Skip to content

armgit / eve

Public Imported from GitHub
0

The Open Framework for Building Agents

feat(slack)!: clearer task card and typing labels (#4054)

Andrew Barbac09c0d669488
diff --git a/.changeset/slack-task-labels.md b/.changeset/slack-task-labels.md
new file mode 100644
index 000000000..d483a9e2b
--- /dev/null
+++ b/.changeset/slack-task-labels.md
@@ -0,0 +1,5 @@
+---
+"eve": minor
+---
+
+Slack task cards and typing indicators name work more clearly. A task the agent hands to its own copy shows its brief instead of `agent: …`. Row titles and results use the first sentence, and results are cut to about 100 characters at a word boundary. The typing indicator shows the same labels as the card and says `Waiting on 3 tasks...` instead of naming one task and counting the rest. The `describeActionRequest` and `describeActionRequests` exports from `eve/channels/slack` are removed. A renderer can call `next()` for eve's default status, or read `presentation` from `actions.requested`.
diff --git a/docs/channels/slack.mdx b/docs/channels/slack.mdx
index 5272bb8c9..1b64619fb 100644
--- a/docs/channels/slack.mdx
+++ b/docs/channels/slack.mdx
@@ -454,19 +454,19 @@ The bases say where a call may go. `apiBaseUrl` is the base Slack Web API method
eve shows an agent's progress in two places. Short work, such as thinking or a tool call, appears in the thread status and never posts a message. [Tasks](/docs/tools/tasks), which keep working while the conversation continues, appear in a task card: one message per turn that eve updates in place. Replies post as ordinary messages, so the thread holds the conversation, the task card, and nothing else.
-| Moment | Thread status |
-| ---------------------------------- | ----------------------------------------------------------------------------- |
-| A mention or DM arrives | `Thinking...`, right away, while the message hook is still deciding |
-| The message hook drops the message | Cleared |
-| The turn starts | `Working...` |
-| The model reasons | The first line of its reasoning |
-| The model calls tools | Its narration before the calls, or an action label such as `grep useEveAgent` |
-| The turn waits on its tasks | `Waiting on researcher...`, or `Waiting on researcher and 2 more tasks...` |
-| A reply posts | Cleared by Slack |
+| Moment | Thread status |
+| ---------------------------------- | ------------------------------------------------------------------------------- |
+| A mention or DM arrives | `Thinking...`, right away, while the message hook is still deciding |
+| The message hook drops the message | Cleared |
+| The turn starts | `Working...` |
+| The model reasons | The first line of its reasoning |
+| The model calls tools | Its narration before the calls, or an action label such as `Search useEveAgent` |
+| The turn waits on its tasks | `Waiting on researcher...`, or `Waiting on 3 tasks...` |
+| A reply posts | Cleared by Slack |
Slack clears a thread status after two minutes without a message, so during a long wait the task card is what shows the work is still going.
-Action labels are the tool name plus its most telling argument, the agent name for agent calls, and `+N more` when the model requests several actions at once. They leave out the model's own `task_wait` and `task_cancel` calls. Reasoning snippets build progressively: extensions of at least four characters appear immediately, while smaller streamed deltas use a five-second refresh interval to avoid one Slack request per token.
+Action labels follow the same rule as the task card titles described below, plus `+N more` when the model requests several actions at once. They leave out the model's own `task_wait` and `task_cancel` calls. Reasoning snippets build progressively: extensions of at least four characters appear immediately, while smaller streamed deltas use a five-second refresh interval to avoid one Slack request per token.
Replies up to 12,000 characters render inline. Longer replies are uploaded unchanged as a Markdown snippet named `eve-response.md` with a short in-thread note, which requires the `files:write` bot scope. Upload failures are logged by the channel dispatcher; they do not trigger a model retry.
@@ -480,9 +480,9 @@ When a turn starts its first task, eve posts a task card in the thread and keeps
Each row shows:
-- **Title:** the tool's [`label.start`](/docs/tools#label-tool-activity), the agent name and the first line of its brief for an agent task (`researcher: Find the incidents behind the checkout spike`), or the tool name.
+- **Title:** the tool's [`label.start`](/docs/tools#label-tool-activity), or the tool name. An agent task shows the agent name and the first sentence of its brief (`researcher: Find the incidents behind the checkout spike`). A task the agent hands to its own copy shows the brief alone.
