@filigran/chatbot 3.16.1 → 3.18.1

package/CHANGELOG.md

@@ -8,7 +8,54 @@ the versions follow [semantic versioning](https://semver.org/spec/v2.0.0.html).
 This file starts at 3.10.0; earlier releases are on the
 [releases page](https://github.com/FiligranHQ/filigran-ui/releases).
 
-## 3.16.1 — unreleased
+## 3.18.1 — unreleased
+
+### Fixed
+
+- A message sent, or a file attached, before the panel's agents have loaded
+  goes to the agent the host asked for (`initialAgent`); it went with no agent
+  before, and the backend answered with the user's default one. The message
+  shows at once and leaves when the agents land, whether from the composer,
+  the handle or a voice call, and Stop lets it go meanwhile. One from the
+  composer or a voice call does not wait for a history still being read back;
+  the handle's `send` still does.
+
+## 3.18.0 — 2026-10-31
+
+### Added
+
+- The status line follows a consult: `consult_progress` frames are read
+  through every consult level, as an agent consulted may consult its own
+  peers. It names the agent consulted and what the agent doing the work does
+  (`Consulting {agent}: {tool}…`, `Consulting {agent}: {tool} (+{count}
+  more)…`, `Consulting {agent}: writing the answer…`), or `Consulting {count}
+  agents…` while several run (`Consulting {agent}: {count} questions…` when
+  they ask one agent), and the consult's elapsed time runs on. Each consult is
+  told apart by the frame's `consult_id`, else by the agent consulted. The
+  consults end when the agent's own loop moves on, whether they answered or
+  not, and a consult frame arriving after that is not shown. Only the agent
+  consulted's own answer reads `writing the answer`, and a consult of a peer
+  in turn shows as that peer's frames, not as `consult_agent`. It said
+  `Thinking...` for the whole consult. A host's `statusLabel` reads the
+  consult as `AgentStatusState.consult` (`ConsultStatus`); a host translating
+  the panel's keys translates the new ones too.
+
+## 3.17.0 — published within 3.18.0
+
+### Added
+
+- `reloadConversation()` on the handle reads the conversation on screen back
+  from the server again, for a change the host made to it outside the panel (a
+  message it stored, another agent taking it over). The messages it added are
+  merged into the thread on screen, which keeps the composer's draft and what
+  only the panel holds; one it edited or deleted shows as it was until the
+  conversation is opened again. The conversation's agent replaces one a turn
+  transferred it to, and the context gauge takes the newest reading. Nothing
+  is read before the conversation is, nor while a turn runs in it; a read-back
+  that fails, retried or not, leaves the thread as it is. `selectConversation`
+  with the id on screen still does nothing.
+
+## 3.16.1 — 2026-10-29
 
 ### Fixed
 

package/README.md

@@ -163,12 +163,13 @@ chat.current?.send('Summarize this report', { extras: { mode: 'deep_investigatio
 | --- | --- |
 | `send(text, { extras, displayText })` | Sends *text* as the user, leaving the composer's draft alone; `extras` join that one message's body, and the user's bubble shows `displayText` in place of *text* (as `onBeforeSend`'s `displayText`, which wins when it gives one). Resolves to `'sent'` (once the backend takes it), `'steered'` (joined the running turn), `'held'` (sent when the turn ends - also, with or without `holdSendsDuringTurn`, while a turn ended provisionally winds down, after `onTurnRunningChange(false)`), `'dropped'` (empty, taken over by `onBeforeSend`, `readOnly`, or the panel unmounted first) or `'failed'` (the backend refused it) |
 | `newConversation({ workspaceId, agent })` | Starts a new chat, in a workspace and with an agent (id or slug) if given |
-| `selectConversation(id)` | Opens a conversation |
+| `selectConversation(id)` | Opens a conversation; the one on screen is left as it is |
 | `setDraft(text)` / `focusComposer()` | Fills / focuses the composer |
 | `getDraft()` | The composer's text as it stands (typed by the user, or set by `setDraft`), e.g. for `onBeforeSend` taking over a held message (`draft.held`) to tell whether the composer still holds what was typed for it |
 | `stop()` | Same as the Stop button; during a voice call it also drops the question awaited and stops speaking |
 | `startVoiceCall()` | Starts a [voice call](#voice-call) over the panel; `false` when this browser has no speech recognition or the panel is `readOnly` |
 | `reloadAgents()` | Reads the agent list again, keeping the agent picked unless it is gone |
+| `reloadConversation()` | Reads the conversation on screen back from the server again, for a change the host made to it outside the panel (a message it stored, another agent taking it over). The messages it added are merged into the thread on screen, which keeps what only the panel holds (earlier pages, a stopped answer, a send that failed) and the composer's draft; one it edited or deleted shows as it was until the conversation is opened again. The conversation's agent replaces one a turn transferred it to, and the context gauge takes the newest reading. Nothing is read before the conversation is, nor while a turn runs in it; a read-back retried after the backend was down or the credentials were refused is the host's too, one that fails leaves the thread as it is, and one answered with a re-created conversation opens that one instead. Does nothing without a conversation, or without the REST sessions route (`legacy`, `ag-ui`, `singleEndpoint`, `sessions: null`) |
 | `scrollToMessage(id)` | Scrolls to a message by its server id and highlights it; `false` when it is not on screen |
 
