@spinabot/brigade 1.0.1 → 1.1.0

package/convex/instance.d.ts

@@ -0,0 +1,34 @@
+export declare const getCoord: import("convex/server").RegisteredQuery<"public", {
+    instanceId: string;
+}, Promise<{
+    _id: import("convex/values").GenericId<"gatewayCoord">;
+    _creationTime: number;
+    pid?: number | undefined;
+    pidAliveAt?: number | undefined;
+    heartbeatTs?: number | undefined;
+    heartbeatPid?: number | undefined;
+    heartbeatUptimeMs?: number | undefined;
+    lockPid?: number | undefined;
+    lockPort?: number | undefined;
+    lockCreatedAt?: string | undefined;
+    lockLeaseUntil?: number | undefined;
+    instanceId: string;
+    updatedAt: number;
+} | null>>;
+export declare const writePid: import("convex/server").RegisteredMutation<"public", {
+    pid: number;
+    instanceId: string;
+}, Promise<void>>;
+export declare const clearPid: import("convex/server").RegisteredMutation<"public", {
+    instanceId: string;
+}, Promise<void>>;
+export declare const writeHeartbeat: import("convex/server").RegisteredMutation<"public", {
+    ts: number;
+    pid: number;
+    uptimeMs: number;
+    instanceId: string;
+}, Promise<void>>;
+export declare const clearHeartbeat: import("convex/server").RegisteredMutation<"public", {
+    instanceId: string;
+}, Promise<void>>;
+//# sourceMappingURL=instance.d.ts.map

package/convex/instance.ts

@@ -0,0 +1,82 @@
+// convex/instance.ts — gatewayCoord (heartbeat/pid; the lock stays LOCAL — fs.open("wx") has no Convex equivalent)
+import { v } from "convex/values";
+import { mutation, query } from "./_generated/server.js";
+
+export const getCoord = query({
+	args: { instanceId: v.string() },
+	handler: async (ctx, args) => {
+		return ctx.db
+			.query("gatewayCoord")
+			.withIndex("by_instance", (q) => q.eq("instanceId", args.instanceId))
+			.first();
+	},
+});
+
+export const writePid = mutation({
+	args: { instanceId: v.string(), pid: v.number() },
+	handler: async (ctx, args) => {
+		const existing = await ctx.db
+			.query("gatewayCoord")
+			.withIndex("by_instance", (q) => q.eq("instanceId", args.instanceId))
+			.first();
+		const now = Date.now();
+		const payload = {
+			instanceId: args.instanceId,
+			pid: args.pid,
+			pidAliveAt: now,
+			updatedAt: now,
+		};
+		if (existing) await ctx.db.patch(existing._id, payload);
+		else await ctx.db.insert("gatewayCoord", payload);
+	},
+});
+
+export const clearPid = mutation({
+	args: { instanceId: v.string() },
+	handler: async (ctx, args) => {
+		const existing = await ctx.db
+			.query("gatewayCoord")
+			.withIndex("by_instance", (q) => q.eq("instanceId", args.instanceId))
+			.first();
+		if (existing) {
+			await ctx.db.patch(existing._id, { pid: undefined, pidAliveAt: undefined, updatedAt: Date.now() });
+		}
+	},
+});
+
+export const writeHeartbeat = mutation({
+	args: { instanceId: v.string(), ts: v.number(), pid: v.number(), uptimeMs: v.number() },
+	handler: async (ctx, args) => {
+		const existing = await ctx.db
+			.query("gatewayCoord")
+			.withIndex("by_instance", (q) => q.eq("instanceId", args.instanceId))
+			.first();
+		const payload = {
+			instanceId: args.instanceId,
+			heartbeatTs: args.ts,
+			heartbeatPid: args.pid,
+			heartbeatUptimeMs: args.uptimeMs,
+			updatedAt: Date.now(),
+		};
+		if (existing) await ctx.db.patch(existing._id, payload);
+		else await ctx.db.insert("gatewayCoord", payload);
+	},
+});
+
+export const clearHeartbeat = mutation({
+	args: { instanceId: v.string() },
+	handler: async (ctx, args) => {
+		const existing = await ctx.db
+			.query("gatewayCoord")
+			.withIndex("by_instance", (q) => q.eq("instanceId", args.instanceId))
+			.first();
+		if (existing) {
+			await ctx.db.patch(existing._id, {
+				heartbeatTs: undefined,
+				heartbeatPid: undefined,
+				heartbeatUptimeMs: undefined,
+				updatedAt: Date.now(),
+			});
+		}
+	},
+});

package/convex/logs.d.ts

@@ -0,0 +1,178 @@