- **Waiting:** while the task waits on an approval, an answer, or a sign-in, the row says so, as in `Waiting for approval`. In a DM or private channel it also shows the request's prompt; elsewhere it shows only what kind of request is waiting, since `approvalChannel` can send the request to one person. The request itself posts wherever `approvalChannel` sends it, and the row goes back to working once the request resolves.
-- **How it ended:** the first line of the result or the agent's reply, `Failed` for a failure, or `Stopped` for a cancelled task. Failed and stopped rows never show a success check. In a DM or private channel, a failure also shows the first line of its error, as in `Failed: Rate limited`. Error text can carry internal details, so everywhere else, including proactive sessions without a known `audience`, shows `Failed` alone.
+- **How it ended:** the first sentence of the result or the agent's reply, cut to about 100 characters, `Failed` for a failure, or `Stopped` for a cancelled task. Failed and stopped rows never show a success check. In a DM or private channel, a failure also shows the first sentence of its error, as in `Failed: Rate limited`. Error text can carry internal details, so everywhere else, including proactive sessions without a known `audience`, shows `Failed` alone.
The plan title summarizes the turn: `Working on 2 tasks`, `1 of 2 tasks done`, `Waiting for approval`, then `Finished 2 tasks` or `Finished 3 tasks: 1 failed, 1 stopped`. A resumable task continued in a later turn appears in that turn's card.
diff --git a/packages/eve/src/channel/task-card.ts b/packages/eve/src/channel/task-card.ts
index 2e478c8d5..6bdad0b48 100644
--- a/packages/eve/src/channel/task-card.ts
+++ b/packages/eve/src/channel/task-card.ts
@@ -1,9 +1,10 @@
import { isTaskControlTool } from "#protocol/task-tools.js";
import type { UnstampedMessageStreamEvent } from "#protocol/message.js";
+import { actionRequestName } from "#shared/action-request-name.js";
import type { RuntimeActionRequest } from "#shared/action-types.js";
import type { ChannelAudience } from "#shared/channel-audience.js";
import type { JsonObject, JsonValue } from "#shared/json.js";
-import { normalizePresentationText } from "#shared/presentation-text.js";
+import { firstSentence, normalizePresentationText } from "#shared/presentation-text.js";
/** A call's input travels to renderers only up to this many characters of JSON. */
const MAX_ACTION_INPUT_LENGTH = 4_096;
@@ -32,13 +33,16 @@ export interface TaskCardTask {
readonly kind: "agent" | "tool";
/** The tool or agent name. */
readonly name: string;
- /** The tool's start label, `agent: brief` for an agent, or the name. */
+ /**
+ * The tool's start label, or the name. An agent's is `researcher: brief`,
+ * or the brief alone for a call to the agent's own copy.
+ */
readonly title: string;
/** `blocked` while the task waits on a person; see {@link TaskCardTask.blockedOn}. */
readonly status: TaskCardStatus | "blocked";
readonly blockedOn?: TaskCardBlocker;
/**
- * One line describing the result once settled. A failure's line appears
+ * The first sentence of the result once settled. A failure's line appears
* only in a private conversation, since error text can carry internals.
*/
readonly summary?: string;
@@ -305,7 +309,7 @@ function requestedCall(
label: string | undefined,
at: string,
): TrackedCall {
- const name = actionName(action);
+ const name = actionRequestName(action);
const call: { -readonly [K in keyof TrackedCall]: TrackedCall[K] } = {
callId: action.callId,
name,
@@ -381,19 +385,6 @@ function bounded(calls: readonly TrackedCall[]): readonly TrackedCall[] {
});
}
-function actionName(action: RuntimeActionRequest): string {
- switch (action.kind) {
- case "load-skill":
- return "load_skill";
- case "subagent-call":
- return action.subagentName;
- case "remote-agent-call":
- return action.remoteAgentName;
- default:
- return action.toolName;
- }
-}
-
function actionStatus(status: string): TaskCardStatus {
switch (status) {
case "completed":
@@ -418,9 +409,9 @@ function taskSummary(data: {
}): string | undefined {
switch (data.status) {
case "completed":
- return firstLine(resultText(data.output));
+ return brief(resultText(data.output));
case "failed":
- return firstLine(data.error?.message);
+ return brief(data.error?.message);
default:
return undefined;
}
