feat: configurable heartbeat session
This commit is contained in:
Zach Knickerbocker
committed by
Peter Steinberger
Peter Steinberger
parent
db61451c67
commit
7725dd6795
@@ -24,7 +24,7 @@ Unknown keys, malformed types, or invalid values cause the Gateway to **refuse t
|
||||
|
||||
When validation fails:
|
||||
- The Gateway does not boot.
|
||||
- Only diagnostic commands are allowed (for example: `clawdbot doctor`, `clawdbot logs`, `clawdbot health`, `clawdbot status`, `clawdbot gateway status`, `clawdbot gateway probe`, `clawdbot help`).
|
||||
- Only diagnostic commands are allowed (for example: `clawdbot doctor`, `clawdbot logs`, `clawdbot health`, `clawdbot status`, `clawdbot service`, `clawdbot help`).
|
||||
- Run `clawdbot doctor` to see the exact issues.
|
||||
- Run `clawdbot doctor --fix` (or `--yes`) to apply migrations/repairs.
|
||||
|
||||
@@ -1414,7 +1414,7 @@ Each `agents.defaults.models` entry can include:
|
||||
- `alias` (optional model shortcut, e.g. `/opus`).
|
||||
- `params` (optional provider-specific API params passed through to the model request).
|
||||
|
||||
`params` is also applied to streaming runs (embedded agent + compaction). Supported keys today: `temperature`, `maxTokens`, `cacheControlTtl` (`"5m"` or `"1h"`, Anthropic API + OpenRouter Anthropic models only; ignored for Anthropic OAuth/Claude Code tokens). These merge with call-time options; caller-supplied values win. `temperature` is an advanced knob—leave unset unless you know the model’s defaults and need a change. Clawdbot includes the `extended-cache-ttl-2025-04-11` beta flag for Anthropic API; keep it if you override provider headers.
|
||||
`params` is also applied to streaming runs (embedded agent + compaction). Supported keys today: `temperature`, `maxTokens`. These merge with call-time options; caller-supplied values win. `temperature` is an advanced knob—leave unset unless you know the model’s defaults and need a change.
|
||||
|
||||
Example:
|
||||
|
||||
@@ -1569,7 +1569,7 @@ Example:
|
||||
}
|
||||
```
|
||||
|
||||
#### `agents.defaults.contextPruning` (TTL-aware tool-result pruning)
|
||||
#### `agents.defaults.contextPruning` (tool-result pruning)
|
||||
|
||||
`agents.defaults.contextPruning` prunes **old tool results** from the in-memory context right before a request is sent to the LLM.
|
||||
It does **not** modify the session history on disk (`*.jsonl` remains complete).
|
||||
@@ -1580,9 +1580,11 @@ High level:
|
||||
- Never touches user/assistant messages.
|
||||
- Protects the last `keepLastAssistants` assistant messages (no tool results after that point are pruned).
|
||||
- Protects the bootstrap prefix (nothing before the first user message is pruned).
|
||||
- Mode:
|
||||
- `cache-ttl`: pruning only runs when the last Anthropic call for the session is **older** than `ttl`.
|
||||
When it runs, it uses the same soft-trim + hard-clear behavior as before.
|
||||
- Modes:
|
||||
- `adaptive`: soft-trims oversized tool results (keep head/tail) when the estimated context ratio crosses `softTrimRatio`.
|
||||
Then hard-clears the oldest eligible tool results when the estimated context ratio crosses `hardClearRatio` **and**
|
||||
there’s enough prunable tool-result bulk (`minPrunableToolChars`).
|
||||
- `aggressive`: always replaces eligible tool results before the cutoff with the `hardClear.placeholder` (no ratio checks).
|
||||
|
||||
Soft vs hard pruning (what changes in the context sent to the LLM):
|
||||
- **Soft-trim**: only for *oversized* tool results. Keeps the beginning + end and inserts `...` in the middle.