+export declare const appendSessionEvent: import("convex/server").RegisteredMutation<"public", {
+    toolName?: string | undefined;
+    content?: ArrayBuffer | undefined;
+    args?: ArrayBuffer | undefined;
+    aborted?: boolean | undefined;
+    inner?: string | undefined;
+    delta?: string | undefined;
+    role?: string | undefined;
+    stopReason?: string | undefined;
+    errorMessage?: string | undefined;
+    toolCallId?: string | undefined;
+    isError?: boolean | undefined;
+    result?: ArrayBuffer | undefined;
+    attempt?: number | undefined;
+    maxAttempts?: number | undefined;
+    delayMs?: number | undefined;
+    success?: boolean | undefined;
+    finalError?: string | undefined;
+    willRetry?: boolean | undefined;
+    messageCount?: number | undefined;
+    type: string;
+    agentId: string;
+    sessionKey: string;
+    ts: string;
+    day: string;
+    ownerId: string;
+}, Promise<void>>;
+export declare const readSessionEventTail: import("convex/server").RegisteredQuery<"public", {
+    limit?: number | undefined;
+    day?: string | undefined;
+    ownerId: string;
+}, Promise<{
+    _id: import("convex/values").GenericId<"sessionEvents">;
+    _creationTime: number;
+    toolName?: string | undefined;
+    content?: ArrayBuffer | undefined;
+    args?: ArrayBuffer | undefined;
+    aborted?: boolean | undefined;
+    inner?: string | undefined;
+    delta?: string | undefined;
+    role?: string | undefined;
+    stopReason?: string | undefined;
+    errorMessage?: string | undefined;
+    toolCallId?: string | undefined;
+    isError?: boolean | undefined;
+    result?: ArrayBuffer | undefined;
+    attempt?: number | undefined;
+    maxAttempts?: number | undefined;
+    delayMs?: number | undefined;
+    success?: boolean | undefined;
+    finalError?: string | undefined;
+    willRetry?: boolean | undefined;
+    messageCount?: number | undefined;
+    type: string;
+    agentId: string;
+    sessionKey: string;
+    ts: string;
+    day: string;
+    ownerId: string;
+}[]>>;
+export declare const findLastError: import("convex/server").RegisteredQuery<"public", {
+    limit?: number | undefined;
+    ownerId: string;
+}, Promise<{
+    _id: import("convex/values").GenericId<"sessionEvents">;
+    _creationTime: number;
+    toolName?: string | undefined;
+    content?: ArrayBuffer | undefined;
+    args?: ArrayBuffer | undefined;
+    aborted?: boolean | undefined;
+    inner?: string | undefined;
+    delta?: string | undefined;
+    role?: string | undefined;
+    stopReason?: string | undefined;
+    errorMessage?: string | undefined;
+    toolCallId?: string | undefined;
+    isError?: boolean | undefined;
+    result?: ArrayBuffer | undefined;
+    attempt?: number | undefined;
+    maxAttempts?: number | undefined;
+    delayMs?: number | undefined;
+    success?: boolean | undefined;
+    finalError?: string | undefined;
+    willRetry?: boolean | undefined;
+    messageCount?: number | undefined;
+    type: string;
+    agentId: string;
+    sessionKey: string;
+    ts: string;
+    day: string;
+    ownerId: string;
+} | null>>;
+export declare const appendSubsystemRecord: import("convex/server").RegisteredMutation<"public", {
+    fields?: any;
+    message: string;
+    time: string;
+    level: "error" | "trace" | "debug" | "info" | "warn" | "fatal";
+    subsystem: string;
+    day: string;
+    ownerId: string;
+}, Promise<void>>;
+export declare const readSubsystemRecords: import("convex/server").RegisteredQuery<"public", {
+    level?: "error" | "trace" | "debug" | "info" | "warn" | "fatal" | undefined;
+    subsystem?: string | undefined;
+    limit?: number | undefined;
+    day?: string | undefined;
+    ownerId: string;
+}, Promise<{
+    _id: import("convex/values").GenericId<"subsystemLog">;
+    _creationTime: number;
+    fields?: any;
+    message: string;
+    time: string;
+    level: string;
+    subsystem: string;
+    day: string;
