II.
AgentCoreImpl overview
Reference · liveagent-core-impl:antigravity-cli.core@current
Antigravity CLI Core (current) overview
Inspect the raw attributes, linked wiki pages, and inbound or outbound graph edges for agent-core-impl:antigravity-cli.core@current.
Attributes
displayName
Antigravity CLI Core (current)
agentVersionId
agentVersion:antigravity:2-0-11
packageRef
loopIteratorPolicy
tool-use-loop
loopIteratorNotes
Tool-use loop over Gemini's `generateContent` (streaming) API, inherited from Gemini CLI architecture. Each turn, the model emits `functionCall` parts; the CLI executes them and feeds back `functionResponse` parts. The loop advances until a chunk arrives with `finishReason: STOP` (or `MAX_TOKENS` / `SAFETY`). Multi-model support routes Claude Sonnet 4.6 and GPT-OSS-120B through the same loop abstraction.
contextManagementStrategy
platform-managed
compactionTriggerNotes
Antigravity CLI relies on the model's context window directly; no in-CLI compaction is performed for the @current pin.
subagentInvokerPolicy
tool-call-handoff
subagentInvokerNotes
Antigravity CLI follows a workflow-driven dispatch model: subagent invocation is a structured tool-call whose target name resolves to another agent or workflow definition; the CLI dispatches a child generateContent loop for the sub-task.
resultSynthesisPolicy
model-direct
resultSynthesisNotes
Final text parts are streamed as they arrive; workflow hooks fire after the last chunk with `finishReason: STOP` is observed.
stopDetectionStrategy
structured-end-event
transportClientLibrary
@google/genai
supportedTransportProtocols
parallelToolCallHandling
native
streamingFidelity
full
thinkingChannelHandling
passthrough
Outgoing edges
composes1
- agent:antigravity-cli·AgentProductAntigravity CLI
emits_message_type7
- protocol-message:gemini-content·ProtocolMessageGemini Content (request/response envelope)
- protocol-message:gemini-part-text·ProtocolMessageGemini Part / text
- protocol-message:gemini-part-inline-data·ProtocolMessageGemini Part / inlineData
- protocol-message:gemini-part-function-call·ProtocolMessageGemini Part / functionCall
- protocol-message:gemini-part-function-response·ProtocolMessageGemini Part / functionResponse
- protocol-message:gemini-candidate·ProtocolMessageGemini Candidate
- protocol-message:gemini-generate-content-response·ProtocolMessageGemini GenerateContentResponse
realizes1
- layer:4-agent-core·LayerAgent-Core
speaks1
- model-transport:gemini-generate-content·ModelTransportProtocolGemini generateContent
supports3
- capability:streaming·CapabilityStreaming
- capability:supports-thinking·CapabilitySupports thinking
- capability:supports-tool-use·CapabilitySupports tool use
supports_effort_level3
- effort-level:low·EffortLevelEffort: low
- effort-level:medium·EffortLevelEffort: medium
- effort-level:high·EffortLevelEffort: high
supports_orchestration_primitive3
- orch-primitive:agent-dispatch·OrchestrationPrimitiveAgent Dispatch
- orch-primitive:multi-agent-dispatch·OrchestrationPrimitiveMulti-Agent Dispatch
- orch-primitive:session-handoff·OrchestrationPrimitiveSession Handoff
Incoming edges
None.