feat(project): add online-insight to project add

[jariy17]

Command structure

Usage: agentcore project add online-insight [options]

adds an online insight config to the current project

Options:
  --name <name>                            the name of the online insight config
  --agent <agent>                          harness/runtime name whose traffic to sample (mutually exclusive with --log-group-name)
  --endpoint <endpoint>                    the agent endpoint qualifier to scope monitoring to (requires --agent)
  --log-group-name <name...>               CloudWatch log group name(s) for custom data sources (1-5; mutually exclusive with --agent)
  --service-name <name...>                 service name(s) to filter traces for custom data sources (requires --log-group-name)
  --insight <insight...>                   insight ID(s) to apply: Builtin.Insight.* identifiers or full ARNs
  --clustering-frequency <freq...>         insight clustering cadence(s): DAILY, WEEKLY, MONTHLY
  --sampling-rate <sampling-rate>          percentage of sessions to sample (0.01-100)
  --description <description>              a description of the config's monitoring purpose
  --enable-on-create <true|false>          enable insights immediately after deploy (default true; pass false to add it paused)
  --tags <tags>                            tags to apply (JSON object of key/value strings)

online-insight is the insights counterpart of online-eval (#2048): same schema, same onlineEvalConfigs array, same spec key — it fills insights instead of evaluators and adds --clustering-frequency. Source options are identical (--agent XOR --log-group-name).

agentcore.json fields written

Each invocation appends one entry to onlineEvalConfigs[] in agentcore/agentcore.json:

Field	From flag	Notes
name	--name	required; 1-48 chars, letter-led alnum/underscore
agent	--agent	runtime to monitor; XOR logGroupNames
endpoint	--endpoint	endpoint qualifier; requires agent
logGroupNames	--log-group-name	custom CloudWatch source (1-5); XOR agent
serviceNames	--service-name	trace filter; requires logGroupNames
insights	--insight	required; each Builtin.Insight.* or an ARN
clusteringConfig.frequencies	--clustering-frequency	subset of DAILY/WEEKLY/MONTHLY; requires insights
samplingRate	--sampling-rate	required; 0.01-100 (percent)
description	--description	<=200 chars
enableOnCreate	--enable-on-create	true/false; false deploys the config paused
tags	--tags	JSON key/value object

Cross-field rules are enforced by OnlineEvalConfigSchema (agent XOR log groups, endpoint→agent, serviceNames→log groups, clustering→insights). Insight-ID format (Builtin.Insight.* / ARN) is validated in the handler. --agent existence is checked project-wide on the post-write ProjectSpecSchema validation in addResource.

Implementation

Mirrors the merged online-eval PR — 1 new handler + 1 co-located test + 3 small wirings:

• new src/handlers/project/add/online-insight/index.ts
• new src/handlers/project/add/online-insight/index.test.ts
• add/index.ts — register the handler
• project/types.ts — add online-insight to AddResourceInput
• core/project/manager.tsx — stack into the no-scaffold case + toProjectSpecKey

Verification

• tsc --noEmit 0 errors; bun test — new suite 15/15, project+manager regression 148/148; prettier --check . clean.
• Bug bash (22/22 pass): 9 happy + 11 error flows local, plus a real deploy to a sandbox account — synthesized AWS::BedrockAgentCore::OnlineEvaluationConfig (Insights wired, ENABLED at sampling 50), a second config deployed paused (--enable-on-create false → DISABLED), both verified via the control plane and torn down.

---

Related: aws/agentcore-l3-cdk-constructs#335 — OnlineEvaluationConfig agent resolves runtimes only (harness targets unsupported). The --agent help text here says "runtime" to match.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.12%. Comparing base (71b0f0a) to head (3a1547e).
⚠️ Report is 1 commits behind head on refactor.

Additional details and impacted files

@@             Coverage Diff              @@
##           refactor    #2063      +/-   ##
============================================
+ Coverage     97.10%   97.12%   +0.01%     
============================================
  Files           384      385       +1     
  Lines         22683    22790     +107     
============================================
+ Hits          22027    22134     +107     
  Misses          656      656              

☔ 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.

[nborges-aws]

The flag here says "harness/runtime", but looking at the projectSchema I'm pretty sure this will always fail for harness.

ProjectSpecSchema builds agentNames list from spec.runtimes. Harnesses are never read into this list. So when if (config.agent && !agentNames.has(config.agent)) runs, agentNames will never see harnesses and this will always fail.

[jariy17]

Good catch — you're right. Properly supporting a harness target needs changes in the CDK constructs (a harness would have to resolve to its underlying runtime in both the validation set and the log-group derivation), so I've filed aws/agentcore-l3-cdk-constructs#335 to track that.
For now I've narrowed the --agent description to runtime-only so it stops promising harness support that isn't wired up. I've also opened a follow-up PR (#2065) to make the same one-line correction on online-eval.

[nborges-aws]

One comment to address about harness source for --agent failing validation. PR looks good otherwise

[tejaskash]

LGTM