feat(config): implement global config to wire up config command by Hweinstock · Pull Request #1789 · aws/agentcore-cli

Hweinstock · 2026-07-17T12:59:42Z

Problem

The CLI is missing a way to view and control global configuration. There is an existing config command, but its not wired up.

Solution

implement a new top level globalConfig/ module that defines an accessor for global config. The global config is decoupled from the source of the data via an injected dependency.
move src/router/schemas.tsx to top level /parsing directory so that logic can be re-used.
global config is type-safe based on the global config schema using json dot notation. For example, valid keys are telemetry, telemetry.enabled, etc. All of which are derived automatically from the schema definition.
wire up config command to be output JSON responses.

Testing

wrote unit tests at the config command level to test e2e, rather than testing implementation directly.
e2e testing using existing config from CLI also works.
Ex.

> bun run start config
{
  "telemetry": {
    "enabled": false,
    "endpoint": "https://telemetry.agentcore.aws.dev"
  },
}
> bun run start config telemetry.endpoint 
"https://telemetry.agentcore.aws.dev"

> bun run start config telemetry.endpoint https://example.com
"https://example.com"

Future Work

add error types for telemetry, and easier reasoning (few todo comments on this).
add logic to setup installationId for telemetry (getOrPut).

codecov-commenter · 2026-07-17T13:00:52Z

Codecov Report

❌ Patch coverage is 95.14170% with 12 lines in your changes missing coverage. Please review.
✅ Project coverage is 94.85%. Comparing base (272e50e) to head (0531cce).

Files with missing lines	Patch %	Lines
src/handlers/config/handler.tsx	85.36%	12 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           refactor    #1789    +/-   ##
==========================================
  Coverage     94.84%   94.85%            
==========================================
  Files           143      149     +6     
  Lines          7120     7360   +240     
==========================================
+ Hits           6753     6981   +228     
- Misses          367      379    +12

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

Hweinstock · 2026-07-17T16:55:25Z

typecheck fixed in #1756 (comment)

Hweinstock · 2026-07-20T13:46:35Z

rebased onto create command work.

AlexanderRichey

I feel that the level of complexity this PR introduces is out of proportion with the functionality gained. Let's sync on this and see if we can develop a simpler approach.

AlexanderRichey · 2026-07-20T15:14:43Z

I have some prior art here that might be useful to look at: https://github.com/AlexanderRichey/yagss/blob/main/internal/builder/config.go

Hweinstock · 2026-07-21T00:29:00Z

converted to draft while I rework based on offline discussion from above, will take out of draft when ready for another review.

AlexanderRichey

This is looking really good. The main things blocking for me at the moment are (1) supporting (what I think are unnecessary) bool types, when JSON has a native bool type, and (2) adding a whole bunch of code to support isValidKey() when simpler options seem to be available.

AlexanderRichey · 2026-07-22T15:28:52Z

+    }
+  }
+
+  public async set(newConfig: Record<string, unknown>): Promise<GlobalConfig> {


Why not use the GlobalConfig type here? That would give us type-safety and any non-GlobalConfig value passed would fail at compile time.

answered below, but I'm aligned with typing it here.

AlexanderRichey · 2026-07-22T15:31:17Z

+export { type GlobalConfigAccessor, type GlobalConfig, type ReadWriteJson } from "./types";
+export { globalConfigSchema, isValidKey } from "./schemas";
+export { DefaultGlobalConfigAccessor } from "./accessor";
+export { FsReadWriteJson } from "./json";


This should probably be in its own json package. We can handle that in a follow up.

+1, I figure we pull it out once we have another consumer.

AlexanderRichey · 2026-07-22T17:45:07Z

@@ -0,0 +1,110 @@
+import z from "zod";
+
+/** Zod boolean schema that coerces string values (`"true"`, `"1"`, `"false"`, `"0"`) to booleans. */


Is this the current behavior? If not, I don't think we should support this. JSON has bools.

discussed in person, this was for coercion, but has a bad side effect of accepting odd looking configs.

