Skip to content

fix: use SDK SessionEvent type for copilot-sdk 0.3.0 compat #100

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
simongdavies merged 1 commit into from
Apr 30, 2026
Merged

fix: use SDK SessionEvent type for copilot-sdk 0.3.0 compat #100

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
236 changes: 118 additions & 118 deletions src/plugin-system/auditor.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,11 @@ import { createWriteStream, mkdirSync, type WriteStream } from "node:fs";
import { createRequire } from "node:module";
import { join } from "node:path";
import { tmpdir } from "node:os";
import { approveAll, type CopilotClient } from "@github/copilot-sdk";
import {
approveAll,
type CopilotClient,
type SessionEvent,
} from "@github/copilot-sdk";
import type {
PluginManifest,
AuditResult,
Expand Down Expand Up @@ -655,130 +659,126 @@ export async function deepAudit(
}
signal?.addEventListener("abort", onAbort, { once: true });

const unsubscribe = auditSession.on(
(event: { type: string; data?: Record<string, unknown> }) => {
// Trace every event type so we can see what's happening
trace(
`event: ${event.type} ${JSON.stringify(event.data ?? {}).slice(0, 300)}`,
);

// Only reset inactivity timer on CONTENT-BEARING events.
// Usage events and other housekeeping must NOT reset the
// timer — otherwise the audit can hang indefinitely getting
// usage events with no actual content.
const isContentEvent =
event.type === "assistant.message" ||
event.type === "assistant.message_delta" ||
event.type === "assistant.reasoning_delta" ||
event.type === "assistant.turn_start" ||
event.type === "assistant.turn_end";

if (isContentEvent) {
clearTimeout(timeoutId);
timeoutId = setTimeout(() => {
unsubscribe();
signal?.removeEventListener("abort", onAbort);
reject(
new Error(
`Audit timed out — no events for ${AUDIT_INACTIVITY_MS / 1000}s`,
),
);
}, AUDIT_INACTIVITY_MS);
}
const unsubscribe = auditSession.on((event: SessionEvent) => {
// Trace every event type so we can see what's happening
trace(
`event: ${event.type} ${JSON.stringify(event.data ?? {}).slice(0, 300)}`,
);

if (event.type === "assistant.message") {
// Final complete message — extract content
const content = (event.data as { content?: string })?.content;
if (content) text = content;
} else if (event.type === "assistant.reasoning_delta") {
// Model is reasoning — forward to progress callback.
// The progress callback handles display logic (e.g.
// suppressing the preview after streaming starts).
const delta = (event.data as { deltaContent?: string })
?.deltaContent;
if (delta) {
progress("reasoning", delta);
}
} else if (event.type === "assistant.message_delta") {
// Streaming delta — accumulate text for the final response.
// Skip whitespace-only deltas for progress display — the
// model can emit "\n\n" before reasoning starts (undocumented
// behaviour). We still accumulate the text but don't trigger
// the "Receiving audit report" UI for whitespace.
const delta = (event.data as { deltaContent?: string })
?.deltaContent;
if (delta) {
text += delta;
if (text.trim().length > 0) {
progress(
"streaming",
`Receiving audit report (${text.length.toLocaleString()} chars)...`,
);
}
}
} else if (
event.type === "assistant.turn_start" ||
event.type === "assistant.turn_end"
) {
// Turn lifecycle events — the CLI server fires these during
// multi-call continuation loops. They're the ONLY reliable
// life sign during dead zones where no message/reasoning
// deltas arrive (known SDK bug — see github/copilot-sdk#524).
// Forward to progress for spinner feedback.
const turnId = (event.data as { turnId?: string })?.turnId;
progress(
"turn",
`${event.type === "assistant.turn_start" ? "Turn" : "Turn complete"} ${turnId ?? ""}`.trim(),
);
} else if (event.type === "assistant.usage") {
// Accumulate usage stats — don't print each one.
// We emit a single aggregate summary at completion.
const d = event.data as {
inputTokens?: number;
outputTokens?: number;
cacheReadTokens?: number;
cost?: number;
duration?: number;
};
totalInputTokens += d.inputTokens ?? 0;
totalOutputTokens += d.outputTokens ?? 0;
totalCacheReadTokens += d.cacheReadTokens ?? 0;
totalCost += d.cost ?? 0;
totalDurationMs += d.duration ?? 0;
usageEventCount++;
// Update spinner so user knows it's still alive
progress(
"usage-tick",
`Analysis in progress (${usageEventCount} API call${usageEventCount === 1 ? "" : "s"})...`,
);
} else if (event.type === "session.idle") {
clearTimeout(timeoutId);
// Only reset inactivity timer on CONTENT-BEARING events.
// Usage events and other housekeeping must NOT reset the
// timer — otherwise the audit can hang indefinitely getting
// usage events with no actual content.
const isContentEvent =
event.type === "assistant.message" ||
event.type === "assistant.message_delta" ||
event.type === "assistant.reasoning_delta" ||
event.type === "assistant.turn_start" ||
event.type === "assistant.turn_end";

