margot-bridge

What problem does it solve?

It bridges Pi-CEO’s orchestrator to Margot’s standalone Gemini-powered research and image-generation MCP server, so research-shaped intents can run with the right tooling and outputs returned reliably.

Core Features & Use Cases

• MCP contract for deep research and diagnostics: Invokes Margot MCP tools like deep_research, deep_research_max, and corpus_status to produce research or corpus readiness signals for the caller.
• Async interaction management: Supports a deep_research_max dispatch-and-poll workflow using interaction_id plus local JSONL inflight/complete tracking for later retrieval and downstream drafting.
• Image generation routing: Calls image_generate(prompt, aspect_ratio, image_size) when the orchestrated intent is image-shaped.
• Safety and isolation guardrails: Avoids modifying Margot’s owned source and prevents accidental secret leakage by keeping Gemini API keys inside Margot’s environment.

Quick Start

Ensure Margot is registered as an MCP server in the runtime that runs Pi-CEO, then invoke this skill with an intent payload containing research (optionally use_corpus: true) or image to receive the MCP tool response or a dispatched interaction_id.