@@ -434,13 +425,8 @@ function resultText(output: JsonValue | undefined): unknown {
return fields.message ?? fields.text ?? fields.summary;
}
-function firstLine(value: unknown): string | undefined {
- if (typeof value !== "string") return undefined;
- const line = value
- .split(/\r?\n/u)
- .map((candidate) => candidate.trim())
- .find((candidate) => candidate.length > 0);
- return presentationText(line);
+function brief(value: unknown): string | undefined {
+ return typeof value === "string" ? firstSentence(value) : undefined;
}
function presentationText(value: string | undefined): string | undefined {
diff --git a/packages/eve/src/client/message-action-parts.ts b/packages/eve/src/client/message-action-parts.ts
index 3d5ea7447..e0d8b119d 100644
--- a/packages/eve/src/client/message-action-parts.ts
+++ b/packages/eve/src/client/message-action-parts.ts
@@ -1,3 +1,4 @@
+import { actionRequestName } from "#shared/action-request-name.js";
import type { RuntimeActionRequest, RuntimeActionResult } from "#shared/action-types.js";
import type { InputRequest } from "#shared/input.js";
import type {
@@ -94,32 +95,16 @@ export function approvedApproval(part: EveDynamicToolPart | undefined):
/** Maps a runtime action request onto its normalized tool descriptor. */
export function normalizeActionRequest(action: RuntimeActionRequest): ActionDescriptor {
+ const name = actionRequestName(action);
switch (action.kind) {
case "load-skill":
- return {
- kind: "load-skill",
- name: "load_skill",
- toolName: "eve:load-skill",
- };
+ return { kind: "load-skill", name, toolName: "eve:load-skill" };
case "tool-call":
case "workflow-tool-call":
- return {
- kind: "tool-call",
- name: action.toolName,
- toolName: action.toolName,
- };
+ return { kind: "tool-call", name, toolName: name };
case "subagent-call":
- return {
- kind: "subagent-call",
- name: action.subagentName,
- toolName: `eve:subagent:${action.subagentName}`,
- };
case "remote-agent-call":
- return {
- kind: "subagent-call",
- name: action.remoteAgentName,
- toolName: `eve:subagent:${action.remoteAgentName}`,
- };
+ return { kind: "subagent-call", name, toolName: `eve:subagent:${name}` };
}
}
diff --git a/packages/eve/src/execution/tasks/task-id-input.test.ts b/packages/eve/src/execution/tasks/task-id-input.test.ts
new file mode 100644
index 000000000..37a24f8e0
--- /dev/null
+++ b/packages/eve/src/execution/tasks/task-id-input.test.ts
@@ -0,0 +1,62 @@
+import { describe, expect, it } from "vitest";
+import { z } from "zod";
+
+import { withTaskIdInput } from "#execution/tasks/task-id-input.js";
+import type { HarnessToolDefinition } from "#harness/execute-tool.js";
+
+function agentTool(name: string): HarnessToolDefinition {
+ return {
+ behavior: {
+ availability: [],
+ handling: {
+ kind: "dispatch",
+ target: {
+ kind: name === "agent" ? "self-agent-call" : "subagent-call",
+ nodeId: "node_1",
+ subagentName: name,
+ },
+ },
+ },
+ description: "Delegate work.",
+ inputSchema: z.object({ message: z.string() }),
+ name,
+ };
+}
+
+describe("agent call label", () => {
+ it.each([
+ [
+ "names a subagent before the first sentence of its brief",
+ "researcher",
+ "Find the March incidents. Include timelines.\nThen rank them.",
+ "researcher: Find the March incidents.",
+ ],
+ [
+ "shows the brief alone for the agent's own copy",
+ "agent",
+ "\nInvestigate issue #4035. Summarize the report.",
+ "Investigate issue #4035.",
+ ],
+ [
+ "keeps versions and abbreviations inside one sentence",
+ "agent",
+ "On [email protected] the build fails, e.g. with an extension mounted. Reproduce it.",
+ "On [email protected] the build fails, e.g. with an extension mounted.",
+ ],
+ [
+ "keeps a leading list marker with its sentence",
+ "agent",
+ "1. Reproduce the build failure on main. Then bisect.",
+ "1. Reproduce the build failure on main.",
+ ],
+ [
+ "keeps a number after an abbreviation inside one sentence",
+ "agent",
+ "See PR No. 4035 for details. Then review it.",
+ "See PR No. 4035 for details.",
+ ],
+ ["falls back to the agent's name for an empty brief", "agent", " \n", "agent"],
+ ])("%s", (_name, tool, message, expected) => {
+ expect(withTaskIdInput(agentTool(tool)).label?.start?.({ message })).toBe(expected);
+ });
+});
diff --git a/packages/eve/src/execution/tasks/task-id-input.ts b/packages/eve/src/execution/tasks/task-id-input.ts
index 0ec736a9c..95c470daa 100644
--- a/packages/eve/src/execution/tasks/task-id-input.ts