if (isContentEvent) {
clearTimeout(timeoutId);
timeoutId = setTimeout(() => {
unsubscribe();
signal?.removeEventListener("abort", onAbort);
// Emit aggregated usage summary
if (usageEventCount > 0) {
reject(
new Error(
`Audit timed out — no events for ${AUDIT_INACTIVITY_MS / 1000}s`,
),
);
}, AUDIT_INACTIVITY_MS);
}

if (event.type === "assistant.message") {
// Final complete message — extract content
const content = (event.data as { content?: string })?.content;
if (content) text = content;
} else if (event.type === "assistant.reasoning_delta") {
// Model is reasoning — forward to progress callback.
// The progress callback handles display logic (e.g.
// suppressing the preview after streaming starts).
const delta = (event.data as { deltaContent?: string })?.deltaContent;
if (delta) {
progress("reasoning", delta);
}
} else if (event.type === "assistant.message_delta") {
// Streaming delta — accumulate text for the final response.
// Skip whitespace-only deltas for progress display — the
// model can emit "\n\n" before reasoning starts (undocumented
// behaviour). We still accumulate the text but don't trigger
// the "Receiving audit report" UI for whitespace.
const delta = (event.data as { deltaContent?: string })?.deltaContent;
if (delta) {
text += delta;
if (text.trim().length > 0) {
progress(
"usage",
JSON.stringify({
inputTokens: totalInputTokens,
outputTokens: totalOutputTokens,
cacheReadTokens: totalCacheReadTokens,
cost: totalCost,
duration: totalDurationMs,
}),
"streaming",
`Receiving audit report (${text.length.toLocaleString()} chars)...`,
);
}
resolve(text);
} else if (event.type === "session.error") {
clearTimeout(timeoutId);
unsubscribe();
signal?.removeEventListener("abort", onAbort);
const msg =
(event.data as { message?: string })?.message ?? "unknown error";
reject(new Error(msg));
}
},
);
} else if (
event.type === "assistant.turn_start" ||
event.type === "assistant.turn_end"
) {
// Turn lifecycle events — the CLI server fires these during
// multi-call continuation loops. They're the ONLY reliable
// life sign during dead zones where no message/reasoning
// deltas arrive (known SDK bug — see github/copilot-sdk#524).
// Forward to progress for spinner feedback.
const turnId = (event.data as { turnId?: string })?.turnId;
progress(
"turn",
`${event.type === "assistant.turn_start" ? "Turn" : "Turn complete"} ${turnId ?? ""}`.trim(),
);
} else if (event.type === "assistant.usage") {
// Accumulate usage stats — don't print each one.
// We emit a single aggregate summary at completion.
const d = event.data as {
inputTokens?: number;
outputTokens?: number;
cacheReadTokens?: number;
cost?: number;
duration?: number;
};
totalInputTokens += d.inputTokens ?? 0;
totalOutputTokens += d.outputTokens ?? 0;
totalCacheReadTokens += d.cacheReadTokens ?? 0;
totalCost += d.cost ?? 0;
totalDurationMs += d.duration ?? 0;
usageEventCount++;
// Update spinner so user knows it's still alive
progress(
"usage-tick",
`Analysis in progress (${usageEventCount} API call${usageEventCount === 1 ? "" : "s"})...`,
);
} else if (event.type === "session.idle") {
clearTimeout(timeoutId);
unsubscribe();
signal?.removeEventListener("abort", onAbort);
// Emit aggregated usage summary
if (usageEventCount > 0) {
progress(
"usage",
JSON.stringify({
inputTokens: totalInputTokens,
outputTokens: totalOutputTokens,
cacheReadTokens: totalCacheReadTokens,
cost: totalCost,
duration: totalDurationMs,
}),
);
}
resolve(text);
} else if (event.type === "session.error") {
clearTimeout(timeoutId);
unsubscribe();
signal?.removeEventListener("abort", onAbort);
const msg =
(event.data as { message?: string })?.message ?? "unknown error";
reject(new Error(msg));
}
});

// Start inactivity timer
timeoutId = setTimeout(() => {
Expand Down
Loading