+    ownerId: string;
+}[]>>;
+export declare const pruneSubsystemLogs: import("convex/server").RegisteredMutation<"public", {
+    ownerId: string;
+    olderThanMs: number;
+}, Promise<{
+    removed: number;
+}>>;
+export declare const appendConfigAudit: import("convex/server").RegisteredMutation<"public", {
+    pid?: number | undefined;
+    sha256: string;
+    ts: string;
+    bytes: number;
+    instanceId: string;
+}, Promise<{
+    pid?: number | undefined;
+    prevHash?: string | undefined;
+    instanceId: string;
+    ts: string;
+    sha256: string;
+    bytes: number;
+    seq: number;
+    lineHash: string;
+}>>;
+export declare const listConfigAudit: import("convex/server").RegisteredQuery<"public", {
+    limit?: number | undefined;
+    instanceId: string;
+}, Promise<{
+    _id: import("convex/values").GenericId<"brigadeConfigAudit">;
+    _creationTime: number;
+    pid?: number | undefined;
+    prevHash?: string | undefined;
+    sha256: string;
+    ts: string;
+    bytes: number;
+    instanceId: string;
+    lineHash: string;
+    seq: number;
+}[]>>;
+export declare const writeConfigHealth: import("convex/server").RegisteredMutation<"public", {
+    sha256: string;
+    mtimeMs: number;
+    ts: string;
+    pid: number;
+    bytes: number;
+    ownerId: string;
+    configPath: string;
+}, Promise<void>>;
+export declare const readConfigHealth: import("convex/server").RegisteredQuery<"public", {
+    ownerId: string;
+}, Promise<{
+    _id: import("convex/values").GenericId<"configHealth">;
+    _creationTime: number;
+    sha256: string;
+    mtimeMs: number;
+    ts: string;
+    pid: number;
+    bytes: number;
+    ownerId: string;
+    configPath: string;
+} | null>>;
+//# sourceMappingURL=logs.d.ts.map

package/convex/logs.ts

@@ -0,0 +1,253 @@
+// convex/logs.ts — sessionEvents + subsystemLog + brigadeConfigAudit + configHealth
+import { v } from "convex/values";
+import { mutation, query } from "./_generated/server.js";
+
+const Level = v.union(
+	v.literal("trace"),
+	v.literal("debug"),
+	v.literal("info"),
+	v.literal("warn"),
+	v.literal("error"),
+	v.literal("fatal"),
+);
+
+// ============================================================================
+// sessionEvents (Pi session events)
+// ============================================================================
+
+export const appendSessionEvent = mutation({
+	args: {
+		ts: v.string(),
+		day: v.string(),
+		ownerId: v.string(),
+		agentId: v.string(),
+		sessionKey: v.string(),
+		type: v.string(),
+		inner: v.optional(v.string()),
+		delta: v.optional(v.string()),
+		toolCallId: v.optional(v.string()),
+		toolName: v.optional(v.string()),
+		args: v.optional(v.bytes()),
+		result: v.optional(v.bytes()),
+		isError: v.optional(v.boolean()),
+		role: v.optional(v.string()),
+		content: v.optional(v.bytes()),
+		stopReason: v.optional(v.string()),
+		errorMessage: v.optional(v.string()),
+		attempt: v.optional(v.number()),
+		maxAttempts: v.optional(v.number()),
+		delayMs: v.optional(v.number()),
+		aborted: v.optional(v.boolean()),
+		willRetry: v.optional(v.boolean()),
+		messageCount: v.optional(v.number()),
+		success: v.optional(v.boolean()),
+		finalError: v.optional(v.string()),
+	},
+	handler: async (ctx, args) => {
+		await ctx.db.insert("sessionEvents", args);
+	},
+});
+
+export const readSessionEventTail = query({
+	args: { ownerId: v.string(), day: v.optional(v.string()), limit: v.optional(v.number()) },
+	handler: async (ctx, args) => {
+		const day = args.day ?? new Date().toISOString().slice(0, 10);
+		const limit = args.limit && args.limit > 0 ? args.limit : 200;
+		// Take the most-recent `limit` (desc) but RETURN them chronological
+		// (oldest-first) to match the disk reader (event-logger.readSession-
+		// EventTail walks the file tail in append order). A consumer replaying
+		// the tail (e.g. convex→filesystem migrate) would otherwise reverse it.
