Skip to Content
Aionis v0.3.2 is the current Runtime baseline for SDK/API hosts, MCP clients, plugins, Substrate-backed recall, and self-managed agent loops.

Guide an Agent

Guide page — how to use this step in your SDK or loop. For the HTTP endpoint parameters, see POST /v1/guide. For the product concept, see Execution Memory.

guide compiles governed context for the next Agent turn. It is where Aionis decides what can be used now, what needs inspection, what should be avoided, and what can be rehydrated when detail is needed.

SDK Example

import { agentPromptFromGuide } from "@aionis/sdk"; const guide = await aionis.execution.guideForRole({ agent_id: "worker-2", team_id: "checkout-team", role: "worker", run_id: "run-002", task_signature: "checkout-migration", query_text: "Continue from the latest accepted checkout route.", context_mode: "compact_agent", guide: { task_context_profile: "multi_agent_handoff", }, }); await agent.run(agentPromptFromGuide(guide));

Agent Context Surfaces

SurfaceMeaning
use_nowCurrent, trusted enough for direct Agent use.
inspect_before_useRelevant but risky, contested, stale, or incomplete.
do_not_useFailed, suppressed, or unsafe as a route.
rehydratePointer to raw evidence that can expand on demand.

Execution Memory renders these surfaces as a contract. That keeps risky or invalidated history visible without turning it into direct instruction.

Task Context Profile

When the host knows what kind of work the Agent is doing, pass task_context_profile with the guide request. The profile changes how the Agent-facing context is phrased, while the same Runtime governance still decides use_now, inspect_before_use, do_not_use, and rehydrate.

const guide = await aionis.guide({ query_text: "Answer from the retained source evidence.", mode: "full_power", context_mode: "compact_agent", task_context_profile: "long_qa", consumer_agent_id: "qa-agent", run_id: "run-qa-001", });

Common profiles:

ProfileUse when
coding_verifierCoding tasks with tests, builds, validators, or acceptance checks.
long_qaSource-grounded long-memory question answering.
multi_agent_handoffPlanner, worker, verifier, reviewer, or team handoff.
loop_engineeringPlan, act, validate, repair loops.
document_integrityFile movement, extraction, migration, archive, or document identity work.

Full guide: Task Context Profiles.

Last updated on