Component
core/ai-tool/agent-invocation
A tool letting one agent delegate a step to another over an agent-to-agent protocol, A2A by default: the caller waits synchronously for a result, asynchronously for a handle it polls, or not at all, fire-and-forget. Its required agent socket resolves the target by topology, not prompt. Marked mutating: a delegated agent may act in the world. Realized on Kubernetes as a Knative Service scaled to zero between calls yet admitting each, it publishes a host, port, endpoint and protocol for a consumer to wire to. The catalog pins no image, so nothing renders until the Release sets `infra.image`. The agent's address reaches the tool's code under default variable names; an `inject:` block on the connection delivers instead, as variables or files, the names that code reads. The component's own `inject:` block adds `env` and `file` entries beside them, and one reading a tool value alone replaces that value's default name. The connection may hold a credential, presented under the agent's published credential_header, its Secret holding the whole value: an API key in a header or cookie, or an HTTP scheme such as Bearer or Basic; a client_secret, for the OAuth 2.0 client-credentials grant, directly or through OpenID Connect; and, for mutual TLS, a client_certificate, mounted as files. A server_ca, also mounted, names the agent's server certificate authority. The build halts on a connection to an agent labelled managed with none of credential, client_secret, client_certificate or unauthenticated: true, and on a credential or client_secret to one labelled external publishing protocol http; to an agent inside the Release, http carries the credential or token unencrypted on the cluster network.
Asal-usul
- Digest konten
- sha256:71f2e29f01cce29f360bc23b17e3d1e64bab3af72bdcb266a219be8e42b538d0
- Dipublikasikan
- 2026-11-10T10:09:46Z
- Identitas penandatangan
- https://token.actions.githubusercontent.com/https://github.com/aggrega-ai/aggrega-hub/.github/workflows/catalog-publish-sign.yml@refs/heads/main
Tanda tangan
- Resmiapplication/vnd.dev.sigstore.bundle.v0.3+json
https://token.actions.githubusercontent.com/https://github.com/aggrega-ai/aggrega-hub/.github/workflows/catalog-publish-sign.yml@refs/heads/main
sha256:26a851d77c1bcfa5b0a18dea6d5db46a1bc633c586bf5d0857a121f462c92071
Referensi pada versi ini
oci://hub.aggrega.ai/component/core/ai-tool/agent-invocation:1.0.0Parameter pengembang
Nilai yang ditetapkan pengembang di Aggregate.
Memperluas BaseComponents#/schemas/BaseToolParameter
| Nama | Tipe | Bawaan | Wajib | Deskripsi |
|---|---|---|---|---|
| protocol | string | a2a | ya | — |
| execution_mode | string (synchronous | asynchronous | fire_and_forget) | synchronous | ya | — |
Penentuan ukuran oleh operator
Ukuran infrastruktur yang disesuaikan operator untuk setiap lingkungan.
| Nama | Tipe | Bawaan | Wajib | Deskripsi |
|---|---|---|---|---|
| image | string | — | tidak | Container image serving the delegation over HTTP. |
| port | integer (min 1024, max 65535) | 8080 | tidak | Port the tool's own container listens on. A consumer does not necessarily connect here: the realization publishes the address callers use, and a control plane that fronts the tool routes from its own port to this one. |
| compute | object | — | tidak | — |
| min_replicas | integer (min 0) | 0 | tidak | Replica floor. Zero is what lets the tool run no replicas between calls, which the realization reaches through its control plane. |
| max_replicas | integer (min 1) | 5 | tidak | Replica ceiling the realization's autoscaler may reach. |
| control_plane_present | boolean | false | tidak | Operator affirmation that Knative Serving, the control plane the realization needs, is installed in the target cluster. The build halts unless this is set true, so a custom resource is never emitted as an orphan into a cluster that cannot reconcile it. |
| overrides | object (open) | — | tidak | Realization-specific overrides. Accepts any nested structure. Each Recipe documents which keys it reads. |
Kontrak keluaran
Nilai yang dipublikasikan komponen ini kepada konsumen melalui topologi uses.
| Nama | Tipe | Bawaan | Wajib | Deskripsi |
|---|---|---|---|---|
| host | string | — | ya | In-cluster DNS name an agent calls to delegate a step to the wired agent. A consumer that composes its own address dials this name; one that does not reads `endpoint` instead. |
| port | integer | — | ya | TCP port the delegation is accepted on. The realization fronts the tool on the control plane's own port rather than the tool container's, so a consumer reads this rather than the port an operator set. |
| endpoint | string | — | ya | Full base URL an agent posts a delegation to, with the host and the port already composed. A consumer hands it to its tool client as the address it calls. |
| protocol | string | — | tidak | URL scheme to reach `host` and `port` over, for a consumer composing its own address rather than reading `endpoint`. |
Soket uses
Soket dependensi yang menjadi jalur koneksi komponen ini. Target setiap koneksi yang dihubungkan melalui sebuah soket harus memenuhi selektor labelnya, dan setiap koneksi dapat membawa lampiran pada slot yang dideklarasikan soket tersebut. Slot wajib harus diisi oleh setiap koneksi yang dihubungkan oleh Aggregate atau Release melalui soket tersebut.
