Google Antigravity Python SDK — v0.1.10 Release Notes

[NTaylorMullen]

The 0.1.10 release introduces support for Gemini Prioritized Inference service tiers, real-time token usage event streaming, stateful context-aware hook decorators, and explicit tool call correlation IDs across hook callbacks. It also standardizes default system instruction merging behavior, fixes WebSocket compaction events, and expands local Gemma model documentation.

🌟 Key Highlights

• Gemini Prioritized Inference Service Tier: Configure agents to utilize Gemini Prioritized Inference service tiers for high-priority model execution with automated graceful fallback.

model_opts = GeminiModelOptions(
service_tier=ServiceTier.PRIORITY,
)
config = LocalAgentConfig(
model=ModelTarget(
name="gemini-3.6-flash",
endpoint=GeminiAPIEndpoint(options=model_opts),
),
)

- **Tool Call ID Correlation in Lifecycle Hooks**: Inspect `call_id` attributes on tool executions, errors, and hooks to correlate multi-step tool invocations across lifecycle callbacks.
```python
from google.antigravity import ToolExecutionError, types
from google.antigravity.hooks import hooks

@hooks.pre_tool_call_decide
async def pre_tool(data: types.ToolCall) -> types.HookResult:
    print(f"Executing tool {data.name} (call_id={data.id})")
    return types.HookResult(allow=True)

@hooks.on_tool_error
async def on_error(error: ToolExecutionError) -> None:
    print(f"Tool {error.tool_name} failed (call_id={error.call_id}): {error}")

• Context-Aware Hook Decorators: Decorate hook handlers (@hooks.pre_turn, @hooks.post_tool_call, etc.) that optionally accept HookContext as a parameter to maintain state and share data across lifecycle callbacks.

  from google.antigravity import types
  from google.antigravity.hooks import HookContext, hooks
  
  @hooks.pre_turn
  async def inspect_prompt(context: HookContext, data: str) -> types.HookResult:
      context.set_data("user_prompt", data)
      return types.HookResult(allow=True)

• ActionCompaction Event Emission & Hook: Track context window compaction notifications over WebSockets and intercept them using @hooks.on_compaction.

  from google.antigravity.hooks import hooks
  
  @hooks.on_compaction
  async def handle_compaction(data) -> None:
      print(f"Context compaction occurred: {data}")

• Standardized System Instructions Strategy: Plain string instructions default to appending to built-in instructions. To override and completely replace built-in instructions, pass CustomSystemInstructions.

  from google.antigravity import CustomSystemInstructions, LocalAgentConfig
  
  # Override and replace built-in instructions completely
  config = LocalAgentConfig(
      system_instructions=CustomSystemInstructions(
          text="You are a specialized code reviewer."
      )
  )

---

📋 Detailed Changes

• Features & Enhancements

  • Tool Execution Call ID Correlation: Added call_id attributes to ToolCall.id, ToolResult.id, ToolExecutionError.call_id, and hook payloads to enable tracing of tool execution steps.
  • Live Token Usage Reporting: Introduced real-time UsageUpdate event streaming so token usage accumulates live during agent execution rather than delaying updates until state transitions.
  • Context-Aware Hook Decorators: Enabled decorated hook handlers to optionally accept HookContext parameters for stateful hook implementations while preserving backward compatibility for stateless handlers.
  • Interactive CLI Spinner: Updated CLI interactive loop spinner to list all active tool names when running concurrent tool calls (e.g., Running tools 'tool_a', 'tool_b').
  • Module Re-exports: Re-exported ReadUrlContentResult and SearchWebResult in connections.local for uniform tool result access.
  • Local Gemma Model Documentation: Added guides and tutorials for running agents locally with Gemma models using LiteRT and OpenAI-compatible endpoints.

• Model & Default Changes

  • System Instruction Append Policy: Standardized string system instructions across all connection types to default to an append strategy (combining custom string instructions with built-in instructions). To override this default and completely replace built-in instructions, pass CustomSystemInstructions(text=...).
  • LiteRT Token Output Limit: Increased max_output_tokens default in LiteRT local server configuration from 8,192 to 16,384 tokens to prevent truncation during complex reasoning and generation tasks.

• Bug Fixes

  • ActionCompaction Event Emission: Fixed issue where compaction notifications were suppressed in external SDK releases, causing @hooks.on_compaction handlers and conversation.compaction_indices tracking to fail; compaction events now emit properly over WebSockets.
  • MCP Test Server Startup: Fixed a race condition where the HTTP port was exposed before uvicorn server startup completed, which previously caused intermittent ConnectionRefusedError failures during test initialization.