AlexanderRichey · 2026-07-22T17:47:22Z

+}, z.boolean());
+
+/** Creates a Zod object schema that coerces JSON strings into objects before validation. */
+const coerceObject = <TShape extends z.ZodRawShape>(shape: TShape) =>


I'm not sure I'm following why we need this. We read a JSON string with JSON.parse, which should parse nested objects just fine, if that's what this is intended to handle.

AlexanderRichey · 2026-07-22T17:48:27Z

+    endpoint: z.string().optional(),
+    audit: coercedBoolean.optional(),
+  }).optional(),
+  installationId: z.uuid().optional(),


What's the argument in favor of using coerceObject and coercedBoolean over standard object and bool?

removed based on offline conversation

AlexanderRichey · 2026-07-22T17:49:36Z

+
+  return paths;
+}
+


I'm not following what all this code is for.

answered offline.

AlexanderRichey · 2026-07-22T17:50:27Z

+  /** Returns the current global config. */
+  get(): Promise<GlobalConfig>;
+  /** Validates and persists a new config. Throws on invalid shape. */
+  set(newConfig: Record<string, unknown>): Promise<GlobalConfig>;


Same as above comment. This should accept a type.

I was originally thinking its more "honest" since we already validate it internally to accept unknown. This also avoids the cast at the config handler call site. However, I think I agree that its easier to type this, and cast for exceptions so that we get the static check in as many places as possible.

AlexanderRichey · 2026-07-22T17:58:39Z

Shouldn't the options be enumerated here? Enumerating the options here would be far simpler than the current isValidKey() code.

I like that idea!

Hweinstock · 2026-07-22T22:48:13Z

Taking into draft while I address the comments and look to simplify.

Hweinstock · 2026-07-22T23:04:05Z

one other approach I considered is duplicating the schema into:

const CONFIG_KEY_TYPES = {
    "telemetry": "object",
    "telemetry.enabled": "boolean",
    "telemetry.endpoint": "string",
    "telemetry.audit": "boolean",
    "installationId": "string",
  };

Then using that instead of walking the zod schema to validate keys and paths. It would remove some complexity at the cost of maintaining two configs.

However, I feel like building this at runtime is easier and avoids the drift (but does introduce more complexity). Especially since if we go with the explicit duplicate mapping, we'd probably want a unit test verifying they don't drift which might result in writing the same code anyway.

… module

Hweinstock · 2026-07-23T22:13:02Z

Main changes:

now treats the file as a source of overrides, rather than source of truth.
maintains a default global config in memory and a resolve function to allow file values to override (env vars too at some point).
all coercion happens in the handler by walking the default config value.

AlexanderRichey · 2026-07-24T03:14:04Z

+/**
+ * Default values for the global config. Includes a unique installationId for each process.
+ */
+export const getDefaultGlobalConfig = once<GlobalConfig>(() => ({


Any reason why we can't have this?

const DEFAULT_GLOBAL_CONFIG = { telemetry: { enabled: true, audit: false, endpoint: "https://telemetry.agentcore.aws.dev", }, installationId: crypto.randomUUID(), } as GlobalConfig

I originally wasn't confident that loading the module in multiple places could cause different installationIds here. It looks like this isn't the case, and module resolution should handle this for us , but thought it was clearer to make this explicit in the code.

However, it sounds like maybe this isn't adding clarity and instead only adds complexity. I'll swap back to the constant.

…andles

github-actions Bot added agentcore-harness-reviewing AgentCore Harness review in progress and removed agentcore-harness-reviewing AgentCore Harness review in progress labels Jul 17, 2026

Hweinstock force-pushed the feat/global-config branch 8 times, most recently from db54dba to b39f26a Compare July 17, 2026 16:53

Hweinstock marked this pull request as ready for review July 17, 2026 16:55

Hweinstock changed the title ~~(wip)feat(config): implement global config to wire up config command~~ feat(config): implement global config to wire up config command Jul 17, 2026

Hweinstock commented Jul 17, 2026

View reviewed changes

Comment thread src/globalConfig/accessor.tsx Outdated

Comment thread src/parsing/index.tsx Outdated

Comment thread src/handlers/config/handler.tsx Outdated

Hweinstock force-pushed the feat/global-config branch from b39f26a to 4b819e1 Compare July 20, 2026 13:25