|
||||
@@ -1596,41 +1598,44 @@ Notes / current limitations:
|
||||
- Tool results containing **image blocks are skipped** (never trimmed/cleared) right now.
|
||||
- The estimated “context ratio” is based on **characters** (approximate), not exact tokens.
|
||||
- If the session doesn’t contain at least `keepLastAssistants` assistant messages yet, pruning is skipped.
|
||||
- `cache-ttl` only activates for Anthropic API calls (and OpenRouter Anthropic models).
|
||||
- After a prune, the TTL window resets so subsequent requests keep cache until `ttl` expires again.
|
||||
- For best results, match `contextPruning.ttl` to the model `cacheControlTtl` you set in `agents.defaults.models.*.params`.
|
||||
- In `aggressive` mode, `hardClear.enabled` is ignored (eligible tool results are always replaced with `hardClear.placeholder`).
|
||||
|
||||
Default (off, unless Anthropic auth profiles are detected):
|
||||
Default (adaptive):
|
||||
```json5
|
||||
{
|
||||
agents: { defaults: { contextPruning: { mode: "adaptive" } } }
|
||||
}
|
||||
```
|
||||
|
||||
To disable:
|
||||
```json5
|
||||
{
|
||||
agents: { defaults: { contextPruning: { mode: "off" } } }
|
||||
}
|
||||
```
|
||||
|
||||
Enable TTL-aware pruning:
|
||||
Defaults (when `mode` is `"adaptive"` or `"aggressive"`):
|
||||
- `keepLastAssistants`: `3`
|
||||
- `softTrimRatio`: `0.3` (adaptive only)
|
||||
- `hardClearRatio`: `0.5` (adaptive only)
|
||||
- `minPrunableToolChars`: `50000` (adaptive only)
|
||||
- `softTrim`: `{ maxChars: 4000, headChars: 1500, tailChars: 1500 }` (adaptive only)
|
||||
- `hardClear`: `{ enabled: true, placeholder: "[Old tool result content cleared]" }`
|
||||
|
||||
Example (aggressive, minimal):
|
||||
```json5
|
||||
{
|
||||
agents: { defaults: { contextPruning: { mode: "cache-ttl" } } }
|
||||
agents: { defaults: { contextPruning: { mode: "aggressive" } } }
|
||||
}
|
||||
```
|
||||
|
||||
Defaults (when `mode` is `"cache-ttl"`):
|
||||
- `ttl`: `"5m"`
|
||||
- `keepLastAssistants`: `3`
|
||||
- `softTrimRatio`: `0.3`
|
||||
- `hardClearRatio`: `0.5`
|
||||
- `minPrunableToolChars`: `50000`
|
||||
- `softTrim`: `{ maxChars: 4000, headChars: 1500, tailChars: 1500 }`
|
||||
- `hardClear`: `{ enabled: true, placeholder: "[Old tool result content cleared]" }`
|
||||
|
||||
Example (cache-ttl tuned):
|
||||
Example (adaptive tuned):
|
||||
```json5
|
||||
{
|
||||
agents: {
|
||||
defaults: {
|
||||
contextPruning: {
|
||||
mode: "cache-ttl",
|
||||
ttl: "5m",
|
||||
mode: "adaptive",
|
||||
keepLastAssistants: 3,
|
||||
softTrimRatio: 0.3,
|
||||
hardClearRatio: 0.5,
|
||||
@@ -1737,12 +1742,9 @@ Z.AI models are available as `zai/<model>` (e.g. `zai/glm-4.7`) and require
|
||||
`30m`. Set `0m` to disable.
|
||||
- `model`: optional override model for heartbeat runs (`provider/model`).
|
||||
- `includeReasoning`: when `true`, heartbeats will also deliver the separate `Reasoning:` message when available (same shape as `/reasoning on`). Default: `false`.
|
||||
- `activeHours`: optional local-time window that controls when heartbeats run.
|
||||
- `start`: start time (HH:MM, 24h). Inclusive.
