google / google/agents-cli

Agent Runtime deploy with A2A agent type

Offen
#63 1 Kommentar 0 Reaktionen 0 zugewiesene Personen Auf GitHub ansehen
Vorherrschende Sprache
Python
Sterne
5.9k
Forks
660
PR-Merge-Kennzahlen
Keine gemergten PRs in 30 T.

Beschreibung

### What is your feature suggestion?

Enable `agents-cli` deployed agents to be discovered by the new Gemini Enterprise App feature called "Agents from Agent Registry".

### What will this enable you to do?

Gemini Enterprise App (I hope I get the freaking naming right here) introduced a new feature that enables one to easily register an agent from Agent Registry. The problem is that this feature is limited to agents supporting A2A. `agents-cli deploy` actually deploys ADK agents with A2A support but Agent Runtime has no idea that the agent actually supports A2A as the agent is deployed with `agent_framework="google-adk"`.

Due to this fact, the automatic discovery feature of Agent Registry/Agent Runtime correctly discovers the agent but again it doesn't discover the A2A capability of the agent. And because of this, these agents are never recognized by the Gemini Enterprise App drawing the "Agents from Agent Registry" useless for agents deployed with `agents-cli` even though the support for A2A is actually in place.

### Additional context

Haven't tested yet but it should be possible to register the `agents-cli` deployed agent to Agent Registry manually by specify the agent card URL but this would make the same agent appear twice in the Agent Registry as the first is auto-discovered ADK agent and the second the manually registered one which is not ideal.

Beitragsleitfaden

Beitragsleitfaden öffnen

Rechercherichtung

Start at the agents-cli deploy entry point and trace how deployed ADK agents are represented for Agent Runtime and Agent Registry discovery. Read the existing deployment and discovery-related documentation or tests, then verify the deployed agent's A2A capability is recognized by Gemini Enterprise App without creating a duplicate registry entry.

Vom Indexierungsmodell aus dem Issue-Text verfasst.

Bewertung

Tech-Stack
google-cloud, python
Bereich
cloud
Issue-Typ
Feature
Schwierigkeit
4/5
Geschätzter Aufwand
3-5 Tage
Aktivitätsstatus
Ruhig
Klarheit
Größtenteils klar
Anfängerfreundlichkeit
48/100

Neue Issues direkt in Ihr Postfach

Eine kurze Übersicht über anfängerfreundliche GitHub-Issues.