The best way to control Claude Code, Codex and more from any computer or phone
| 1 | import * as aiops from "@distilled.cloud/aws/aiops"; |
| 2 | import { unownedUnlessPrevious } from "alchemy/AdoptPolicy"; |
| 3 | import { isResolved } from "alchemy/Diff"; |
| 4 | import { createPhysicalName } from "alchemy/PhysicalName"; |
| 5 | import * as Provider from "alchemy/Provider"; |
| 6 | import { Resource } from "alchemy/Resource"; |
| 7 | import { createInternalTags, diffTags, hasAlchemyTags } from "alchemy/Tags"; |
| 8 | import { toWireDays } from "alchemy/Util/Duration"; |
| 9 | import type * as Duration from "effect/Duration"; |
| 10 | import * as Effect from "effect/Effect"; |
| 11 | import * as Schedule from "effect/Schedule"; |
| 12 | import * as Stream from "effect/Stream"; |
| 13 | import type { PolicyStatement } from "../IAM/Policy.ts"; |
| 14 | import type { Providers } from "../Providers.ts"; |
| 15 | |
| 16 | export interface InvestigationGroupEncryptionConfiguration { |
| 17 | /** |
| 18 | * How investigation data is encrypted — `"AWS_OWNED_KEY"` (the default, |
| 19 | * an AWS owned key) or `"CUSTOMER_MANAGED_KMS_KEY"`. |
| 20 | * @default "AWS_OWNED_KEY" |
| 21 | */ |
| 22 | type?: aiops.EncryptionConfigurationType; |
| 23 | /** |
| 24 | * ID or ARN of the customer managed KMS key to encrypt investigation data |
| 25 | * with. Required when `type` is `"CUSTOMER_MANAGED_KMS_KEY"`. |
| 26 | */ |
| 27 | kmsKeyId?: string; |
| 28 | } |
| 29 | |
| 30 | export interface InvestigationGroupCrossAccountConfiguration { |
| 31 | /** |
| 32 | * ARN of an IAM role in a source account that CloudWatch investigations |
| 33 | * assumes to retrieve telemetry during cross-account investigations. |
| 34 | */ |
| 35 | sourceRoleArn: string; |
| 36 | } |
| 37 | |
| 38 | export interface InvestigationGroupProps { |
| 39 | /** |
| 40 | * Name of the investigation group. If omitted, a deterministic physical |
| 41 | * name is generated from the app, stage, and logical ID. |
| 42 | * |
| 43 | * Changing the name replaces the investigation group. |
| 44 | * @default ${app}-${stage}-${id} |
| 45 | */ |
| 46 | name?: string; |
| 47 | /** |
| 48 | * ARN of the IAM role that CloudWatch investigations assumes to access |
| 49 | * telemetry (CloudWatch, X-Ray, CloudTrail, ...) during investigations. |
| 50 | * The role's trust policy must allow the `aiops.amazonaws.com` service |
| 51 | * principal to assume it. |
| 52 | */ |
| 53 | roleArn: string; |
| 54 | /** |
| 55 | * How long investigations and their data are retained (e.g. `"7 days"` |
| 56 | * or `Duration.days(7)`). Rounded to whole days on the wire |
| 57 | * (`retentionInDays`). |
| 58 | * The retention period cannot be updated in place — changing it replaces |
| 59 | * the investigation group. |
| 60 | * @default 90 days |
| 61 | */ |
| 62 | retention?: Duration.Input; |
| 63 | /** |
| 64 | * Encryption configuration for investigation data. Omit to use an AWS |
| 65 | * owned key. |
| 66 | * @default AWS owned key |
| 67 | */ |
| 68 | encryptionConfiguration?: InvestigationGroupEncryptionConfiguration; |
| 69 | /** |
| 70 | * Tag keys whose values CloudWatch investigations uses as boundaries to |
| 71 | * narrow the search for related telemetry (useful when resource names |
| 72 | * repeat across applications). |
| 73 | */ |
| 74 | tagKeyBoundaries?: string[]; |
| 75 | /** |