|
||||
- `end`: end time (HH:MM, 24h). Exclusive. Use `"24:00"` for end-of-day.
|
||||
- `timezone`: `"user"` (default), `"local"`, or an IANA timezone id.
|
||||
- `target`: optional delivery channel (`last`, `whatsapp`, `telegram`, `discord`, `slack`, `signal`, `imessage`, `none`). Default: `last`.
|
||||
- `session`: optional session key to control which session the heartbeat runs in. Default: `main`.
|
||||
- `to`: optional recipient override (channel-specific id, e.g. E.164 for WhatsApp, chat id for Telegram).
|
||||
- `target`: optional delivery channel (`last`, `whatsapp`, `telegram`, `discord`, `slack`, `msteams`, `signal`, `imessage`, `none`). Default: `last`.
|
||||
- `prompt`: optional override for the heartbeat body (default: `Read HEARTBEAT.md if it exists (workspace context). Follow it strictly. Do not infer or repeat old tasks from prior chats. If nothing needs attention, reply HEARTBEAT_OK.`). Overrides are sent verbatim; include a `Read HEARTBEAT.md` line if you still want the file read.
|
||||
- `ackMaxChars`: max chars allowed after `HEARTBEAT_OK` before delivery (default: 300).
|
||||
|
||||
@@ -1773,7 +1775,6 @@ Note: `applyPatch` is only under `tools.exec`.
|
||||
- `tools.web.fetch.maxChars` (default 50000)
|
||||
- `tools.web.fetch.timeoutSeconds` (default 30)
|
||||
- `tools.web.fetch.cacheTtlMinutes` (default 15)
|
||||
- `tools.web.fetch.maxRedirects` (default 3)
|
||||
- `tools.web.fetch.userAgent` (optional override)
|
||||
- `tools.web.fetch.readability` (default true; disable to use basic HTML cleanup only)
|
||||
- `tools.web.fetch.firecrawl.enabled` (default true when an API key is set)
|
||||
@@ -1840,7 +1841,7 @@ Example:
|
||||
|
||||
`agents.defaults.subagents` configures sub-agent defaults:
|
||||
- `model`: default model for spawned sub-agents (string or `{ primary, fallbacks }`). If omitted, sub-agents inherit the caller’s model unless overridden per agent or per call.
|
||||
- `maxConcurrent`: max concurrent sub-agent runs (default 8)
|
||||
- `maxConcurrent`: max concurrent sub-agent runs (default 1)
|
||||
- `archiveAfterMinutes`: auto-archive sub-agent sessions after N minutes (default 60; set `0` to disable)
|
||||
- Per-subagent tool policy: `tools.subagents.tools.allow` / `tools.subagents.tools.deny` (deny wins)
|
||||
|
||||
@@ -1974,7 +1975,7 @@ Notes:
|
||||
|
||||
`agents.defaults.maxConcurrent` sets the maximum number of embedded agent runs that can
|
||||
execute in parallel across sessions. Each session is still serialized (one run
|
||||
per session key at a time). Default: 4.
|
||||
per session key at a time). Default: 1.
|
||||
|
||||
### `agents.defaults.sandbox`
|
||||
|
||||
@@ -2452,9 +2453,6 @@ Controls session scoping, reset policy, reset triggers, and where the session st
|
||||
dm: { mode: "idle", idleMinutes: 240 },
|
||||
group: { mode: "idle", idleMinutes: 120 }
|
||||
},
|
||||
resetByChannel: {
|
||||
discord: { mode: "idle", idleMinutes: 10080 }
|
||||
},
|
||||
resetTriggers: ["/new", "/reset"],
|
||||
// Default is already per-agent under ~/.clawdbot/agents/<agentId>/sessions/sessions.json
|
||||
// You can override with {agentId} templating:
|
||||
@@ -2490,7 +2488,7 @@ Fields:
|
||||
- `idleMinutes`: sliding idle window in minutes. When daily + idle are both configured, whichever expires first wins.