 #### Host extensions
@@ -747,6 +748,7 @@ Download failures (403/404/5xx/network) are reported through the optional `onDow
 - `tool_start` — Agent is using tools (with `tools` array)
 - `tool_heartbeat` — Liveness signal during a long tool execution (with `tools` and `elapsed_s`). The widget keeps the current status label and renders a live elapsed-time indicator next to it once the execution exceeds ~15 s, so long operations (background tasks, agent consults) never look stuck
 - `wait_start` — The agent pauses (with `seconds` and `reason`): the reason and a countdown show, with `Continue now` when the host names `skipWait`. The pause stays on screen through `wait_status` (what it watches changed state), `tool_heartbeat`, `tool_progress` and `consult_progress`; streamed text or any other status ends it
+- `consult_progress` — A frame of an agent the agent consulted (with `agent_id`, `agent_name`, `consult_id` and the frame under `inner`); a consulted agent may consult its own peers, each level wrapping the frame once more. The status line names the agent consulted and what the agent doing the work does (its first tool and how many more, or writing the answer), or how many agents are consulted at once (several questions to one agent: how many), and keeps the consult's elapsed time. A consult ends when the agent consulted streams or composes its own answer, and every consult ends when the agent's own loop moves on (a new step, other tools, its answer) or its turn is replayed. Each consult is told apart by its `consult_id`; frames without one (an older backend) by the agent consulted, so two consults of one agent at once show as one
 - `analyzing` — Agent is analyzing tool results
 - `composing` — Agent is composing the response
 - `streaming` — Content is being streamed

package/dist/index.d.ts

@@ -829,6 +829,11 @@ interface ChatPanelHandle {
     startVoiceCall: () => boolean;
     /** Read the agent list again, keeping the agent picked unless it is gone. */
     reloadAgents: () => void;
+    /**
+     * Read the conversation on screen back again, for a change the host made to it. Not `selectConversation`,
+     * which does nothing for the id on screen, so a host's sync with its own state never resets it.
+     */
+    reloadConversation: () => void;
     /** Scroll to a message on screen (by its server id) and highlight it; false when it is not there. */
     scrollToMessage: (messageId: string) => boolean;
 }
@@ -1098,6 +1103,26 @@ interface AgentStatusState {
     waitSeconds?: number;
     waitReason?: string;
     waitStartMs?: number;
+    /** `consulting`: the consult the agent is waiting on. */
+    consult?: ConsultStatus;
+}
+/**
+ * A consult the agent is waiting on, as the status line shows it. A consulted
+ * agent may consult its own peers: the name is the agent the conversation's
+ * agent consulted, the status and tools those of the agent doing the work.
+ */
+interface ConsultStatus {
+    agentName: string;
+    /** The working agent's status: `thinking`, `tool_start`, `streaming`... */
+    status: string;
+    /** Its tools, on `tool_start`. */
+    tools?: string[];
+    /** The agent consulted writes its own answer (not a peer it consulted in turn). */
+    answering?: boolean;
+    /** The consults running at once, this one included. */
+    count: number;
+    /** The agents they consult: one when every consult running asks the same agent. */
+    agents: number;
 }
 interface ChatFile {
     name: string;
@@ -1195,4 +1220,4 @@ interface TransferredAgent {
 declare const voiceCallSupported: () => boolean;
 
 export { ChatPanel, ChatToggleButton, voiceCallSupported };
-export type { AgentStatusState, ApiEndpoints, BackendType, ChatAnsweredBy, ChatArtifact, ChatArtifactKind, ChatAttachment, ChatCodeBlockContext, ChatCodeBlockRenderer, ChatContextBreakdown, ChatContextUsage, ChatConversationRef, ChatConversationReferenceCandidate, ChatConversationSummary, ChatFile, ChatHeldOutcome, ChatMarkerContext, ChatMarkerRenderer, ChatMessage, ChatMessagePersistedFeedback, ChatMessageSlotContext, ChatMode, ChatPanelHandle, ChatPanelMode, ChatPanelProps, ChatPanelTheme, ChatPromptTemplate, ChatQuotaStatus, ChatSendDecision, ChatSendDraft, ChatSendOptions, ChatSendOutcome, ChatStreamEventInfo, ChatToggleButtonProps, ChatTurnEnd, ChatTurnOutcome, ChatWorkspace, MessageFeedback, ToolApprovalDecision, ToolApprovalProposal, ToolApprovalVerdict, TransferredAgent, Translate, XtmAgent };
+export type { AgentStatusState, ApiEndpoints, BackendType, ChatAnsweredBy, ChatArtifact, ChatArtifactKind, ChatAttachment, ChatCodeBlockContext, ChatCodeBlockRenderer, ChatContextBreakdown, ChatContextUsage, ChatConversationRef, ChatConversationReferenceCandidate, ChatConversationSummary, ChatFile, ChatHeldOutcome, ChatMarkerContext, ChatMarkerRenderer, ChatMessage, ChatMessagePersistedFeedback, ChatMessageSlotContext, ChatMode, ChatPanelHandle, ChatPanelMode, ChatPanelProps, ChatPanelTheme, ChatPromptTemplate, ChatQuotaStatus, ChatSendDecision, ChatSendDraft, ChatSendOptions, ChatSendOutcome, ChatStreamEventInfo, ChatToggleButtonProps, ChatTurnEnd, ChatTurnOutcome, ChatWorkspace, ConsultStatus, MessageFeedback, ToolApprovalDecision, ToolApprovalProposal, ToolApprovalVerdict, TransferredAgent, Translate, XtmAgent };