+++ b/packages/eve/src/execution/tasks/task-id-input.ts
@@ -6,7 +6,8 @@ import {
import { isAgentTool } from "#execution/tasks/tool-entry-point.js";
import type { HarnessToolDefinition } from "#harness/execute-tool.js";
import type { JsonObject } from "#shared/json.js";
-import { normalizePresentationText } from "#shared/presentation-text.js";
+import { firstSentence } from "#shared/presentation-text.js";
+import { AGENT_TOOL_NAME } from "#tools/framework/agent-contract.js";
import { isToolSchema, withOptionalStringProperty } from "#tools/schema.js";
// A `serve` tool's model input is the tool's own input plus an optional
@@ -18,7 +19,7 @@ export const TASK_ID_INPUT = "taskId";
/**
* The tool as the model sees it: its input gains `taskId`, and its description
* explains it. An agent tool also gets its activity label, the agent and the
- * first line of its message: `researcher: Find the March incidents`.
+ * first sentence of its message: `researcher: Find the March incidents`.
*/
export function withTaskIdInput(definition: HarnessToolDefinition): HarnessToolDefinition {
const schema = definition.inputSchema;
@@ -44,22 +45,19 @@ export function withTaskIdInput(definition: HarnessToolDefinition): HarnessToolD
};
}
-/** `researcher: Find the March incidents`, or the agent's name when the message is empty. */
+/**
+ * `researcher: Find the March incidents`, from the first sentence of the
+ * message. A call to the agent's own copy shows the brief alone, since its
+ * name, `agent`, says nothing. The agent's name when the message is empty.
+ */
function agentCallLabel(name: string, input: unknown): string {
const message =
typeof input === "object" && input !== null
? (input as { message?: unknown }).message
: undefined;
- const line = typeof message === "string" ? firstLine(message) : undefined;
- return line === undefined ? name : `${name}: ${line}`;
-}
-
-function firstLine(text: string): string | undefined {
- for (const line of text.split(/\r?\n/u)) {
- const normalized = normalizePresentationText(line);
- if (normalized !== "") return normalized;
- }
- return undefined;
+ const brief = typeof message === "string" ? firstSentence(message) : undefined;
+ if (brief === undefined) return name;
+ return name === AGENT_TOOL_NAME ? brief : `${name}: ${brief}`;
}
/** A `serve` tool call's model input, split into the task it names and the tool's own input. */
diff --git a/packages/eve/src/public/channels/slack/action-status.test.ts b/packages/eve/src/public/channels/slack/action-status.test.ts
deleted file mode 100644
index 7f7588c7f..000000000
--- a/packages/eve/src/public/channels/slack/action-status.test.ts
+++ /dev/null
@@ -1,103 +0,0 @@
-import { describe, expect, it } from "vitest";
-
-import {
- describeActionRequest,
- describeActionRequests,
-} from "#public/channels/slack/action-status.js";
-import type { RuntimeActionRequest } from "#shared/action-types.js";
-import type { JsonObject } from "#shared/json.js";
-
-function toolCall(toolName: string, input: JsonObject = {}): RuntimeActionRequest {
- return { callId: "c1", input, kind: "tool-call", toolName };
-}
-
-describe("describeActionRequest", () => {
- it("labels tool calls with the tool name plus the salient argument", () => {
- expect(
- describeActionRequest(toolCall("grep", { path: "packages", pattern: "useEveAgent" })),
- ).toBe("grep useEveAgent");
- expect(describeActionRequest(toolCall("read_file", { filePath: "agent/agent.ts" }))).toBe(
- "read_file agent/agent.ts",
- );
- expect(describeActionRequest(toolCall("bash", { command: "pnpm test\npnpm lint" }))).toBe(
- "bash pnpm test",
- );
- });
-
- it("keeps the tail of long paths and clips long commands", () => {
- expect(
- describeActionRequest(
- toolCall("read_file", {
- filePath: "/workspace/eve/packages/eve/src/runtime/actions/executor-registry.ts",
- }),
- ),
- ).toBe("read_file actions/executor-registry.ts");
-
- const clipped = describeActionRequest(toolCall("bash", { command: "x".repeat(120) }));
- expect(clipped.length).toBeLessThanOrEqual("bash ".length + 40);
- expect(clipped.endsWith("...")).toBe(true);
- });
-
- it("probes generic keys for tools without a salient-key entry", () => {
- expect(describeActionRequest(toolCall("close_issue", { issueNumber: 42 }))).toBe(
- "close_issue 42",
- );
- });
-
- it("falls back to the bare tool name without a telling argument", () => {