|
||||
- `resetByType`: per-session overrides for `dm`, `group`, and `thread`.
|
||||
- If you only set legacy `session.idleMinutes` without any `reset`/`resetByType`, Clawdbot stays in idle-only mode for backward compatibility.
|
||||
- `resetByChannel`: channel-specific reset policy overrides (keyed by channel id, applies to all session types for that channel; overrides `reset`/`resetByType`).
|
||||
- `heartbeatIdleMinutes`: optional idle override for heartbeat checks (daily reset still applies when enabled).
|
||||
- `agentToAgent.maxPingPongTurns`: max reply-back turns between requester/target (0–5, default 5).
|
||||
- `sendPolicy.default`: `allow` or `deny` fallback when no rule matches.
|
||||
- `sendPolicy.rules[]`: match by `channel`, `chatType` (`direct|group|room`), or `keyPrefix` (e.g. `cron:`). First deny wins; otherwise allow.
|
||||
@@ -2617,13 +2615,10 @@ Defaults:
|
||||
// noSandbox: false,
|
||||
// executablePath: "/Applications/Brave Browser.app/Contents/MacOS/Brave Browser",
|
||||
// attachOnly: false, // set true when tunneling a remote CDP to localhost
|
||||
// snapshotDefaults: { mode: "efficient" }, // tool/CLI default snapshot preset
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Note: `browser.snapshotDefaults` only affects Clawdbot's browser tool + CLI. Direct HTTP clients must pass `mode` explicitly.
|
||||
|
||||
### `ui` (Appearance)
|
||||
|
||||
Optional accent color used by the native apps for UI chrome (e.g. Talk Mode bubble tint).
|
||||
@@ -2647,13 +2642,6 @@ Defaults:
|
||||
- bind: `loopback`
|
||||
- port: `18789` (single port for WS + HTTP)
|
||||
|
||||
Bind modes:
|
||||
- `loopback`: `127.0.0.1` (local-only)
|
||||
- `lan`: `0.0.0.0` (all interfaces)
|
||||
- `tailnet`: Tailscale IPv4 address (100.64.0.0/10)
|
||||
- `auto`: prefer loopback, fall back to LAN if loopback cannot bind
|
||||
- `custom`: `gateway.customBindHost` (IPv4), fallback to LAN if unavailable
|
||||
|
||||
```json5
|
||||
{
|
||||
gateway: {
|
||||
@@ -2684,15 +2672,14 @@ Notes:
|
||||
- `clawdbot gateway` refuses to start unless `gateway.mode` is set to `local` (or you pass the override flag).
|
||||
- `gateway.port` controls the single multiplexed port used for WebSocket + HTTP (control UI, hooks, A2UI).
|
||||
- OpenAI Chat Completions endpoint: **disabled by default**; enable with `gateway.http.endpoints.chatCompletions.enabled: true`.
|
||||
- OpenResponses endpoint: **disabled by default**; enable with `gateway.http.endpoints.responses.enabled: true`.
|
||||
- Precedence: `--port` > `CLAWDBOT_GATEWAY_PORT` > `gateway.port` > default `18789`.
|
||||
- Non-loopback binds (`lan`/`tailnet`/`custom`, or `auto` when loopback is unavailable) require auth. Use `gateway.auth.token` (or `CLAWDBOT_GATEWAY_TOKEN`).
|
||||
- Non-loopback binds (`lan`/`tailnet`/`auto`) require auth. Use `gateway.auth.token` (or `CLAWDBOT_GATEWAY_TOKEN`).
|
||||
- The onboarding wizard generates a gateway token by default (even on loopback).
|
||||
- `gateway.remote.token` is **only** for remote CLI calls; it does not enable local gateway auth. `gateway.token` is ignored.
|
||||
|
||||
Auth and Tailscale:
|
||||
- `gateway.auth.mode` sets the handshake requirements (`token` or `password`).
