vinta-ai-workflows 0.7.0-alpha6 → 0.7.0-alpha7
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 +33 -0
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -7,6 +7,7 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
|
|
|
7
7
|
|
|
8
8
|
## [0.7.0] — YYYY-MM-DD
|
|
9
9
|
|
|
10
|
+
<!-- pre-release: 0.7.0-alpha7 on 2026-09-14 -->
|
|
10
11
|
<!-- pre-release: 0.7.0-alpha6 on 2026-09-13 -->
|
|
11
12
|
<!-- pre-release: 0.7.0-alpha5 on 2026-09-13 -->
|
|
12
13
|
<!-- pre-release: 0.7.0-alpha4 on 2026-09-13 -->
|
|
@@ -16,6 +17,38 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
|
|
|
16
17
|
|
|
17
18
|
### Added
|
|
18
19
|
|
|
20
|
+
- **A run has a spokesperson you can ask.** `serve` grows a **Monitor** panel on
|
|
21
|
+
the run view: one agent that reads the journal and answers in prose — what is
|
|
22
|
+
blocked, why a phase failed, what it would take to move on. It runs on the
|
|
23
|
+
dearest tier on the crew, because it is read by a person deciding what to do
|
|
24
|
+
about a failing run, and it is asked a handful of times rather than hundreds.
|
|
25
|
+
|
|
26
|
+
Three things it deliberately is not.
|
|
27
|
+
|
|
28
|
+
**It is not in the permission path.** The obvious thought, once an agent has
|
|
29
|
+
been refused something, is to put a smarter agent in front of the refusals —
|
|
30
|
+
but a phase makes hundreds of tool calls (one node in one real run made 179)
|
|
31
|
+
and a model turn before each is latency and cost spent on questions like "may
|
|
32
|
+
I run ruff", which should never have been questions. Permissions are settled
|
|
33
|
+
structurally, once per spawn, for free.
|
|
34
|
+
|
|
35
|
+
**It does not read the transcripts.** Piping every agent's output into a
|
|
36
|
+
second agent would make the monitor the most expensive thing in the run and
|
|
37
|
+
tie its cost to the work rather than to the questions. It reads a bounded
|
|
38
|
+
digest — statuses, failure reasons, pending questions, the last few refusals
|
|
39
|
+
with their sentences, one trimmed last word per phase — so a one-hour run and
|
|
40
|
+
a one-day run cost about the same to ask about.
|
|
41
|
+
|
|
42
|
+
**It has no authority.** It can explain a §9.1 pause; it cannot answer one.
|
|
43
|
+
An agent that could quietly approve its colleagues' work would turn a
|
|
44
|
+
checkpoint into a formality, and the checkpoint is the point.
|
|
45
|
+
|
|
46
|
+
It answers about **finished runs too**, which is when "why did this fail" is
|
|
47
|
+
usually asked, and it needs no lane: it writes nothing.
|
|
48
|
+
|
|
49
|
+
- **`serve` reads `--permission`.** The flag has been accepted and ignored since
|
|
50
|
+
it existed. It decides how the monitor is spawned, so it is read now.
|
|
51
|
+
|
|
19
52
|
- **A plan staffs a team, instead of picking a model per phase.** Choosing a tier
|
|
20
53
|
phase by phase answers "what runs this one" ten times and never adds it up, so the
|
|
21
54
|
two questions that decide what a feature costs go unasked: how many agents does
|
package/package.json
CHANGED