- expect(describeActionRequest(toolCall("list_new_issues"))).toBe("list_new_issues");
- });
-
- it("labels dispatched calls with the target name and skill loads with the skill", () => {
- expect(
- describeActionRequest({
- callId: "c1",
- description: "Reason about the issue",
- input: {},
- kind: "subagent-call",
- name: "reasoner",
- nodeId: "n1",
- subagentName: "reasoner",
- }),
- ).toBe("reasoner");
- expect(
- describeActionRequest({
- callId: "c1",
- description: "Triage remotely",
- input: {},
- kind: "remote-agent-call",
- name: "triage",
- nodeId: "n1",
- remoteAgentName: "triage",
- }),
- ).toBe("triage");
- expect(
- describeActionRequest({ callId: "c1", input: { skill: "arena" }, kind: "load-skill" }),
- ).toBe("load_skill arena");
- expect(
- describeActionRequest({
- callId: "c1",
- input: { path: "agent/agent.ts" },
- kind: "workflow-tool-call",
- toolName: "publish",
- workflowId: "publish-workflow",
- }),
- ).toBe("publish agent/agent.ts");
- });
-});
-
-describe("describeActionRequests", () => {
- it("shows the first label and a count for the rest of the batch", () => {
- expect(
- describeActionRequests([
- toolCall("grep", { pattern: "digest" }),
- toolCall("read_file", { filePath: "a.ts" }),
- toolCall("read_file", { filePath: "b.ts" }),
- ]),
- ).toBe("grep digest +2 more");
- });
-
- it("returns a generic label for an empty batch", () => {
- expect(describeActionRequests([])).toBe("Working...");
- });
-});
diff --git a/packages/eve/src/public/channels/slack/action-status.ts b/packages/eve/src/public/channels/slack/action-status.ts
index 7582282e5..0a1c38936 100644
--- a/packages/eve/src/public/channels/slack/action-status.ts
+++ b/packages/eve/src/public/channels/slack/action-status.ts
@@ -1,100 +1,36 @@
/**
- * Typing-indicator labels for requested actions. The Slack indicator is
- * plain text (`assistant.threads.setStatus` renders no markdown), so the
- * label is the action's name plus its most telling argument — `grep useEve`
- * or `read_file agent/agent.ts` instead of `Running grep...`.
+ * Typing-indicator text for the turn's calls. It uses each call's start label,
+ * the same text its task card row shows, so both surfaces name a call the same
+ * way: `Run pnpm test`, `researcher: Find the March incidents`.
*/
+import type { ActionPresentationByCallId } from "#protocol/message.js";
+import { actionRequestName } from "#shared/action-request-name.js";
import type { RuntimeActionRequest } from "#shared/action-types.js";
-
-/** Argument keys worth surfacing, most telling first, per tool. */
-const SALIENT_KEYS: Readonly<Record<string, readonly string[]>> = {
- bash: ["command"],
- glob: ["pattern"],
- grep: ["pattern", "path"],
- load_skill: ["skill"],
- read_file: ["filePath"],
- web_fetch: ["url"],
- web_search: ["query"],
- write_file: ["filePath"],
-};
-
-/** Fallback keys probed on tools without a {@link SALIENT_KEYS} entry. */
-const GENERIC_KEYS: readonly string[] = [
- "filePath",
- "path",
- "pattern",
- "command",
- "query",
- "name",
- "repo",
- "url",
- "issueNumber",
- "number",
-];
-
-const MAX_ARG_CHARS = 40;
-
-function salientArg(toolName: string, input: Readonly<Record<string, unknown>>): string | null {
- for (const key of SALIENT_KEYS[toolName] ?? GENERIC_KEYS) {
- const value = input[key];
- if (typeof value === "string" && value.trim().length > 0) return value;
- if (typeof value === "number") return String(value);
- }
- return null;
-}
-
-/** Keep the tail of a path: `agent/lib/triage/cards.ts` -> `triage/cards.ts`. */
-function shortenArg(text: string): string {
- const oneLine = text.split(/\r?\n/u, 1)[0]?.trim() ?? "";
- if (oneLine.length <= MAX_ARG_CHARS) return oneLine;
- if (oneLine.includes("/") && !oneLine.includes(" ")) {
- const segments = oneLine.split("/").filter((s) => s.length > 0);
- const tail = segments.slice(-2).join("/");
- if (tail.length <= MAX_ARG_CHARS) return tail;
- }
- return `${oneLine.slice(0, MAX_ARG_CHARS - 3).trimEnd()}...`;
-}
-
-function toolCallLabel(toolName: string, input: Readonly<Record<string, unknown>>): string {
- const arg = salientArg(toolName, input);
- return arg === null ? toolName : `${toolName} ${shortenArg(arg)}`;
-}
+import { AGENT_TOOL_NAME } from "#tools/framework/agent-contract.js";
/**
- * One action's typing-indicator label: `grep useEveAgent` for tool calls,
- * the subagent or remote-agent name for dispatched calls, and
- * `load_skill <name>` for skill loads.
