feat(agent): auto compact — proactive session compression to reduce token cost and latency (#2982)

When a user is idle for longer than a configured TTL, nanobot **proactively** compresses the session context into a summary. This reduces token cost and first-token latency when the user returns — instead of re-processing a long stale context with an expired KV cache, the model receives a compact summary and fresh input.
2026-04-11 15:56:41 +08:00
parent e392c27f7e
commit fb6dd111e1
10 changed files with 1091 additions and 7 deletions
@@ -77,6 +77,7 @@ class AgentDefaults(Base):
    reasoning_effort: str | None = None  # low / medium / high / adaptive - enables LLM thinking mode
    timezone: str = "UTC"  # IANA timezone, e.g. "Asia/Shanghai", "America/New_York"
    unified_session: bool = False  # Share one session across all channels (single-user multi-device)
+    session_ttl_minutes: int = Field(default=0, ge=0)  # Auto /new after idle (0 = disabled)
    dream: DreamConfig = Field(default_factory=DreamConfig)