Amphibious¶
Bridgic Amphibious is a dual-mode agent framework that lets you build agents operating in both LLM-driven and deterministic modes, with automatic fallback between them. Instead of choosing between full autonomy and rigid workflows, Amphibious gives you both — in the same agent.
The framework is built on two core design principles:
- Agent = Think Units + Context Orchestration — An agent is defined by declaring
CognitiveWorkerthink units and orchestrating them with scoped context, rather than wiring low-level LLM calls manually. - Functional Execution vs. Decision Making — Decoupled — Tools and skills are pure capabilities, while when and how to invoke them is handled by either LLM reasoning (
on_agent) or developer-defined workflows (on_workflow). - Human-in-the-Loop — Three entry points for human interaction: code-level
request_human(), workflowHumanCallyield, and the auto-injectedrequest_humantool for LLM-autonomous requests.
Essential¶
These tutorials cover the fundamental concepts you need to build amphibious agents:
- Quick Start: Build your first amphibious agent in 5 minutes — run both Agent mode and Workflow mode to see the dual-mode experience firsthand.
- Dual-Mode Orchestration: Master the two orchestration modes —
on_agentfor LLM-driven decision making andon_workflowfor deterministic step-by-step execution. - RunMode: Explore the four run modes and learn how Amphiflow mode automatically switches between Workflow and Agent when things go wrong.
- Customizing the OTC Cycle: Override hooks in the Observe-Think-Act cycle — inject custom observations, reshape LLM messages, intercept tool calls, and post-process outputs.
- Execution Tracing: Record, export, and analyze the full execution trace of your agent for debugging and optimization.
Advanced¶
These tutorials dive deeper into specific capabilities for advanced use cases:
- Phase Annotation: Organize complex agent execution into structured phases using the
snapshotcontext manager. - CognitiveWorker & think_unit: Understand the framework's atomic building block — the pure thinking unit that only decides what to do, plus its declarative execution configuration and error strategies.
- Context & Exposure: Control what the LLM sees — manage agent state with Context and choose between full disclosure and progressive reveal with Exposure strategies.
- Cognitive Policies: Enable multi-round deliberation before action — let the agent gather details (Acquiring), mentally rehearse plans (Rehearsal), and assess information quality (Reflection).
- Cognitive History: Understand the three-tier memory system — Working Memory, Short-term Memory, and Long-term Memory — that automatically manages execution history.
This architecture makes Bridgic Amphibious a powerful platform for building agents that are both reliable and adaptive — bridging the precision of deterministic workflows with the creative problem-solving of LLM reasoning.