Skip to content

AGENT_AINL_OPERATING_MODEL.md

Latest commit

 

History

History
57 lines (37 loc) · 3.93 KB

AGENT_AINL_OPERATING_MODEL.md

File metadata and controls

57 lines (37 loc) · 3.93 KB
Raw

Agent + AINL operating model (long-term)

This page fixes roles, defaults, and evidence for humans and coding agents that use AINL alongside OpenClaw (or other hosts). It complements TOKEN_AND_USAGE_OBSERVABILITY.md (where signals live) and AINL_PROFILES.md (named env bundles).

Two roles

Role Owns Good fit
AINL Compiled graphs, adapters, monitors, budgets, tested pipelines Cron, guardrails, repeatable transforms, retrieval pilots
Agent Planning, synthesis, ambiguous tradeoffs, human coordination Exploration, prioritization, “what good looks like”

Push down into AINL what benefits from structure, tests, and schedules. Keep up judgment until a pattern is stable enough to formalize.

Host contract (highest leverage)

Curated artifacts only help if the host loads them:

  • Prefer .openclaw/bootstrap/session_context.md (or equivalent) over dumping full MEMORY.md when token-aware startup has run.
  • Prefer rolling budget keys and workflow/token_budget (after hydration) over re-parsing long markdown windows for gates.

If the scheduler never runs intelligence programs, or the UI never reads bootstrap output, token cost stays high regardless of AINL quality.

Default operational loop

Use this order unless measurement says otherwise:

  1. Profiles — ainl profile list / AINL_PROFILES.md; isolate paths per workspace — WORKSPACE_ISOLATION.md.
  2. Observability — identify which layer dominates spend — TOKEN_AND_USAGE_OBSERVABILITY.md.
  3. Rolling budget + hydration — bridge publishes the rolling aggregate to memory_records (workflow / budget.aggregate / record id weekly_remaining_v1); run_intelligence.py merges into cache — INTELLIGENCE_PROGRAMS.md, TOKEN_CAPS_STAGING.md. A legacy weekly_remaining_v1 SQLite table may still exist from install; ainl status prefers it when non-null, else reads weekly_remaining_tokens from the memory row.
  4. Caps — bridge (AINL_BRIDGE_REPORT_MAX_CHARS) then gateway (PROMOTER_LLM_*) — TOKEN_CAPS_STAGING.md.
  5. Intelligence programs — token-aware startup + summarizer on a real schedule.
  6. Optional accelerants — one path at a time: embedding pilot — EMBEDDING_RETRIEVAL_PILOT.md; WASM — WASM_OPERATOR_NOTES.md; TTL — TTL_MEMORY_TUNER.md.

Automate vs. pause

Automate Pause / measure first
Scheduled context build, summarization, rolling publish Provider-only “sparse attention” and similar unless documented for your API
Index refresh where you opted in One global cap for every feature without staging
Cap enforcement at bridge/gateway WASM rewrites of hot paths without traces

Agent habits (checklist)

Before loading or suggesting large context:

  1. Check budget_hydrate output from scripts/run_intelligence.py (JSON) or cache workflow/token_budget.
  2. Use ainl bridge-sizing-probe when tuning report caps.
  3. Prefer rolling budget JSON / memory.get on aggregate keys over scanning many daily files in the model prompt.

North star

Evidence over aspiration: every change should be checkable against real signals (gateway usage, markdown heuristics, memory operations, bridge reports) — not against headline savings alone.

See also