Hweinstock marked this pull request as draft July 20, 2026 13:25

Hweinstock force-pushed the feat/global-config branch 4 times, most recently from d60cf0d to 7fdb445 Compare July 20, 2026 13:43

Hweinstock marked this pull request as ready for review July 20, 2026 13:46

AlexanderRichey requested changes Jul 20, 2026

View reviewed changes

Hweinstock marked this pull request as draft July 21, 2026 00:28

Hweinstock force-pushed the feat/global-config branch 3 times, most recently from c9c3f17 to c0c3d7a Compare July 21, 2026 00:47

Hweinstock marked this pull request as ready for review July 21, 2026 22:06

Hweinstock force-pushed the feat/global-config branch 4 times, most recently from e393ae5 to 548979d Compare July 21, 2026 22:24

AlexanderRichey requested changes Jul 22, 2026

View reviewed changes

Hweinstock marked this pull request as draft July 22, 2026 22:48

Hweinstock force-pushed the feat/global-config branch from 7188c20 to cd99d77 Compare July 22, 2026 23:24

Hweinstock marked this pull request as ready for review July 22, 2026 23:31

Hweinstock marked this pull request as draft July 23, 2026 11:32

Hweinstock marked this pull request as ready for review July 23, 2026 12:07

Hweinstock marked this pull request as draft July 23, 2026 14:12

feat(globalConfig): implement config command through new globalConfig…

b298004

… module

Hweinstock force-pushed the feat/global-config branch from e190e1b to b298004 Compare July 23, 2026 22:10

Hweinstock marked this pull request as ready for review July 23, 2026 22:13

AlexanderRichey reviewed Jul 24, 2026

View reviewed changes

AlexanderRichey previously approved these changes Jul 24, 2026

View reviewed changes

Comment thread src/globalConfig/accessor.tsx Outdated

refactor(globalConfig): avoid redundant once that module resolution h…

2259b29

…andles

Hweinstock dismissed AlexanderRichey’s stale review via 0c63bb5 July 24, 2026 13:49

Hweinstock force-pushed the feat/global-config branch from 0c63bb5 to 1501976 Compare July 24, 2026 13:53

refactor(globalConfig): rename resolveConfig to applyOverrides

0531cce

Hweinstock force-pushed the feat/global-config branch from 1501976 to 0531cce Compare July 24, 2026 14:03

AlexanderRichey approved these changes Jul 24, 2026

View reviewed changes

nborges-aws reviewed Jul 24, 2026

View reviewed changes

Comment thread src/handlers/config/handler.tsx

nborges-aws approved these changes Jul 24, 2026

View reviewed changes

Hweinstock merged commit 1763609 into aws:refactor Jul 24, 2026
8 checks passed

Hweinstock deleted the feat/global-config branch July 24, 2026 19:07

		@@ -0,0 +1,110 @@
		import z from "zod";

		/** Zod boolean schema that coerces string values (`"true"`, `"1"`, `"false"`, `"0"`) to booleans. */

Uh oh!

Conversation

Hweinstock commented Jul 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Solution

Testing

Future Work

Uh oh!

codecov-commenter commented Jul 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Hweinstock commented Jul 17, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Hweinstock commented Jul 20, 2026

Uh oh!

AlexanderRichey left a comment

Choose a reason for hiding this comment

Uh oh!

AlexanderRichey commented Jul 20, 2026

Uh oh!

Hweinstock commented Jul 21, 2026

Uh oh!

AlexanderRichey left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Hweinstock Jul 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Hweinstock commented Jul 22, 2026

Uh oh!

Hweinstock commented Jul 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Hweinstock commented Jul 23, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Hweinstock Jul 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Hweinstock commented Jul 17, 2026 •

edited

Loading

codecov-commenter commented Jul 17, 2026 •

edited

Loading

Hweinstock Jul 22, 2026 •

edited

Loading

Hweinstock commented Jul 22, 2026 •

edited

Loading

Hweinstock Jul 24, 2026 •

edited

Loading