@getmarrow/sdk 3.7.12 → 3.7.13
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +4 -70
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -58,7 +58,9 @@ Four measured deltas: `attempts_per_success`, `time_to_success_seconds`, `drift_
|
|
|
58
58
|
|
|
59
59
|
---
|
|
60
60
|
|
|
61
|
-
## What's New in v3.7.
|
|
61
|
+
## What's New in v3.7.13
|
|
62
|
+
|
|
63
|
+
v3.7.13 is a README-only patch release for npmjs. It keeps the package page focused on the current install surface and sends full feature history, examples, and API reference to [getmarrow.ai/docs](https://getmarrow.ai/docs/).
|
|
62
64
|
|
|
63
65
|
### Decision Brief: One Pre-Action Call
|
|
64
66
|
|
|
@@ -82,75 +84,7 @@ Use this before deploys, publishes, merges, audits, patches, secret changes, or
|
|
|
82
84
|
|
|
83
85
|
`decisionBrief()` does not replace the Marrow loop. Agents should still log intent with `think()` or passive auto-logging before acting, then call `commit()` after verification so the outcome teaches the next run.
|
|
84
86
|
|
|
85
|
-
|
|
86
|
-
|
|
87
|
-
Agent Status: Prove Marrow Is Working
|
|
88
|
-
|
|
89
|
-
Agents can now call `agentStatus()` to verify Marrow is active without sending a user to a dashboard. It returns connection state, signal quality, non-sensitive proof, and next actions.
|
|
90
|
-
|
|
91
|
-
```typescript
|
|
92
|
-
const status = await marrow.agentStatus('7d');
|
|
93
|
-
|
|
94
|
-
if (!status.active) {
|
|
95
|
-
console.log(status.next_actions[0]);
|
|
96
|
-
}
|
|
97
|
-
|
|
98
|
-
if (status.quality.enough_signal) {
|
|
99
|
-
console.log(status.summary);
|
|
100
|
-
}
|
|
101
|
-
```
|
|
102
|
-
|
|
103
|
-
No raw action, context, or outcome text is returned. The response is aggregated proof only.
|
|
104
|
-
|
|
105
|
-
### Previous: v3.7.8 — Auto-Logging + Docs
|
|
106
|
-
|
|
107
|
-
Auto-logging section added to README. Docs link added at bottom. README now mirrors MCP structure with consistent layout.
|
|
108
|
-
|
|
109
|
-
### Previous: v3.7.5 — Multi-API-Key Management
|
|
110
|
-
|
|
111
|
-
Manage API keys for your entire fleet of agents directly from your code:
|
|
112
|
-
|
|
113
|
-
- `createApiKey()` — Create named, scoped API keys (`live`/`test`)
|
|
114
|
-
- `listApiKeys()` — List all keys with masked display
|
|
115
|
-
- `getApiKey()` — Get key details and usage stats
|
|
116
|
-
- `revokeApiKey()` — Permanently revoke a key
|
|
117
|
-
- `rotateApiKey()` — Atomically rotate (revoke + create new)
|
|
118
|
-
- `getKeyAudit()` — Paginated audit log for key operations
|
|
119
|
-
|
|
120
|
-
```typescript
|
|
121
|
-
const key = await marrow.createApiKey({
|
|
122
|
-
name: 'Production Agent',
|
|
123
|
-
key_type: 'live',
|
|
124
|
-
scopes: ['decisions:write', 'memories:read']
|
|
125
|
-
});
|
|
126
|
-
// { id, key: 'mrw_live_...', name, scopes, ... }
|
|
127
|
-
// ⚠️ Full key shown once — store it securely
|
|
128
|
-
|
|
129
|
-
const keys = await marrow.listApiKeys();
|
|
130
|
-
// [{ name, key: 'mrw_live_abc1...d4ef', key_type, status, ... }]
|
|
131
|
-
```
|
|
132
|
-
|
|
133
|
-
---
|
|
134
|
-
|
|
135
|
-
### Previous: v3.6.0 — Agent-Narrated Marrow Contribution
|
|
136
|
-
|
|
137
|
-
Marrow now tells the agent exactly what it contributed to each decision, so the agent can surface that contribution to the user in plain English — no dashboard required.
|
|
138
|
-
|
|
139
|
-
- `think()` returns `marrow_contributed` describing intelligence Marrow surfaced (warnings consulted, hive patterns, similar decisions, workflow templates, loop detection, collective insight).
|
|
140
|
-
- `commit()` returns `marrow_contributed` with concrete signals (pattern reused, warning avoided, workflow step).
|
|
141
|
-
- `endSession()` returns `session_summary` aggregating Marrow's contribution across the session, with a one-line `narrative` for the agent to surface.
|
|
142
|
-
|
|
143
|
-
Each object has `has_signal: boolean` — agent narrates when true, stays quiet when false. The MCP `marrow-always-on` system prompt instructs agents on tone and timing.
|
|
144
|
-
|
|
145
|
-
```typescript
|
|
146
|
-
const result = await marrow.think({ action: 'deploy auth refactor', type: 'implementation' });
|
|
147
|
-
if (result.marrow_contributed?.has_signal) {
|
|
148
|
-
// Agent can mention: "Pulling 12 similar patterns from the hive..."
|
|
149
|
-
// Or: "Marrow flagged this approach failed 4× last week..."
|
|
150
|
-
}
|
|
151
|
-
```
|
|
152
|
-
|
|
153
|
-
Counts and booleans only — no decision content, no PII echoed back.
|
|
87
|
+
Full feature history, examples, and API reference live at [getmarrow.ai/docs](https://getmarrow.ai/docs/).
|
|
154
88
|
|
|
155
89
|
---
|
|
156
90
|
|