| 76 | * Map of Amazon Q Developer in chat applications configuration ARNs to |
| 77 | * SNS topic ARNs, used to send investigation updates to chat channels. |
| 78 | */ |
| 79 | chatbotNotificationChannel?: Record<string, string[]>; |
| 80 | /** |
| 81 | * Whether investigations can query CloudTrail event history for the past |
| 82 | * seven days of events. |
| 83 | * @default true |
| 84 | */ |
| 85 | isCloudTrailEventHistoryEnabled?: boolean; |
| 86 | /** |
| 87 | * Source-account role configurations for cross-account investigations. |
| 88 | */ |
| 89 | crossAccountConfigurations?: InvestigationGroupCrossAccountConfiguration[]; |
| 90 | /** |
| 91 | * IAM resource-policy statements attached to the investigation group, |
| 92 | * granting other principals or AWS services (for example |
| 93 | * `aiops.alarms.cloudwatch.amazonaws.com`, so CloudWatch alarms can start |
| 94 | * investigations) access to it. Serialized as a `2012-10-17` policy |
| 95 | * document via `PutInvestigationGroupPolicy`. |
| 96 | * |
| 97 | * Omit to leave any existing policy unmanaged; pass `[]` to delete a |
| 98 | * previously attached policy. |
| 99 | */ |
| 100 | policy?: PolicyStatement[]; |
| 101 | /** |
| 102 | * Tags to apply to the investigation group. Merged with internal Alchemy |
| 103 | * tags. |
| 104 | */ |
| 105 | tags?: Record<string, string>; |
| 106 | } |
| 107 | |
| 108 | export interface InvestigationGroup extends Resource< |
| 109 | "AWS.AIOps.InvestigationGroup", |
| 110 | InvestigationGroupProps, |
| 111 | { |
| 112 | /** Name of the investigation group. */ |
| 113 | name: string; |
| 114 | /** ARN of the investigation group. */ |
| 115 | arn: string; |
| 116 | /** ARN of the IAM role investigations assume to access telemetry. */ |
| 117 | roleArn: string | undefined; |
| 118 | /** |
| 119 | * How long investigations and their data are retained, in whole days |
| 120 | * (the AWS wire unit for the `retention` prop). |
| 121 | */ |
| 122 | retentionInDays: number | undefined; |
| 123 | }, |
| 124 | never, |
| 125 | Providers |
| 126 | > {} |
| 127 | |
| 128 | /** |
| 129 | * A CloudWatch investigations *investigation group* — the one-time, |
| 130 | * per-Region container that configures who can run AI-assisted operational |
| 131 | * investigations, which IAM role is used to access telemetry, how long |
| 132 | * investigation data is retained, and how it is encrypted. |
| 133 | * |
| 134 | * You can have at most one investigation group per Region in an account, so |
| 135 | * replacements are performed delete-first. |
| 136 | * ### Creating an Investigation Group |
| 137 | * **Example:** Basic Investigation Group |
| 138 | * ```typescript |
| 139 | * import * as AIOps from "@alchemy.run/aws/AIOps"; |
| 140 | * import * as IAM from "@alchemy.run/aws/IAM"; |
| 141 | * |
| 142 | * const role = yield* IAM.Role("InvestigationsRole", { |
| 143 | * assumeRolePolicyDocument: { |
| 144 | * Version: "2012-10-17", |
| 145 | * Statement: [{ |
| 146 | * Effect: "Allow", |
| 147 | * Principal: { Service: "aiops.amazonaws.com" }, |
| 148 | * Action: ["sts:AssumeRole"], |
| 149 | * }], |
| 150 | * }, |
| 151 | * managedPolicyArns: ["arn:aws:iam::aws:policy/AIOpsAssistantPolicy"], |
| 152 | * }); |
| 153 | * |
| 154 | * const group = yield* AIOps.InvestigationGroup("Investigations", { |
| 155 | * roleArn: role.roleArn, |
| 156 | * }); |
| 157 | * ``` |