+		const rows = await ctx.db
+			.query("sessionEvents")
+			.withIndex("by_owner_day", (q) => q.eq("ownerId", args.ownerId).eq("day", day))
+			.order("desc")
+			.take(limit);
+		return rows.reverse();
+	},
+});
+
+export const findLastError = query({
+	args: { ownerId: v.string(), limit: v.optional(v.number()) },
+	handler: async (ctx, args) => {
+		// Mirror the disk `getLastLoggedError` scan: the most recent error is not
+		// always a tool error (isError) — it can be an auto_retry_start with an
+		// errorMessage or an aborted compaction_end. Scan today's events
+		// newest-first and return the first error-shaped row. Falls back to the
+		// isError index for older days when today has none.
+		const day = new Date().toISOString().slice(0, 10);
+		const scan = args.limit && args.limit > 0 ? args.limit : 400;
+		const recent = await ctx.db
+			.query("sessionEvents")
+			.withIndex("by_owner_day", (q) => q.eq("ownerId", args.ownerId).eq("day", day))
+			.order("desc")
+			.take(scan);
+		for (const row of recent) {
+			if (row.type === "tool_execution_end" && row.isError === true) return row;
+			if (row.type === "auto_retry_start" && typeof row.errorMessage === "string") return row;
+			if (row.type === "compaction_end" && row.aborted === true && typeof row.errorMessage === "string") {
+				return row;
+			}
+		}
+		// Nothing today — fall back to the newest tool error across all days.
+		return ctx.db
+			.query("sessionEvents")
+			.withIndex("by_owner_error", (q) => q.eq("ownerId", args.ownerId).eq("isError", true))
+			.order("desc")
+			.first();
+	},
+});
+
+// ============================================================================
+// subsystemLog
+// ============================================================================
+
+export const appendSubsystemRecord = mutation({
+	args: {
+		time: v.string(),
+		day: v.string(),
+		ownerId: v.string(),
+		level: Level,
+		subsystem: v.string(),
+		message: v.string(),
+		fields: v.optional(v.any()),
+	},
+	handler: async (ctx, args) => {
+		await ctx.db.insert("subsystemLog", args);
+	},
+});
+
+export const readSubsystemRecords = query({
+	args: {
+		ownerId: v.string(),
+		day: v.optional(v.string()),
+		level: v.optional(Level),
+		subsystem: v.optional(v.string()),
+		limit: v.optional(v.number()),
+	},
+	handler: async (ctx, args) => {
+		const limit = args.limit && args.limit > 0 ? args.limit : 200;
+		const day = args.day ?? new Date().toISOString().slice(0, 10);
+		let rows = await ctx.db
+			.query("subsystemLog")
+			.withIndex("by_owner_day", (q) => q.eq("ownerId", args.ownerId).eq("day", day))
+			.order("desc")
+			.take(limit * 4);
+		if (args.level) rows = rows.filter((r) => r.level === args.level);
+		if (args.subsystem) rows = rows.filter((r) => r.subsystem === args.subsystem);
+		return rows.slice(0, limit);
+	},
+});
+
+export const pruneSubsystemLogs = mutation({
+	args: { ownerId: v.string(), olderThanMs: v.number() },
+	handler: async (ctx, args) => {
+		const cutoff = new Date(Date.now() - args.olderThanMs).toISOString();
+		// Prune ONE bounded page of expired rows; the client loops until
+		// {removed:0}. `.collect()` of the whole (unbounded, high-volume)
+		// subsystemLog blows the 16 MiB read + 8k-delete caps once it's large —
+		// which is exactly when prune runs. Expired rows sort FIRST on
+		// by_owner_day (oldest day first), so each page makes progress until
+		// only kept rows remain. Page size is small + count-bounded (log rows
+		// are tiny) to stay under the read cap. Lossless.
