Filters
Rate limiting, deduplication and token budget — on by default, fully tunable.
Last updated October 1, 2026
Three mechanical filters run on every message. All are enabled by default. You can tune or disable each one from the gateway edit page.
Rate limiting
Limits how many messages a single user_id can send in a time window.
| Parameter | Default | Range |
|---|---|---|
| Max messages | 10 | 10–30 |
| Window | 10s | 10–60 seconds |
Deduplication
Catches identical messages (same user_id + message hash) within a time window.
| Parameter | Default | Range |
|---|---|---|
| Window | 1 min | 1–3 minutes |
Token budget
Blocks messages that exceed a token count. Prevents token-stuffed payloads from reaching your agent.
| Parameter | Default | Range |
|---|---|---|
| Max tokens | 4,096 | 50–10,000 |
Above 4,096 tokens a warning is shown — most LLM context windows cap well below 10K per message.