| 158 | * |
| 159 | * **Example:** Short Retention and Tag Boundaries |
| 160 | * ```typescript |
| 161 | * const group = yield* AIOps.InvestigationGroup("Investigations", { |
| 162 | * roleArn: role.roleArn, |
| 163 | * retention: "7 days", |
| 164 | * tagKeyBoundaries: ["Application"], |
| 165 | * tags: { Environment: "test" }, |
| 166 | * }); |
| 167 | * ``` |
| 168 | * |
| 169 | * ### Resource Policy |
| 170 | * **Example:** Let CloudWatch Alarms Start Investigations |
| 171 | * ```typescript |
| 172 | * const group = yield* AIOps.InvestigationGroup("Investigations", { |
| 173 | * roleArn: role.roleArn, |
| 174 | * policy: [{ |
| 175 | * Effect: "Allow", |
| 176 | * Principal: { Service: "aiops.alarms.cloudwatch.amazonaws.com" }, |
| 177 | * Action: ["aiops:CreateInvestigation", "aiops:CreateInvestigationEvent"], |
| 178 | * Resource: "*", |
| 179 | * Condition: { |
| 180 | * StringEquals: { "aws:SourceAccount": "111122223333" }, |
| 181 | * ArnLike: { "aws:SourceArn": "arn:aws:cloudwatch:us-east-1:111122223333:alarm:*" }, |
| 182 | * }, |
| 183 | * }], |
| 184 | * }); |
| 185 | * ``` |
| 186 | * |
| 187 | * @resource |
| 188 | */ |
| 189 | export const InvestigationGroup = Resource<InvestigationGroup>("AWS.AIOps.InvestigationGroup"); |
| 190 | |
| 191 | /** |
| 192 | * Bounded retry for `createInvestigationGroup` while the freshly created IAM |
| 193 | * role propagates — AIOps validates that it can assume the role at create |
| 194 | * time, which surfaces as a ValidationException/AccessDeniedException that |
| 195 | * mentions the role for the first seconds of the role's life. |
| 196 | * |
| 197 | * Explicitly annotated so the conditional `Retry.Return` type never leaks |
| 198 | * into declaration emit (it would widen `AWS.providers()` for consumers). |
| 199 | */ |
| 200 | const retryWhileRolePropagates = <A, R>( |
| 201 | self: Effect.Effect<A, aiops.CreateInvestigationGroupError, R>, |
| 202 | ): Effect.Effect<A, aiops.CreateInvestigationGroupError, R> => |
| 203 | Effect.retry(self, { |
| 204 | while: (e) => |
| 205 | (e._tag === "ValidationException" || e._tag === "AccessDeniedException") && |
| 206 | /role/i.test(e.message ?? ""), |
| 207 | schedule: Schedule.max([Schedule.fixed("3 seconds"), Schedule.recurs(8)]), |
| 208 | }); |
| 209 | |
| 210 | const sameStringArray = (l: readonly string[], r: readonly string[]) => |
| 211 | l.length === r.length && l.every((v, i) => v === r[i]); |
| 212 | |
| 213 | export const InvestigationGroupProvider = () => |
| 214 | Provider.effect( |
| 215 | InvestigationGroup, |
| 216 | Effect.gen(function* () { |
| 217 | const createName = Effect.fn(function* ( |
| 218 | id: string, |
| 219 | props: Pick<InvestigationGroupProps, "name">, |
| 220 | ) { |
| 221 | return props.name ?? (yield* createPhysicalName({ id, maxLength: 64 })); |
| 222 | }); |
| 223 | |
| 224 | const observeByArn = (arn: string) => |
| 225 | aiops |
| 226 | .getInvestigationGroup({ identifier: arn }) |
| 227 | .pipe(Effect.catchTag("ResourceNotFoundException", () => Effect.undefined)); |
| 228 | |
| 229 | // At most one investigation group exists per Region, so scanning the |
| 230 | // list for our name is a single cheap call. |
| 231 | const findArnByName = (name: string) => |
| 232 | aiops.listInvestigationGroups.items({}).pipe( |
| 233 | Stream.runCollect, |
| 234 | Effect.map((items) => Array.from(items).find((g) => g.name === name)?.arn), |
