cctally 1.56.2 → 1.57.1
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 +13 -0
- package/bin/_lib_pricing.py +6 -2
- package/dashboard/static/assets/index-CIMIZmQ2.js +80 -0
- package/dashboard/static/assets/index-D3Xnfr7K.css +1 -0
- package/dashboard/static/dashboard.html +2 -2
- package/package.json +1 -1
- package/dashboard/static/assets/index-CtgGA2MX.js +0 -78
- package/dashboard/static/assets/index-DBy0SKfB.css +0 -1
package/CHANGELOG.md
CHANGED
|
@@ -5,6 +5,19 @@ based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
|
|
|
5
5
|
|
|
6
6
|
## [Unreleased]
|
|
7
7
|
|
|
8
|
+
## [1.57.1] - 2026-06-28
|
|
9
|
+
|
|
10
|
+
### Fixed
|
|
11
|
+
- Conversation viewer: a tool call that follows a to-do/checklist update inside the same assistant turn is no longer silently dropped — when an assistant message updated its task checklist and then made a further tool call (for example a Codex agent call that errored), only the checklist card rendered and every tool call after it vanished entirely; the leading checklist now collapses to its card as before and the trailing tool call(s) render normally right after it (#245).
|
|
12
|
+
|
|
13
|
+
## [1.57.0] - 2026-06-27
|
|
14
|
+
|
|
15
|
+
### Added
|
|
16
|
+
- Conversation viewer: Codex tool calls (`mcp__codex__codex` / `codex-reply`) now render as a dedicated card — the prompt and the response render as Markdown (the response was previously a raw JSON blob), with an OpenAI-green "agent run" header showing model / reasoning effort / sandbox, a clamped response with show-full, a dedicated error block, and reply-thread chips.
|
|
17
|
+
|
|
18
|
+
### Fixed
|
|
19
|
+
- Model chips no longer mislabel or miscolour Fable and unrecognized models. A conversation whose main model was `claude-fable-5` previously rendered a browse-rail chip literally labelled "sonnet" in the sonnet green (the rail chip's text is its class name), and any unrecognized model — including the internal `<synthetic>` placeholder — silently borrowed that same "sonnet" identity across every chip surface. Fable now gets its own dedicated rose chip everywhere (the conversation rail and reader, the Sessions panel, the session/block/projects modals, and the Weekly/Monthly model-cost bars and detail cards), and a genuinely-unrecognized model now gets a neutral grey "other" chip — labelled in the rail by the model's own abbreviation — instead of impersonating sonnet (#244).
|
|
20
|
+
|
|
8
21
|
## [1.56.2] - 2026-06-27
|
|
9
22
|
|
|
10
23
|
### Fixed
|
package/bin/_lib_pricing.py
CHANGED
|
@@ -32,9 +32,11 @@ TIERED_THRESHOLD = 200_000
|
|
|
32
32
|
def _chip_for_model(name: str) -> str:
|
|
33
33
|
"""Bucket a canonical model id into a small chip palette.
|
|
34
34
|
|
|
35
|
-
Returns one of 'opus' | 'sonnet' | 'haiku' | 'other'. Used by the
|
|
35
|
+
Returns one of 'opus' | 'sonnet' | 'haiku' | 'fable' | 'other'. Used by the
|
|
36
36
|
dashboard's Weekly / Monthly panels and modals so per-model
|
|
37
|
-
coloring stays consistent across the UI.
|
|
37
|
+
coloring stays consistent across the UI. #244 — `fable` is a dedicated
|
|
38
|
+
family (Fable is a current first-class model); it must NOT fall through to
|
|
39
|
+
the neutral 'other' bucket, mirroring the frontend `modelChipClass`.
|
|
38
40
|
"""
|
|
39
41
|
n = (name or "").lower()
|
|
40
42
|
if "opus" in n:
|
|
@@ -43,6 +45,8 @@ def _chip_for_model(name: str) -> str:
|
|
|
43
45
|
return "sonnet"
|
|
44
46
|
if "haiku" in n:
|
|
45
47
|
return "haiku"
|
|
48
|
+
if "fable" in n:
|
|
49
|
+
return "fable"
|
|
46
50
|
return "other"
|
|
47
51
|
|
|
48
52
|
|