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

POST /v1/guide

Compile governed context for an Agent before it acts.

Guide is where Aionis turns recalled memory and execution state into use_now, inspect_before_use, do_not_use, and rehydrate surfaces.

Request body

FieldTypeRequiredDescription
query_textstringYesWhat the Agent is about to do.
tenant_idstringNoTenant boundary.
scopestringNoProject or workspace boundary.
modestandard or full_powerNofull_power merges execution context and structured execution memory.
context_modestandard, full_power, or compact_agentNoUse compact_agent for prompt-ready context.
task_context_profilegeneral, coding_verifier, document_integrity, long_qa, multi_agent_handoff, or loop_engineeringNoTask posture for Agent context rendering. Defaults to general.
agent_rolestringNoRole such as agent, planner, worker, verifier, reviewer.
run_idstringNoCurrent execution session.
consumer_agent_idstringNoAgent that will read the guide.
consumer_team_idstringNoShared team lane.
limitnumberNoCandidate limit, max 200.
context_char_budgetnumberNoMaximum compact context character budget.
context_compaction_profilebalanced or aggressiveNoContext compression profile.
include_packetsbooleanNoInclude internal memory and guide packets for debugging.

Example

curl -s http://127.0.0.1:3001/v1/guide \ -H 'content-type: application/json' \ -d '{ "tenant_id": "default", "scope": "checkout-migration", "query_text": "Continue from the current accepted migration state.", "mode": "full_power", "context_mode": "compact_agent", "task_context_profile": "coding_verifier", "consumer_agent_id": "worker-1", "consumer_team_id": "checkout-team", "run_id": "run-001" }'

SDK

const guide = await aionis.guide({ query_text: "Continue from the current accepted migration state.", mode: "full_power", context_mode: "compact_agent", task_context_profile: "coding_verifier", consumer_agent_id: "worker-1", consumer_team_id: "checkout-team", run_id: "run-001", });

Response

Important fields:

FieldMeaning
contract_versionaionis_guide_result_v1.
guide_trace_idStable handle for feedback attribution. Preserve it.
agent_context.prompt_textPrompt-ready governed context.
agent_context.task_context_profileProfile used to render the Agent context.
agent_context.use_now_memory_idsMemory allowed to directly guide action.
agent_context.inspect_before_use_memory_idsRelevant memory that needs verification.
agent_context.do_not_use_memory_idsMemory blocked from direct action.
agent_context.rehydrate_memory_idsEvidence pointers to expand on demand.
source_mapRuntime surfaces used to compile the guide.

Feedback requirement

Use guide_trace_id when you call POST /v1/feedback. Only attribute outcome to memory IDs that this guide exposed.

Task context profiles

task_context_profile is selected by your host or adapter. It does not change recall, storage, lifecycle state, or admission authority. It only changes the Agent-facing posture of the compiled context.

Use common mappings like:

Host taskProfile
Coding task with required tests or validatorscoding_verifier
Source-grounded long-memory question answeringlong_qa
Planner/worker/reviewer handoffmulti_agent_handoff
Plan-execute-validate-repair looploop_engineering
File movement, extraction, or document identity workdocument_integrity

See Task Context Profiles.

Common errors

ProblemFix
Empty or weak contextCheck embeddings, scope, and whether evidence was observed.
Useful memory appears under inspectVerify it before direct use, or send feedback after actual use.
Feedback later failsPreserve the exact guide_trace_id returned by this endpoint.
Last updated on