|
||||
- `gateway.auth.token` stores the shared token for token auth (used by the CLI on the same machine and as the bootstrap credential for device pairing).
|
||||
- `gateway.auth.token` stores the shared token for token auth (used by the CLI on the same machine).
|
||||
- When `gateway.auth.mode` is set, only that method is accepted (plus optional Tailscale headers).
|
||||
- `gateway.auth.password` can be set here, or via `CLAWDBOT_GATEWAY_PASSWORD` (recommended).
|
||||
- `gateway.auth.allowTailscale` allows Tailscale Serve identity headers
|
||||
@@ -2701,9 +2688,6 @@ Auth and Tailscale:
|
||||
`true`, Serve requests do not need a token/password; set `false` to require
|
||||
explicit credentials. Defaults to `true` when `tailscale.mode = "serve"` and
|
||||
auth mode is not `password`.
|
||||
- After pairing, the Gateway issues **device tokens** scoped to the device role + scopes.
|
||||
These are returned in `hello-ok.auth.deviceToken`; clients should persist and reuse them
|
||||
instead of the shared token. Rotate/revoke via `device.token.rotate`/`device.token.revoke`.
|
||||
- `gateway.tailscale.mode: "serve"` uses Tailscale Serve (tailnet only, loopback bind).
|
||||
- `gateway.tailscale.mode: "funnel"` exposes the dashboard publicly; requires auth.
|
||||
- `gateway.tailscale.resetOnExit` resets Serve/Funnel config on shutdown.
|
||||
@@ -2712,7 +2696,6 @@ Remote client defaults (CLI):
|
||||
- `gateway.remote.url` sets the default Gateway WebSocket URL for CLI calls when `gateway.mode = "remote"`.
|
||||
- `gateway.remote.token` supplies the token for remote calls (leave unset for no auth).
|
||||
- `gateway.remote.password` supplies the password for remote calls (leave unset for no auth).
|
||||
- `gateway.remote.tlsFingerprint` pins the gateway TLS cert fingerprint (sha256).
|
||||
|
||||
macOS app behavior:
|
||||
- Clawdbot.app watches `~/.clawdbot/clawdbot.json` and switches modes live when `gateway.mode` or `gateway.remote.url` changes.
|
||||
@@ -2726,36 +2709,12 @@ macOS app behavior:
|
||||
remote: {
|
||||
url: "ws://gateway.tailnet:18789",
|
||||
token: "your-token",
|
||||
password: "your-password",
|
||||
tlsFingerprint: "sha256:ab12cd34..."
|
||||
password: "your-password"
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### `gateway.nodes` (Node command allowlist)
|
||||
|
||||
The Gateway enforces a per-platform command allowlist for `node.invoke`. Nodes must both
|
||||
**declare** a command and have it **allowed** by the Gateway to run it.
|
||||
|
||||
Use this section to extend or deny commands:
|
||||
|
||||
```json5
|
||||
{
|
||||
gateway: {
|
||||
nodes: {
|
||||
allowCommands: ["custom.vendor.command"], // extra commands beyond defaults
|
||||
denyCommands: ["sms.send"] // block a command even if declared
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Notes:
|
||||
- `allowCommands` extends the built-in per-platform defaults.
|
||||
- `denyCommands` always wins (even if the node claims the command).
|
||||
- `node.invoke` rejects commands that are not declared by the node.
|
||||
|
||||
### `gateway.reload` (Config hot reload)
|
||||
|
||||
The Gateway watches `~/.clawdbot/clawdbot.json` (or `CLAWDBOT_CONFIG_PATH`) and applies changes automatically.
