Skip to content

Update kagent and kmcp API reference docs #387

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

Open
github-actions wants to merge 1 commit into
base: main
Choose a base branch
from
+3 −1
Open

Update kagent and kmcp API reference docs #387

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
4 changes: 3 additions & 1 deletion src/app/docs/kagent/resources/api-ref/page.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -645,6 +645,8 @@ _Appears in:_
| --- | --- | --- | --- |
| `region` _string_ | AWS region where the Bedrock model is available (e.g., us-east-1, us-west-2) | | |
| `additionalModelRequestFields` _[JSON](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.31/#json-v1-apiextensions-k8s-io)_ | AdditionalModelRequestFields passes model-specific parameters to Bedrock's<br />additionalModelRequestFields in the Converse API. Use this for provider-specific<br />options that are not part of the standard InferenceConfiguration block, such as<br />Claude extended thinking or top_k. Values are forwarded as-is to the API.<br />Example: \{"top_k": 5, "thinking": \{"type": "enabled", "budget_tokens": 16000\}\} | | |
| `promptCaching` _boolean_ | PromptCaching enables Bedrock prompt caching by appending a CachePoint<br />block at the end of the Converse request's `system` content array and<br />the end of the `toolConfig.tools` array. Bedrock will cache the prefix up to and<br />including those cache points across requests in the same region for<br />roughly 5 minutes after first use, billing the cached portion at a<br />reduced rate on cache hits.<br /><br />Recommended for tool-using agents that make many Converse calls per<br />task with a stable system prompt and tool set — the per-call input<br />token count can drop by 70-90% on hit. Has no effect on models that<br />don't support caching; the marker is ignored by Bedrock for those.<br /><br />See https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html<br />for the current list of supported models and minimum prefix sizes. | false | |
| `cacheTTL` _string_ | CacheTTL controls how long Bedrock retains a cached prefix when<br />PromptCaching is enabled. Only meaningful when PromptCaching is true.<br /><br /> - "5m" (default): Bedrock's standard 5-minute sliding cache. Each cache<br /> hit refreshes the window. Supported by all prompt-caching models.<br /> - "1h": extended-TTL caching, useful for tasks whose Converse calls are<br /> spaced more than 5 minutes apart.<br /><br />NOTE: "1h" is NOT strictly better than "5m". Extended-TTL cache writes are<br />billed at a higher per-token rate than 5-minute writes, and 1h is supported<br />on a narrower set of models. Only choose "1h" when calls are spaced far<br />enough apart that a 5-minute cache would expire between them; otherwise the<br />higher write cost is wasted. See the AWS prompt-caching docs above. | 5m | Enum: [5m 1h] <br /> |


#### ByoDeploymentSpec
Expand Down Expand Up @@ -1291,7 +1293,7 @@ _Appears in:_
| `timeout` _[Duration](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.31/#duration-v1-meta)_ | | 30s | |
| `sseReadTimeout` _[Duration](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.31/#duration-v1-meta)_ | | | |
| `terminateOnClose` _boolean_ | | true | |
| `allowedNamespaces` _[AllowedNamespaces](#allowednamespaces)_ | AllowedNamespaces defines which namespaces are allowed to reference this RemoteMCPServer.<br />This follows the Gateway API pattern for cross-namespace route attachments.<br />If not specified, only Agents in the same namespace can reference this RemoteMCPServer.<br />See: https://gateway-api.sigs.k8s.io/guides/multiple-ns/#cross-namespace-route-attachment | | |
| `allowedNamespaces` _[AllowedNamespaces](#allowednamespaces)_ | AllowedNamespaces defines which namespaces are allowed to reference this RemoteMCPServer.<br />This follows the Gateway API pattern for cross-namespace route attachments.<br />If not specified, only Agents in the same namespace can reference this RemoteMCPServer.<br />See: https://gateway-api.sigs.k8s.io/guides/multiple-ns/#cross-namespace-route-attachment<br /><br />A cross-namespace-permitting value (from: All or from: Selector) is<br />mutually exclusive with spec.tls.caCertSecretRef (enforced by a spec-level<br />XValidation rule): a pinned CA Secret is mounted onto the consuming agent's<br />pod by bare name and Kubernetes resolves it in the agent's namespace, not<br />this RemoteMCPServer's, so a CA-pinning RemoteMCPServer cannot be referenced<br />cross-namespace. from: Same (the default) is always allowed. | | |
| `tls` _[TLSConfig](#tlsconfig)_ | TLS configuration for the upstream MCP server connection.<br />Use this for HTTPS upstreams that present a certificate the agent's<br />system trust store does not include (corporate CA, self-signed cert<br />on a test fixture, internal MCP gateway). Reuses the same TLSConfig<br />type as ModelConfig.spec.tls — disableVerify turns off certificate<br />validation entirely, caCertSecretRef + caCertSecretKey point at a<br />PEM bundle Secret in the same namespace, and disableSystemCAs<br />trusts only the named bundle.<br /><br />Note one asymmetry with ModelConfig: a spec-level XValidation rule<br />on RemoteMCPServer rejects spec.tls when spec.url has the http://<br />scheme (a TLS opinion contradicts a plaintext URL). ModelConfig has<br />no equivalent rule, so a TLS block can sit alongside any baseUrl. | | |


Expand Down