+		const rows = await ctx.db
+			.query("subsystemLog")
+			.withIndex("by_owner_day", (q) => q.eq("ownerId", args.ownerId))
+			.order("asc")
+			.take(500);
+		let removed = 0;
+		for (const r of rows) {
+			if (r.time < cutoff) {
+				await ctx.db.delete(r._id);
+				removed += 1;
+			}
+		}
+		return { removed };
+	},
+});
+
+// ============================================================================
+// brigadeConfigAudit (hash-chained)
+// ============================================================================
+
+export const appendConfigAudit = mutation({
+	args: {
+		instanceId: v.string(),
+		ts: v.string(),
+		sha256: v.string(),
+		bytes: v.number(),
+		pid: v.optional(v.number()),
+	},
+	handler: async (ctx, args) => {
+		const tail = await ctx.db
+			.query("brigadeConfigAudit")
+			.withIndex("by_instance_seq", (q) => q.eq("instanceId", args.instanceId))
+			.order("desc")
+			.first();
+		const seq = (tail?.seq ?? 0) + 1;
+		const prevHash = tail?.lineHash;
+		const lineHashInput = `${args.instanceId}|${args.ts}|${args.sha256}|${args.bytes}|${seq}|${prevHash ?? ""}`;
+		const enc = new TextEncoder().encode(lineHashInput);
+		const buf = await crypto.subtle.digest("SHA-256", enc);
+		const lineHash = Array.from(new Uint8Array(buf))
+			.map((b) => b.toString(16).padStart(2, "0"))
+			.join("");
+		const payload = {
+			instanceId: args.instanceId,
+			ts: args.ts,
+			sha256: args.sha256,
+			bytes: args.bytes,
+			seq,
+			lineHash,
+			...(prevHash !== undefined ? { prevHash } : {}),
+			...(args.pid !== undefined ? { pid: args.pid } : {}),
+		};
+		await ctx.db.insert("brigadeConfigAudit", payload);
+		return payload;
+	},
+});
+
+export const listConfigAudit = query({
+	args: { instanceId: v.string(), limit: v.optional(v.number()) },
+	handler: async (ctx, args) => {
+		const limit = args.limit && args.limit > 0 ? args.limit : 100;
+		return ctx.db
+			.query("brigadeConfigAudit")
+			.withIndex("by_instance_seq", (q) => q.eq("instanceId", args.instanceId))
+			.order("asc")
+			.take(limit);
+	},
+});
+
+// ============================================================================
+// configHealth
+// ============================================================================
+
+export const writeConfigHealth = mutation({
+	args: {
+		ownerId: v.string(),
+		ts: v.string(),
+		configPath: v.string(),
+		bytes: v.number(),
+		sha256: v.string(),
+		mtimeMs: v.number(),
+		pid: v.number(),
+	},
+	handler: async (ctx, args) => {
+		const existing = await ctx.db
+			.query("configHealth")
+			.withIndex("by_owner", (q) => q.eq("ownerId", args.ownerId))
+			.first();
+		if (existing) await ctx.db.replace(existing._id, args);
+		else await ctx.db.insert("configHealth", args);
+	},
+});
+
+export const readConfigHealth = query({
+	args: { ownerId: v.string() },
+	handler: async (ctx, args) => {
+		return ctx.db
+			.query("configHealth")
+			.withIndex("by_owner", (q) => q.eq("ownerId", args.ownerId))
+			.first();
+	},
+});