| 235 | ); |
| 236 | |
| 237 | const observe = Effect.fn(function* (name: string, arnHint: string | undefined) { |
| 238 | if (arnHint !== undefined) { |
| 239 | const found = yield* observeByArn(arnHint); |
| 240 | if (found !== undefined) return found; |
| 241 | } |
| 242 | const arn = yield* findArnByName(name); |
| 243 | return arn === undefined ? undefined : yield* observeByArn(arn); |
| 244 | }); |
| 245 | |
| 246 | const observedTags = (arn: string) => |
| 247 | aiops.listTagsForResource({ resourceArn: arn }).pipe( |
| 248 | Effect.map((r) => { |
| 249 | const tags: Record<string, string> = {}; |
| 250 | for (const [key, value] of Object.entries(r.tags ?? {})) { |
| 251 | if (value !== undefined) tags[key] = value; |
| 252 | } |
| 253 | return tags; |
| 254 | }), |
| 255 | Effect.catchTag("ResourceNotFoundException", () => |
| 256 | Effect.succeed({} as Record<string, string>), |
| 257 | ), |
| 258 | ); |
| 259 | |
| 260 | const toAttributes = (live: aiops.GetInvestigationGroupResponse) => ({ |
| 261 | name: live.name!, |
| 262 | arn: live.arn!, |
| 263 | roleArn: live.roleArn, |
| 264 | retentionInDays: live.retentionInDays, |
| 265 | }); |
| 266 | |
| 267 | return InvestigationGroup.Provider.of({ |
| 268 | stables: ["name", "arn"], |
| 269 | list: () => |
| 270 | Effect.gen(function* () { |
| 271 | const items = yield* aiops.listInvestigationGroups.items({}).pipe(Stream.runCollect); |
| 272 | const groups: { |
| 273 | name: string; |
| 274 | arn: string; |
| 275 | roleArn: string | undefined; |
| 276 | retentionInDays: number | undefined; |
| 277 | }[] = []; |
| 278 | for (const item of Array.from(items)) { |
| 279 | if (item.arn === undefined) continue; |
| 280 | const live = yield* observeByArn(item.arn); |
| 281 | if (live?.arn !== undefined && live.name !== undefined) { |
| 282 | groups.push(toAttributes(live)); |
| 283 | } |
| 284 | } |
| 285 | return groups; |
| 286 | }), |
| 287 | read: Effect.fn(function* ({ id, olds, output, previous }) { |
| 288 | const name = output?.name ?? (yield* createName(id, olds ?? {})); |
| 289 | const live = yield* observe(name, output?.arn); |
| 290 | if (live?.arn === undefined || live.name === undefined) { |
| 291 | return undefined; |
| 292 | } |
| 293 | const attrs = toAttributes(live); |
| 294 | const tags = yield* observedTags(live.arn); |
| 295 | return (yield* hasAlchemyTags(id, tags)) |
| 296 | ? attrs |
| 297 | : unownedUnlessPrevious(attrs, previous, "name"); |
| 298 | }), |
| 299 | diff: Effect.fn(function* ({ id, news, olds }) { |
| 300 | if (!isResolved(news)) return undefined; |
| 301 | const oldName = yield* createName(id, olds ?? {}); |
| 302 | const newName = yield* createName(id, news ?? {}); |
| 303 | // Only one investigation group may exist per Region, so a |
| 304 | // replacement must delete the old group before creating the new. |
| 305 | if (oldName !== newName) { |
| 306 | return { action: "replace", deleteFirst: true } as const; |
| 307 | } |
| 308 | // The retention period has no update API — replace to change it. |
| 309 | if (toWireDays(olds?.retention) !== toWireDays(news?.retention)) { |
| 310 | return { action: "replace", deleteFirst: true } as const; |
| 311 | } |
| 312 | // fall through: engine default update logic for mutable fields |
| 313 | }), |
| 314 | reconcile: Effect.fn(function* ({ id, news, output, session }) { |