+ * The first call's label, or its tool or agent name, plus `+N more` when the
+ * model requested several calls at once.
*/
-export function describeActionRequest(action: RuntimeActionRequest): string {
- switch (action.kind) {
- case "load-skill":
- return toolCallLabel("load_skill", action.input);
- case "remote-agent-call":
- return action.remoteAgentName;
- case "subagent-call":
- return action.subagentName;
- case "tool-call":
- case "workflow-tool-call":
- return toolCallLabel(action.toolName, action.input);
- }
-}
-
-/**
- * Typing-indicator text for one requested batch: the first action's
- * {@link describeActionRequest} label, plus `+N more` when the model
- * requested several actions at once.
- */
-export function describeActionRequests(actions: readonly RuntimeActionRequest[]): string {
+export function describeActionRequests(
+ actions: readonly RuntimeActionRequest[],
+ presentation: ActionPresentationByCallId | undefined,
+): string {
const [first] = actions;
if (first === undefined) return "Working...";
- const label = describeActionRequest(first);
- return actions.length === 1 ? label : `${label} +${actions.length - 1} more`;
+ const label = presentation?.[first.callId]?.label ?? actionRequestName(first);
+ return actions.length === 1 ? label : `${label} +${String(actions.length - 1)} more`;
}
-/** `Waiting on researcher...`, `Waiting on researcher and deploy...`, or `... and 2 more tasks...`. */
+/**
+ * `Waiting on researcher...` for one named task, otherwise a count such as
+ * `Waiting on 3 tasks...`. The agent's own copy is never named, since its
+ * name, `agent`, says nothing.
+ */
export function waitingOnTasks(names: readonly string[]): string {
- const [first, second] = names;
- if (names.length === 1) return `Waiting on ${first!}...`;
- if (names.length === 2) return `Waiting on ${first!} and ${second!}...`;
- return `Waiting on ${first!} and ${String(names.length - 1)} more tasks...`;
+ const [only] = names;
+ if (names.length === 1 && only !== AGENT_TOOL_NAME) return `Waiting on ${only!}...`;
+ return names.length === 1
+ ? "Waiting on a task..."
