof document === 'undefined') return;
+			if (document.visibilityState !== 'visible') return;
+			if (streamFinished) return;
+			if (!conversationId) return;
+			// the bytes have been quiet for too long, the OS likely killed the socket
+			// kicking the reader unblocks reader.read with done=true so the outer loop can resume
+			if (Date.now() - lastByteAt > STREAM_VISIBILITY_KICK_MS) {
+				reader!.cancel().catch(() => {});
+			}
+		};
+		if (typeof document !== 'undefined') {
+			document.addEventListener('visibilitychange', onVisibilityChange);
+		}
+
 		try {
 			let chunk = '';
+			// outer loop drives the resume cycle, swaps reader on premature end of stream
 			while (true) {
-				if (abortSignal?.aborted) break;
-
-				const { done, value } = await reader.read();
-				if (done) break;
-
-				if (abortSignal?.aborted) break;
-
-				chunk += decoder.decode(value, { stream: true });
-				const lines = chunk.split(SSE_LINE_SEPARATOR);
-				chunk = lines.pop() || '';
-
-				for (const line of lines) {
+				while (true) {
 					if (abortSignal?.aborted) break;
 
-					if (line.startsWith(SSE_DATA_PREFIX)) {
-						const data = line.slice(SSE_DATA_PREFIX.length).trim();
-						if (data === SSE_DONE_MARKER) {
-							streamFinished = true;
-
-							continue;
+					let done: boolean;
+					let value: Uint8Array | undefined;
+					try {
+						const r = await reader.read();
+						done = r.done;
+						value = r.value;
+					} catch (readErr) {
+						// reader.read() rejects with TypeError when the underlying connection drops
+						// instead of just resolving with done=true. treat it like done so the outer
+						// loop swaps reader via the resume path
+						if (isAbortError(readErr)) {
+							throw readErr;
 						}
+						console.warn('reader.read() rejected, treating as premature end:', readErr);
+						done = true;
+						value = undefined;
+					}
+					if (done) break;
 
-						try {
-							const parsed: ApiChatCompletionStreamChunk = JSON.parse(data);
-							const choice = parsed.choices?.[0];
-							const content = choice?.delta?.content;
-							const reasoningContent = choice?.delta?.reasoning_content;
-							const toolCalls = choice?.delta?.tool_calls;
-							const timings = parsed.timings;
-							const promptProgress = parsed.prompt_progress;
+					if (abortSignal?.aborted) break;
 
-							const chunkModel = ChatService.extractModelName(parsed);
-							if (chunkModel && !modelEmitted) {
-								modelEmitted = true;
-								onModel?.(chunkModel);
-							}
-
-							if (parsed.id && !idEmitted) {
-								idEmitted = true;
-								onCompletionId?.(parsed.id);
-							}
-
-							if (promptProgress) {
-								ChatService.notifyTimings(undefined, promptProgress, onTimings);
-							}
-
-							if (timings) {
-								ChatService.notifyTimings(timings, promptProgress, onTimings);
-								lastTimings = timings;
-							}
-
-							if (content) {
-								finalizeOpenToolCallBatch();
-								aggregatedContent += content;
-								if (!abortSignal?.aborted) {
-									onChunk?.(content);
-								}
-							}
-
-							if (reasoningContent) {
-								finalizeOpenToolCallBatch();
-								fullReasoningContent += reasoningContent;
-								if (!abortSignal?.aborted) {
-									onReasoningChunk?.(reasoningContent);
-								}
-							}
-
-							processToolCallDelta(toolCalls);
-						} catch (e) {
-							console.error('Error parsing JSON chunk:', e);
+					if (value && value.byteLength > 0) {
+						segmentBytesRead += value.byteLength;
+						lastByteAt = Date.now();
+						if (!madeProgress) {
+							madeProgress = true;
+							onConnectionState?.(StreamConnectionState.STREAMING);
 						}
 					}
+
+					chunk += decoder.decode(value, { stream: true });
+					const lines = chunk.split(SSE_LINE_SEPARATOR);
+					chunk = lines.pop() || '';
+
+					// the persisted offset must point right after the last fully parsed line,
+					// the trailing `chunk` is partial bytes still waiting for a newline
+					if (conversationId) {
+						const tailBytes = encoder.encode(chunk).byteLength;
+						bytesParsed = segmentStartOffset + segmentBytesRead - tailBytes;
+						ChatService.saveStreamState(conversationId, bytesParsed, streamModel);
+					}
+
+					for (const line of lines) {
+						if (abortSignal?.aborted) break;
+
+						if (line.startsWith(SSE_DATA_PREFIX)) {
+							const data = line.slice(SSE_DATA_PREFIX.length).trim();
+							if (data === SSE_DONE_MARKER) {
+								streamFinished = true;
+
+								continue;
+							}
+
+							try {
+								const parsed: ApiChatCompletionStreamChunk = JSON.parse(data);
+								const choice = parsed.choices?.[0];
+								const content = choice?.delta?.content;
+								const reasoningContent = choice?.delta?.reasoning_content;
+								const toolCalls = choice?.delta?.tool_calls;
+								const timings = parsed.timings;
+								const promptProgress = parsed.prompt_progress;
+
+								const chunkModel = ChatService.extractModelName(parsed);
+								if (chunkModel && !modelEmitted) {
+									modelEmitted = true;
+									onModel?.(chunkModel);
+								}
+
+								if (parsed.id && !idEmitted) {
+									idEmitted = true;
+									onCompletionId?.(parsed.id);
+								}
+
+								if (promptProgress) {
+									ChatService.notifyTimings(undefined, promptProgress, onTimings);
+								}
+
+								if (timings) {
+									ChatService.notifyTimings(timings, promptProgress, onTimings);
+									lastTimings = timings;
+								}
+
+								if (content) {
+									finalizeOpenToolCallBatch();
+									aggregatedContent += content;
+									if (!abortSignal?.aborted) {
+										onChunk?.(content);
+									}
+								}
+
+								if (reasoningContent) {
+									finalizeOpenToolCallBatch();
+									fullReasoningContent += reasoningContent;
+									if (!abortSignal?.aborted) {
+										onReasoningChunk?.(reasoningContent);
+									}
+								}
+
+								processToolCallDelta(toolCalls);
+							} catch (e) {
+								console.error('Error parsing JSON chunk:', e);
+							}
+						}
+					}
+
+					if (abortSignal?.aborted) break;
+					if (streamFinished) break;
 				}
 