target_agent
- Selektor
- component.aggrega.ai/category=ai-agent
- Injeksi
- env, file
Slot yang dapat diisi oleh setiap koneksi melalui soket ini
| Slot | Selektor | Wajib | Injeksi | Sensitif | Deskripsi |
|---|---|---|---|---|---|
| credential | feature.aggrega.ai/kind=secret | tidak | env | ya | Optional credential the tool presents to the agent this connection reaches, under the request header that agent publishes as credential_header. When attached, the Recipe delivers it to the tool's container as AGENT_TOOL_TARGET_CREDENTIAL, a Kubernetes secretKeyRef to the Secret the credential publishes, and, on a connection writing no inject block, the header as AGENT_TOOL_TARGET_CREDENTIAL_HEADER where the agent publishes one; a credential publishing no Secret name and data key halts the build, and so does one on a connection to an agent labelled component.aggrega.ai/hosting external that publishes protocol http, to which it would travel unencrypted. An agent inside the Release that publishes protocol http receives it unencrypted on the cluster network. The Secret's value is the whole header value, delivered unchanged for the container to present as it is, with no prefix added: the key, for an API key the agent's card places in a header; the cookie's name, "=" and the key, under a credential_header of Cookie, for one it places in a cookie; and the scheme's name, a space and the credential, as "Bearer " and a token, for an HTTP scheme whose value is fixed, such as Bearer or Basic. An API key the card places in the query string, an HTTP scheme whose value answers a challenge from the agent, such as Digest, and two header credentials required together are not covered; OAuth 2.0 and OpenID Connect take client_secret, and mutual TLS client_certificate. Fill it with core/referenced-secret mapped to helm/k8s-existing-secret, naming a Secret the namespace already holds: a realization that names its Secret after the tool, such as k8s-external-secret or k8s-secret-from-parameter, halts the build here, because the Recipe renders its Service from an artifacts template and gives the tool no rendered name. Unattached, the tool runs without an injected credential. |
| client_secret | feature.aggrega.ai/kind=secret | tidak | env | ya | Optional client secret the tool's container sends, with client_id, to the agent's authorization server to obtain an access token by the OAuth 2.0 client-credentials grant, at the token_url the agent publishes or the token endpoint its oidc_discovery_url names. Delivered as AGENT_TOOL_TARGET_CLIENT_SECRET, a Kubernetes secretKeyRef to the Secret it publishes, beside AGENT_TOOL_TARGET_CLIENT_ID and AGENT_TOOL_TARGET_AUDIENCE and, on a connection writing no inject block, AGENT_TOOL_TARGET_TOKEN_URL, AGENT_TOOL_TARGET_OIDC_DISCOVERY_URL and AGENT_TOOL_TARGET_SCOPES, each where it has a value. Obtaining, caching and refreshing the token, and presenting it on each request, are the container's. The other OAuth 2.0 flows a card may declare, authorization code, device code, implicit and password, are not covered. The build halts on a client_secret with no client_id, on one wired to an agent publishing neither token_url nor oidc_discovery_url, on either address when it is not an https URL naming a host or carries a user name or password, and on one wired to an agent labelled component.aggrega.ai/hosting external that publishes protocol http, to which the token obtained with it would travel unencrypted. An agent inside the Release that publishes protocol http receives the token unencrypted on the cluster network. Fill it as credential is filled; it may sit beside credential on one connection. |
| client_certificate | feature.aggrega.ai/kind=secret | tidak | file | ya | Optional client certificate the tool's container presents in the TLS handshake to an agent whose card requires mutual TLS. Its Secret holds tls.crt, the certificate, and tls.key, its private key, as a kubernetes.io/tls Secret does. The Recipe mounts the whole Secret read-only into the tool's container and delivers AGENT_TOOL_TARGET_CLIENT_CERT_FILE and AGENT_TOOL_TARGET_CLIENT_KEY_FILE naming the two files; presenting the certificate, and reading the files again when a renewal updates them, are the container's. The build halts on a client_certificate publishing no Secret name, on one wired to an agent publishing protocol http, and on one wired to an agent publishing no protocol and an endpoint that is not an https URL. Fill it with core/referenced-secret mapped to helm/k8s-existing-secret, naming a Secret the namespace already holds; the data key it publishes is not read, since the whole Secret is mounted. It may sit beside credential or client_secret on one connection. |
| server_ca | feature.aggrega.ai/kind=secret | tidak | file | ya | Optional authority the tool's container checks the agent's server certificate against, where the image does not already trust the authority that signs it. Its Secret holds ca.crt: the agent's server authority, not the issuer of a client certificate, so a ca.crt written beside a client certificate is not this one. The Recipe mounts the whole Secret read-only into the tool's container and delivers AGENT_TOOL_TARGET_SERVER_CA_FILE naming the file; checking the agent's certificate against it is the container's. It serves a connection holding any credential, or none. The build halts on a server_ca publishing no Secret name, on one wired to an agent publishing protocol http, and on one wired to an agent publishing no protocol and an endpoint that is not an https URL. Fill it as client_certificate is filled. |
Slot lampiran
Titik pengikatan bernama tempat sebuah Feature dilampirkan ke komponen ini. Selektor sebuah slot membatasi Feature mana yang dapat mengisinya, dan daftar injeksinya menyebutkan cara materi Feature dapat dikirimkan. Slot wajib harus terikat pada setiap instans, melalui lampiran pada instans tersebut atau pada cakupan di atasnya, dan slot sensitif dideklarasikan untuk menangani materi rahasia. Sebuah Feature juga dapat dilampirkan dengan nama yang tidak dideklarasikan oleh slot mana pun.
