ur-agent 1.57.5 → 1.58.1
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 +36 -0
- package/dist/cli.js +724 -516
- package/docs/USAGE.md +8 -0
- package/docs/providers.md +8 -1
- package/documentation/index.html +1 -1
- package/extensions/jetbrains-ur/build.gradle.kts +1 -1
- package/extensions/vscode-ur-inline-diffs/package.json +1 -1
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,41 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## 1.58.1
|
|
4
|
+
|
|
5
|
+
- Unified vision-capability detection behind
|
|
6
|
+
`src/utils/model/visionCapability.ts`. Three implementations disagreed: the
|
|
7
|
+
Ollama adapter's `modelCapabilityEnabled` returned `has(x) ?? true`, so a
|
|
8
|
+
model advertising nothing was assumed capable; `ur model-doctor` matched
|
|
9
|
+
names privately; the router read a precomputed flag. The binary shape was the
|
|
10
|
+
defect — absence of evidence was reported as evidence, in opposite directions.
|
|
11
|
+
- Vision support is now tri-state. A capability list is authoritative both ways;
|
|
12
|
+
a recognised name can confirm support but never rule it out; anything else is
|
|
13
|
+
`unknown`. Images are withheld only on a confirmed no, so servers without a
|
|
14
|
+
capabilities endpoint keep working, and the note distinguishes "this model
|
|
15
|
+
cannot see" from "support could not be confirmed" — advice that had been
|
|
16
|
+
backwards for models like `kimi-k2.7-code:cloud`.
|
|
17
|
+
|
|
18
|
+
- Fixed choice menus where all three fields said the same thing. Neither the
|
|
19
|
+
schema nor the tool prompt stated that `header`, `label` and `description`
|
|
20
|
+
must carry different information, so the header restated the question and the
|
|
21
|
+
description paraphrased the label — leaving the one field with room to be
|
|
22
|
+
informative saying nothing. Each field now has a defined job, and the prompt
|
|
23
|
+
carries a contrasted bad/good example rather than an abstract instruction.
|
|
24
|
+
|
|
25
|
+
## 1.58.0
|
|
26
|
+
|
|
27
|
+
- Added per-agent cost and token attribution: `ur agent-inspect --costs`.
|
|
28
|
+
`stats.ts` already read every `{sessionId}/subagents/agent-{agentId}.jsonl`
|
|
29
|
+
transcript, but only to fold those tokens into a single total, so a fan-out
|
|
30
|
+
that consumed most of a session's budget was indistinguishable from one that
|
|
31
|
+
did not. Rows are attributed by filename rather than by joining turns back to
|
|
32
|
+
the spawning `tool_use`, because the Agent tool's input carries no agent id
|
|
33
|
+
and any such join would be a guess.
|
|
34
|
+
- The breakdown reports tokens and omits the money column on a local runtime.
|
|
35
|
+
`calculateUSDCost` returns 0 for Ollama, so a cost column would have rendered
|
|
36
|
+
as a wall of `$0.00` and read as a broken feature; cost appears only when a
|
|
37
|
+
provider actually billed.
|
|
38
|
+
|
|
3
39
|
## 1.57.5
|
|
4
40
|
|
|
5
41
|
- Fixed the release gate failing on `repoEditImports`. The gate runs
|