+				// inner reader done, decide whether to try a resume
 				if (abortSignal?.aborted) break;
+				if (streamFinished) break;
+				if (!conversationId) break;
+
+				if (!madeProgress) {
+					onConnectionState?.(StreamConnectionState.LOST);
+					onError?.(new Error('Stream resume produced no new bytes, giving up'));
+					break;
+				}
+
+				onConnectionState?.(StreamConnectionState.RESUMING);
+				madeProgress = false;
+
+				// the server resends starting at bytesParsed, discard any partial line we held, it
+				// will be retransmitted from a clean line boundary. reuse the frozen model, not the
+				// live dropdown
+				const resumeResp = await ChatService.resumeStream(
+					conversationId,
+					abortSignal,
+					streamModel
+				).catch(() => null);
+				// an abort landing during the resume request is intentional, not a lost connection
+				if (abortSignal?.aborted) break;
+				if (!resumeResp || resumeResp.status !== 200) {
+					onConnectionState?.(StreamConnectionState.LOST);
+					onError?.(new Error('Stream connection lost and could not be resumed'));
+					break;
+				}
+				const newReader = resumeResp.body?.getReader();
+				if (!newReader) break;
+
+				try {
+					reader.releaseLock();
+				} catch {
+					/* ignore */
+				}
+				reader = newReader;
+				decoder = new TextDecoder();
+				chunk = '';
+				segmentStartOffset = bytesParsed;
+				segmentBytesRead = 0;
+				lastByteAt = Date.now();
 			}
 
 			if (abortSignal?.aborted) return;
@@ -718,6 +972,10 @@ export class ChatService {
 			if (streamFinished) {
 				finalizeOpenToolCallBatch();
 
+				if (conversationId) {
+					ChatService.clearStreamState(conversationId);
+				}
+
 				const finalToolCalls =
 					aggregatedToolCalls.length > 0 ? JSON.stringify(aggregatedToolCalls) : undefined;
 
@@ -735,7 +993,14 @@ export class ChatService {
 
 			throw err;
 		} finally {
-			reader.releaseLock();
+			if (typeof document !== 'undefined') {
+				document.removeEventListener('visibilitychange', onVisibilityChange);
+			}
+			try {
+				reader.releaseLock();
+			} catch {
+				/* ignore */
+			}
 		}
 	}
 

tools/ui/src/lib/services/mcp.service.ts

@@ -628,19 +628,20 @@ export class MCPService {
 		);
 
 		const runtimeErrorHandler = (error: Error) => {
-			// Ignore errors that are expected when the SDK's transport is closed,
-			// or when connecting to servers that don't support SSE (stateless-only
-			// endpoints returning 405). The SDK wraps the original AbortError in
-			// a new Error with the message "SSE stream disconnected: AbortError",
-			// and also produces "Cannot cancel a stream locked by a reader".
-			// DOMException is thrown by the browser when aborting fetch requests.
-			const msg = error.message || String(error);
+			// the SDK reports any post initialize error here, including the abort we trigger
+			// ourselves on the next health check cycle, on tab unload, or on server teardown.
+			// these are lifecycle aborts, not actionable errors, so we keep them out of the red console.
+			// the SDK wraps the original AbortError in a generic Error like
+			//   "SSE stream disconnected: AbortError: The operation was aborted."
+			// which isAbortError cannot recognize by name alone, so we also pattern match on the message
+			if (isAbortError(error)) {
+				return;
+			}
+			const msg = error?.message ?? '';
 			if (
-				error.name === 'AbortError' ||
-				error instanceof DOMException ||
-				msg.includes('SSE stream disconnected') ||
-				msg.includes('stream locked by a reader') ||
-				msg.includes('The operation was aborted')
+				/SSE stream disconnected:.*AbortError/i.test(msg) ||
+				/AbortError: .*aborted/i.test(msg) ||
+				/stream locked by a reader/i.test(msg)
 			) {
 				return;
 			}