Tidak ada slot lampiran yang dideklarasikan.
Daftar izin injeksi mandiri
env, file
Relasi
Sisi bertipe ke dan dari entitas lain yang dipublikasikan. Referensi tanpa target yang dipublikasikan ditampilkan sebagai sisi yang terputus, bukan sebagai tautan.
Menggunakan
- menggunakanmenggunakan soket «target_agent»
component.aggrega.ai/category=ai-agent
Slot lampiran: credential, client_secret, client_certificate, server_ca
core/ai-agent/a2a-remote v1.0.0 →core/ai-agent/agent-framework v1.0.0 →core/ai-agent/agent-framework v1.1.0 →core/ai-agent/agenta-api v1.0.0 →core/ai-agent/agenta-cron v1.0.0 →core/ai-agent/agenta-migrations v1.0.0 →core/ai-agent/agenta-runner v1.0.0 →core/ai-agent/agenta-services v1.0.0 →core/ai-agent/agenta-web v1.0.0 →core/ai-agent/agenta-web-mobile v1.0.0 →core/ai-agent/agenta-worker-queues v1.0.0 →core/ai-agent/agenta-worker-streams v1.0.0 →core/ai-agent/anthropic v1.0.0 →core/ai-agent/autogen v1.0.0 →core/ai-agent/autogpt v1.0.0 →core/ai-agent/bedrock v1.0.0 →core/ai-agent/bland v1.0.0 →core/ai-agent/claude-agent-sdk v1.0.0 →core/ai-agent/crewai v1.0.0 →core/ai-agent/crewai v1.1.0 →core/ai-agent/dify v1.0.0 →core/ai-agent/dify v1.1.0 →core/ai-agent/dify-plugin-daemon v1.0.0 →core/ai-agent/dify-sandbox v1.0.0 →core/ai-agent/dify-ssrf-proxy v1.0.0 →core/ai-agent/dify-web v1.0.0 →core/ai-agent/dify-web v1.1.0 →core/ai-agent/dify-worker v1.0.0 →core/ai-agent/dify-worker v1.1.0 →core/ai-agent/dify-worker-beat v1.0.0 →core/ai-agent/dify-worker-beat v1.1.0 →core/ai-agent/dynamiq v1.0.0 →core/ai-agent/flowise v1.0.0 →core/ai-agent/flowise v1.1.0 →core/ai-agent/gemini v1.0.0 →core/ai-agent/google-adk v1.0.0 →core/ai-agent/google-adk v1.1.0 →core/ai-agent/griptape v1.0.0 →core/ai-agent/gumloop v1.0.0 →core/ai-agent/haystack v1.0.0 →core/ai-agent/langflow v1.0.0 →core/ai-agent/langflow v1.1.0 →core/ai-agent/langgraph v1.0.0 →core/ai-agent/langgraph v1.1.0 →core/ai-agent/letta v1.0.0 →core/ai-agent/librechat v1.0.0 →core/ai-agent/librechat v1.1.0 →core/ai-agent/librechat v1.2.0 →core/ai-agent/lindy v1.0.0 →core/ai-agent/livekit-agent-worker v1.0.0 →core/ai-agent/llamaindex v1.0.0 →core/ai-agent/mastra v1.0.0 →core/ai-agent/metagpt v1.0.0 →core/ai-agent/n8n v1.0.0 →core/ai-agent/n8n v1.1.0 →core/ai-agent/open-webui v1.0.0 →core/ai-agent/openai-agents-sdk v1.0.0 →core/ai-agent/openclaw v1.0.0 →core/ai-agent/openhands v1.0.0 →core/ai-agent/phidata v1.0.0 →core/ai-agent/promptflow v1.0.0 →core/ai-agent/pydantic-ai v1.0.0 →core/ai-agent/retell v1.0.0 →core/ai-agent/rivet v1.0.0 →core/ai-agent/semantic-kernel v1.0.0 →core/ai-agent/smolagents v1.0.0 →core/ai-agent/vapi v1.0.0 →core/ai-agent/vertex-ai v1.0.0 → - $refmemperluas skema (values)
BaseComponents#/schemas/BaseToolParameter
Digunakan oleh
berlaku untuk
mengimplementasikan