+ : `Waiting on ${String(names.length)} tasks...`;
}
diff --git a/packages/eve/src/public/channels/slack/defaults.ts b/packages/eve/src/public/channels/slack/defaults.ts
index 51be58abe..8461765cc 100644
--- a/packages/eve/src/public/channels/slack/defaults.ts
+++ b/packages/eve/src/public/channels/slack/defaults.ts
@@ -477,7 +477,9 @@ export const defaultEvents: SlackChannelInternalEvents = {
(action) => action.kind !== "tool-call" || !isTaskControlTool(action.toolName),
);
if (actions.length === 0) return;
- await channel.thread.startTyping(truncateTypingStatus(describeActionRequests(actions)));
+ await channel.thread.startTyping(
+ truncateTypingStatus(describeActionRequests(actions, event.presentation)),
+ );
},
async "message.completed"(event, channel, _ctx) {
diff --git a/packages/eve/src/public/channels/slack/index.ts b/packages/eve/src/public/channels/slack/index.ts
index 7f5ad4965..69ef2c6b0 100644
--- a/packages/eve/src/public/channels/slack/index.ts
+++ b/packages/eve/src/public/channels/slack/index.ts
@@ -92,11 +92,6 @@ export type {
TaskCardView,
} from "#channel/task-card.js";
-export {
- describeActionRequest,
- describeActionRequests,
-} from "#public/channels/slack/action-status.js";
-
export {
loadThreadContextMessages,
type LoadThreadContextMessagesOptions,
diff --git a/packages/eve/src/public/channels/slack/slackChannel.test.ts b/packages/eve/src/public/channels/slack/slackChannel.test.ts
index 0079f715f..b208dcd91 100644
--- a/packages/eve/src/public/channels/slack/slackChannel.test.ts
+++ b/packages/eve/src/public/channels/slack/slackChannel.test.ts
@@ -695,7 +695,7 @@ describe("slackChannel() default event handlers", () => {
expect(fetchMock).toHaveBeenCalledTimes(method === "files.getUploadURLExternal" ? 1 : 3);
});
- it("names the working tasks in the status while a turn waits on them", async () => {
+ it("names one working task in the status while a turn waits, and counts several", async () => {
const adapter = withState(
getAdapter(slackChannel({ credentials: { botToken: "xoxb-test" } })),
THREAD_STATE,
@@ -730,10 +730,23 @@ describe("slackChannel() default event handlers", () => {
ctx,
);
await callEvent(adapter, waiting, ctx);
+ await callEvent(
+ adapter,
+ makeEvent("task.settled", {
+ callId: "c2",
+ status: "completed",
+ taskId: "reviewer-1",
+ turnId: "t1",
+ }),
+ ctx,
+ );
+ await callEvent(adapter, taskStarted("c4", "agent"), ctx);
+ await callEvent(adapter, waiting, ctx);
expect(slackStatuses(fetchMock)).toEqual([
- "Waiting on researcher and 2 more tasks...",
+ "Waiting on 3 tasks...",
"Waiting on reviewer...",
+ "Waiting on a task...",
]);
});
@@ -846,7 +859,7 @@ describe("slackChannel() default event handlers", () => {
expect(ctx.state.pendingToolCallMessage).toBeNull();
});
- it("keeps the model's own task calls out of the typing status", async () => {
+ it("shows the first call's start label in the typing status, without the model's own task calls", async () => {
const adapter = withState(
getAdapter(slackChannel({ credentials: { botToken: "xoxb-test" } })),
THREAD_STATE,
@@ -854,12 +867,13 @@ describe("slackChannel() default event handlers", () => {
const ctx = buildAdapterContext(adapter, stubAccessor());
const requested = (toolNames: readonly string[]) =>
makeEvent("actions.requested", {
- actions: toolNames.map((toolName, index) => ({
- callId: `call_${String(index)}`,
+ actions: toolNames.map((toolName) => ({
+ callId: `call_${toolName}`,
input: {},
kind: "tool-call" as const,
toolName,
})),
+ presentation: { call_search: { label: "Search checkout incidents" } },
sequence: 0,
stepIndex: 0,
turnId: "t1",
@@ -868,11 +882,9 @@ describe("slackChannel() default event handlers", () => {
await callEvent(adapter, requested(["task_wait"]), ctx);
expect(fetchMock).not.toHaveBeenCalled();
- await callEvent(adapter, requested(["task_cancel", "search"]), ctx);
- expect(fetchMock).toHaveBeenCalledTimes(1);
- expect(parseSlackRequestBody(fetchMock.mock.calls[0]![1] as RequestInit)).toMatchObject({
- status: "search",
- });
+ await callEvent(adapter, requested(["task_cancel", "search", "deploy"]), ctx);
+ await callEvent(adapter, requested(["task_cancel", "deploy"]), ctx);