|
||||
@@ -3003,7 +2962,7 @@ Auto-generated certs require `openssl` on PATH; if generation fails, the bridge
|
||||
|
||||
### `discovery.wideArea` (Wide-Area Bonjour / unicast DNS‑SD)
|
||||
|
||||
When enabled, the Gateway writes a unicast DNS-SD zone for `_clawdbot-gw._tcp` under `~/.clawdbot/dns/` using the standard discovery domain `clawdbot.internal.`
|
||||
When enabled, the Gateway writes a unicast DNS-SD zone for `_clawdbot-bridge._tcp` under `~/.clawdbot/dns/` using the standard discovery domain `clawdbot.internal.`
|
||||
|
||||
To make iOS/Android discover across networks (Vienna ⇄ London), pair this with:
|
||||
- a DNS server on the gateway host serving `clawdbot.internal.` (CoreDNS is recommended)
|
||||
@@ -3033,9 +2992,6 @@ Template placeholders are expanded in `tools.media.*.models[].args` and `tools.m
|
||||
| `{{From}}` | Sender identifier (E.164 for WhatsApp; may differ per channel) |
|
||||
| `{{To}}` | Destination identifier |
|
||||
| `{{MessageSid}}` | Channel message id (when available) |
|
||||
| `{{MessageSidFull}}` | Provider-specific full message id when `MessageSid` is shortened |
|
||||
| `{{ReplyToId}}` | Reply-to message id (when available) |
|
||||
| `{{ReplyToIdFull}}` | Provider-specific full reply-to id when `ReplyToId` is shortened |
|
||||
| `{{SessionId}}` | Current session UUID |
|
||||
| `{{IsNewSession}}` | `"true"` when a new session was created |
|
||||
| `{{MediaUrl}}` | Inbound media pseudo-URL (if present) |
|
||||
|
||||
@@ -127,18 +127,26 @@ Example: two agents, only the second agent runs heartbeats.
|
||||
- `every`: heartbeat interval (duration string; default unit = minutes).
|
||||
- `model`: optional model override for heartbeat runs (`provider/model`).
|
||||
- `includeReasoning`: when enabled, also deliver the separate `Reasoning:` message when available (same shape as `/reasoning on`).
|
||||
- `session`: optional session key for heartbeat runs.
|
||||
- `main` (default): agent main session.
|
||||
- Explicit session key (copy from `clawdbot sessions --json` or the [sessions CLI](/cli/sessions)).
|
||||
- Session key formats: see [Sessions](/concepts/session) and [Groups](/concepts/groups).
|
||||
- `target`:
|
||||
- `last` (default): deliver to the last used external channel.
|
||||
- explicit channel: `whatsapp` / `telegram` / `discord` / `slack` / `signal` / `imessage`.
|
||||
- explicit channel: `whatsapp` / `telegram` / `discord` / `slack` / `msteams` / `signal` / `imessage`.
|
||||
- `none`: run the heartbeat but **do not deliver** externally.
|
||||
- `to`: optional recipient override (E.164 for WhatsApp, chat id for Telegram, etc.).
|
||||
- `to`: optional recipient override (channel-specific id, e.g. E.164 for WhatsApp or a Telegram chat id).
|
||||
- `prompt`: overrides the default prompt body (not merged).
|
||||
- `ackMaxChars`: max chars allowed after `HEARTBEAT_OK` before delivery.
|
||||
|
||||
## Delivery behavior
|
||||
|
||||
- Heartbeats run in each agent’s **main session** (`agent:<id>:<mainKey>`), or `global`
|
||||
when `session.scope = "global"`.
|
||||
- Heartbeats run in the agent’s main session by default (`agent:<id>:<mainKey>`),
|
||||
or `global` when `session.scope = "global"`. Set `session` to override to a
|
||||
specific channel session (Discord/WhatsApp/etc.).
|
||||
- `session` only affects the run context; delivery is controlled by `target` and `to`.
|
||||
- To deliver to a specific channel/recipient, set `target` + `to`. With
|
||||
`target: "last"`, delivery uses the last external channel for that session.
|
||||
- If the main queue is busy, the heartbeat is skipped and retried later.
|
||||
- If `target` resolves to no external destination, the run still happens but no
|
||||
outbound message is sent.
|
||||
|
||||
Reference in New Issue
Block a user