| 315 | const name = output?.name ?? (yield* createName(id, news)); |
| 316 | const internalTags = yield* createInternalTags(id); |
| 317 | const desiredTags = { ...news.tags, ...internalTags }; |
| 318 | |
| 319 | // 1. OBSERVE — cloud state is authoritative; output.arn is only a |
| 320 | // cache of the identifier. |
| 321 | let live = yield* observe(name, output?.arn); |
| 322 | |
| 323 | // 2. ENSURE — create when missing. A ConflictException here means |
| 324 | // a concurrent create raced us (one group per Region): if a |
| 325 | // group with OUR name now exists we converge on it, otherwise |
| 326 | // the conflict is a real error (a foreign group occupies the |
| 327 | // Region's single slot) and is propagated. |
| 328 | if (live === undefined) { |
| 329 | live = yield* aiops |
| 330 | .createInvestigationGroup({ |
| 331 | name, |
| 332 | roleArn: news.roleArn, |
| 333 | retentionInDays: toWireDays(news.retention), |
| 334 | encryptionConfiguration: news.encryptionConfiguration, |
| 335 | tagKeyBoundaries: news.tagKeyBoundaries, |
| 336 | chatbotNotificationChannel: news.chatbotNotificationChannel, |
| 337 | isCloudTrailEventHistoryEnabled: news.isCloudTrailEventHistoryEnabled, |
| 338 | crossAccountConfigurations: news.crossAccountConfigurations, |
| 339 | tags: desiredTags, |
| 340 | }) |
| 341 | .pipe( |
| 342 | retryWhileRolePropagates, |
| 343 | Effect.flatMap((created) => |
| 344 | created.arn === undefined ? observe(name, undefined) : observeByArn(created.arn), |
| 345 | ), |
| 346 | Effect.catchTag("ConflictException", (error) => |
| 347 | observe(name, undefined).pipe( |
| 348 | Effect.flatMap((existing) => |
| 349 | existing === undefined ? Effect.fail(error) : Effect.succeed(existing), |
| 350 | ), |
| 351 | ), |
| 352 | ), |
| 353 | ); |
| 354 | } |
| 355 | |
| 356 | const arn = live?.arn ?? output?.arn; |
| 357 | |
| 358 | // 3. SYNC — diff each OBSERVED mutable aspect against the desired |
| 359 | // state and PATCH only on drift. Aspects the user leaves |
| 360 | // undefined are unmanaged (left as observed). |
| 361 | if (live !== undefined && arn !== undefined) { |
| 362 | const update: Omit<aiops.UpdateInvestigationGroupRequest, "identifier"> = {}; |
| 363 | if (live.roleArn !== news.roleArn) { |
| 364 | update.roleArn = news.roleArn; |
| 365 | } |
| 366 | if ( |
| 367 | news.encryptionConfiguration !== undefined && |
| 368 | ((live.encryptionConfiguration?.type ?? "AWS_OWNED_KEY") !== |
| 369 | (news.encryptionConfiguration.type ?? "AWS_OWNED_KEY") || |
| 370 | live.encryptionConfiguration?.kmsKeyId !== news.encryptionConfiguration.kmsKeyId) |
| 371 | ) { |
| 372 | update.encryptionConfiguration = news.encryptionConfiguration; |
| 373 | } |
| 374 | if ( |
| 375 | news.tagKeyBoundaries !== undefined && |
| 376 | !sameStringArray(live.tagKeyBoundaries ?? [], news.tagKeyBoundaries) |
| 377 | ) { |
| 378 | update.tagKeyBoundaries = news.tagKeyBoundaries; |
| 379 | } |
| 380 | if ( |
| 381 | news.isCloudTrailEventHistoryEnabled !== undefined && |
| 382 | live.isCloudTrailEventHistoryEnabled !== news.isCloudTrailEventHistoryEnabled |
| 383 | ) { |
| 384 | update.isCloudTrailEventHistoryEnabled = news.isCloudTrailEventHistoryEnabled; |
| 385 | } |
| 386 | if ( |
| 387 | news.chatbotNotificationChannel !== undefined && |
| 388 | JSON.stringify(live.chatbotNotificationChannel ?? {}) !== |
