ur-agent 1.61.1 → 1.62.0

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/CHANGELOG.md CHANGED
@@ -1,5 +1,29 @@
1
1
  # Changelog
2
2
 
3
+ ## 1.62.0
4
+
5
+ - `ur agent-inspect --costs` now labels each row with what the agent was
6
+ actually doing. A real 62-agent fan-out reported opaque hex ids, so you could
7
+ see that one agent burned 810k input tokens — 14% of the session — without
8
+ being able to tell which of your subagents it was. The description already
9
+ existed in the `agent-{id}.meta.json` sidecar and was simply never read.
10
+ - Agents with missing or malformed metadata still appear, keyed by id.
11
+ Historical sessions predate the description field, and dropping those rows
12
+ would lose spend in order to avoid losing a label.
13
+
14
+ ## 1.61.2
15
+
16
+ - Fixed `ur agent-inspect --costs` reporting nothing, always. It resolved the
17
+ *live* session, but every `ur` invocation mints a new session id, so bare it
18
+ pointed at a session created milliseconds earlier that had by definition
19
+ spawned nothing — two consecutive runs in the same directory produced two
20
+ different empty session ids and no data either time. It now prefers the live
21
+ session when it has transcripts (the `/agent-inspect` case, inside a session
22
+ that already fanned out) and otherwise falls back to the most recent session
23
+ in the project that does.
24
+ - The transcript writer was never at fault: Agent-tool subagents do write
25
+ `agent-*.jsonl`. Only the reader's choice of session was wrong.
26
+
3
27
  ## 1.61.1
4
28
 
5
29
  - Fixed secondary model queries failing on any Ollama setup whose session model