At a glance
- Did you know
- Compaction summarises a long context so the agent can keep going without hitting the window limit.
- Rule of thumb
- Keep decisions, open tasks and file paths verbatim; summarise the rest.
- Watch out
- A model call per compaction; loses details it should keep.
Its settings in ScaleMySystems
What you can change on it in the lab, and what each number does.
- Concurrency (slots)
- Calls it can work on at once (per instance when autoscaling). Needed slots ≈ calls per second × seconds each takes. Extra calls wait in the queue, then are rejected.
- Queue capacity
- Calls allowed to wait for a free slot. A bigger queue absorbs bursts but adds waiting time; 0 rejects straight away.
- vCPU
- CPU cores available. Needed ≈ calls per second × CPU per request ÷ 1000. Above about 90% busy it slows down and may crash.
- CPU per request (ms)
- CPU time each call burns. Part of what the component is; make the code cheaper to lower it.
- Latency (p50) (ms)
- Typical time one call takes here, before any calls it makes itself. Slots are held for this long.
- Latency (p99) (ms)
- The slowest 1% of calls. Callers must set their timeout above this or they cut off healthy calls.
- Timeout to dependencies (ms)
- How long it waits for each call it makes before giving up. Set it above the dependency’s p99 but low enough that a dead dependency does not hold every slot.
- Error rate
- Share of calls that fail on their own (bugs, declines, flaky tools). An agent that calls it on every step multiplies this.
- Quota (req/min)
- Calls allowed per minute (an API key or provider limit). Calls over it get 429s. Needed = calls per second × 60, with headroom.
Problems that start with it
- Compaction calls overflow Long sessions are compacted every step by a summariser with 20 slots, about 600 ms each.
- Harness compaction backs up The harness compacts context on every turn with a summariser that has 4 slots.
- Deep research: from launch build to real traffic A multi-agent research product on its first servers: the lead agent fans each question out to sub-agents that search and browse the web, and the browser tool, sub-agent pool and model capacity are an eighth of their real size. Questions arrive one a second and each takes minutes. One in ten fails.
- Coding agent platform: the model provider goes down A coding agent platform: each task is a loop of model calls, tool calls, code runs in a sandbox and checkpoints, minutes long. Twenty seconds in, the model provider has an outage for 30 seconds. Every step that needs the model fails, and with it the whole task: two in three tasks are lost.
Full lesson coming soon
Lessons are being added one component at a time: how it works, its settings, a live exercise to break and fix it, interview questions and a short check. See the Circuit breaker lesson for what one looks like.