| 389 | JSON.stringify(news.chatbotNotificationChannel) |
| 390 | ) { |
| 391 | update.chatbotNotificationChannel = news.chatbotNotificationChannel; |
| 392 | } |
| 393 | if ( |
| 394 | news.crossAccountConfigurations !== undefined && |
| 395 | JSON.stringify(live.crossAccountConfigurations ?? []) !== |
| 396 | JSON.stringify(news.crossAccountConfigurations) |
| 397 | ) { |
| 398 | update.crossAccountConfigurations = news.crossAccountConfigurations; |
| 399 | } |
| 400 | if (Object.keys(update).length > 0) { |
| 401 | yield* aiops.updateInvestigationGroup({ |
| 402 | identifier: arn, |
| 403 | ...update, |
| 404 | }); |
| 405 | live = (yield* observeByArn(arn)) ?? live; |
| 406 | } |
| 407 | } |
| 408 | |
| 409 | // 3b. SYNC POLICY — diff the OBSERVED resource policy against the |
| 410 | // desired policy document and put/delete only on drift. |
| 411 | // `policy: undefined` leaves any existing policy unmanaged; |
| 412 | // `policy: []` deletes it. |
| 413 | if (arn !== undefined && news.policy !== undefined) { |
| 414 | const observedPolicy = yield* aiops |
| 415 | .getInvestigationGroupPolicy({ identifier: arn }) |
| 416 | .pipe( |
| 417 | Effect.map((r) => r.policy), |
| 418 | Effect.catchTag("ResourceNotFoundException", () => Effect.undefined), |
| 419 | ); |
| 420 | if (news.policy.length === 0) { |
| 421 | if (observedPolicy !== undefined) { |
| 422 | yield* aiops |
| 423 | .deleteInvestigationGroupPolicy({ identifier: arn }) |
| 424 | .pipe(Effect.catchTag("ResourceNotFoundException", () => Effect.void)); |
| 425 | } |
| 426 | } else { |
| 427 | const desiredPolicy = JSON.stringify({ |
| 428 | Version: "2012-10-17", |
| 429 | Statement: news.policy, |
| 430 | }); |
| 431 | const inSync = yield* Effect.sync(() => { |
| 432 | if (observedPolicy === undefined) return false; |
| 433 | try { |
| 434 | return JSON.stringify(JSON.parse(observedPolicy)) === desiredPolicy; |
| 435 | } catch { |
| 436 | return false; |
| 437 | } |
| 438 | }); |
| 439 | if (!inSync) { |
| 440 | yield* aiops.putInvestigationGroupPolicy({ |
| 441 | identifier: arn, |
| 442 | policy: desiredPolicy, |
| 443 | }); |
| 444 | } |
| 445 | } |
| 446 | } |
| 447 | |
| 448 | // 3c. SYNC TAGS — diff against OBSERVED cloud tags so adoption |
| 449 | // converges (create-time tags only apply on first create). |
| 450 | if (arn !== undefined) { |
| 451 | const currentTags = yield* observedTags(arn); |
| 452 | const { upsert, removed } = diffTags(currentTags, desiredTags); |
| 453 | if (upsert.length > 0) { |
| 454 | yield* aiops.tagResource({ |
| 455 | resourceArn: arn, |
| 456 | tags: Object.fromEntries(upsert.map((t) => [t.Key, t.Value])), |
| 457 | }); |
| 458 | } |
| 459 | if (removed.length > 0) { |
| 460 | yield* aiops.untagResource({ |
| 461 | resourceArn: arn, |
| 462 | tagKeys: removed, |
| 463 | }); |
| 464 | } |
| 465 | } |
| 466 | |
| 467 | yield* session.note(name); |
| 468 | return { |
| 469 | name: live?.name ?? name, |
| 470 | arn: arn!, |
| 471 | roleArn: live?.roleArn ?? news.roleArn, |
| 472 | retentionInDays: live?.retentionInDays ?? toWireDays(news.retention), |
| 473 | }; |
| 474 | }), |
| 475 | delete: Effect.fn(function* ({ output }) { |
| 476 | yield* aiops |
| 477 | .deleteInvestigationGroup({ identifier: output.arn }) |
| 478 | .pipe(Effect.catchTag("ResourceNotFoundException", () => Effect.void)); |
| 479 | }), |
| 480 | }); |
| 481 | }), |
| 482 | ); |