+ expect(slackStatuses(fetchMock)).toEqual(["Search checkout incidents +1 more", "deploy"]);
});
it("message.completed clears typing without posting for empty delivery", async () => {
diff --git a/packages/eve/src/public/channels/slack/task-card.test.ts b/packages/eve/src/public/channels/slack/task-card.test.ts
index bed6e57e6..ceed2eaa2 100644
--- a/packages/eve/src/public/channels/slack/task-card.test.ts
+++ b/packages/eve/src/public/channels/slack/task-card.test.ts
@@ -133,7 +133,8 @@ const TWO_TASKS_STARTED: readonly UnstampedMessageStreamEvent[] = [
const BOTH_TASKS_SETTLED: readonly UnstampedMessageStreamEvent[] = [
createTaskSettledEvent({
callId: DEPLOY_CALL,
- output: "3 deploys; 14:02 changed the cache TTL\nFull log attached.",
+ output:
+ "The 14:02 deploy changed the cache TTL for every storefront route, so checkout requests missed the cache and queued at the origin. Full log attached.",
status: "completed",
taskId: "deploy-4hd8sa",
turnId: TURN_ID,
@@ -193,7 +194,15 @@ describe("Slack task card", () => {
tasks: [
{
output: {
- elements: [{ elements: [{ text: "3 deploys; 14:02 changed the cache TTL" }] }],
+ elements: [
+ {
+ elements: [
+ {
+ text: "The 14:02 deploy changed the cache TTL for every storefront route, so checkout requests missed the…",
+ },
+ ],
+ },
+ ],
},
status: "complete",
},
diff --git a/packages/eve/src/public/channels/slack/task-card.ts b/packages/eve/src/public/channels/slack/task-card.ts
index 229871ae5..69cd64bfc 100644
--- a/packages/eve/src/public/channels/slack/task-card.ts
+++ b/packages/eve/src/public/channels/slack/task-card.ts
@@ -28,7 +28,7 @@ const log = createLogger("slack.task-card");
/** A plan block holds at most 50 tasks. */
const MAX_PLAN_ROWS = 50;
const MAX_TITLE_LENGTH = 80;
-const MAX_LINE_LENGTH = 200;
+const MAX_LINE_LENGTH = 100;
/** Unfinished turns eve keeps tracking; the oldest drop first. */
const MAX_TRACKED_TURNS = 20;
const WRITE_RETRY_MS = 1_000;
@@ -180,8 +180,12 @@ function richText(line: string): BlockKitBlock {
};
}
+/** Cuts at a word boundary when one is near, so a row doesn't end mid-word. */
function truncate(text: string, max: number): string {
- return text.length <= max ? text : `${text.slice(0, max - 1)}…`;
+ if (text.length <= max) return text;
+ const cut = text.slice(0, max - 1);
+ const space = cut.lastIndexOf(" ");
+ return `${(space > max / 2 ? cut.slice(0, space) : cut).trimEnd()}…`;
}
/** Slack task ids must be unique in a plan; call ids are, but can carry characters Slack rejects. */
diff --git a/packages/eve/src/shared/action-request-name.ts b/packages/eve/src/shared/action-request-name.ts
new file mode 100644
index 000000000..7b263a6f5
--- /dev/null
+++ b/packages/eve/src/shared/action-request-name.ts
@@ -0,0 +1,16 @@
+import type { RuntimeActionRequest } from "#shared/action-types.js";
+
+/** The tool, agent, or `load_skill` a request targets. */
+export function actionRequestName(action: RuntimeActionRequest): string {
+ switch (action.kind) {
+ case "load-skill":
+ return "load_skill";
+ case "subagent-call":
+ return action.subagentName;
+ case "remote-agent-call":
+ return action.remoteAgentName;
+ case "tool-call":
+ case "workflow-tool-call":
+ return action.toolName;
+ }
+}
diff --git a/packages/eve/src/shared/presentation-text.ts b/packages/eve/src/shared/presentation-text.ts
index 9c879ac07..a8f3a1bf8 100644
--- a/packages/eve/src/shared/presentation-text.ts
+++ b/packages/eve/src/shared/presentation-text.ts
@@ -11,3 +11,19 @@ export function normalizePresentationText(text: string): string {
.slice(0, MAX_ACTIVITY_TEXT_LENGTH)
);
}
+
+/**
+ * The first sentence of the text's first non-empty line, normalized: a brief
+ * for a message or result. Returns `undefined` when there is no text.
+ */
+export function firstSentence(text: string): string | undefined {
+ for (const line of text.split(/\r?\n/u)) {
+ const normalized = normalizePresentationText(line);
+ if (normalized === "") continue;
+ // A sentence ends at `.`, `!`, or `?` before a capital or quote, so
+ // `e.g. the`, `No. 4035`, and `[email protected]` stay whole. A list marker such
+ // as `1.` never ends one.
+ return normalized.split(/(?<=[.!?])(?<!(?:^|\s)\d{1,2}\.)\s+(?=[\p{Lu}"'`(])/u, 1)[0];
+ }
+ return undefined;
+}