@margay/ccl-core 1.3.5-rc.7 → 1.3.5-rc.8
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 +14 -0
- package/dist/cli.js +1249 -590
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -178,6 +178,20 @@ ccl -p "summarize this diff" --model sonnet
|
|
|
178
178
|
ccl -p "generate a migration plan" --model deepseek-v4-pro --effort high
|
|
179
179
|
```
|
|
180
180
|
|
|
181
|
+
Capability pools use explicit `cost-first ranking` and `quality-first ranking`
|
|
182
|
+
policies. For `visual-ui-analysis`, the current cost-first ranking is Qwen 3.8
|
|
183
|
+
Flash → DeepSeek V4 Flash Vision Exp → Grok 4.5; the quality-first ranking is
|
|
184
|
+
Qwen 3.8 Flash → Grok 4.5 → K3 / Moonshot Platform. DeepSeek Vision remains
|
|
185
|
+
experimental. The hidden K3 visual route is reported as canonical model
|
|
186
|
+
`kimi-platform/kimi-k3`.
|
|
187
|
+
|
|
188
|
+
Anthropic Messages remains the transport for the main Agent, planning, editing,
|
|
189
|
+
tool calling, Advisor, Critic, and general subagent loops. Native OpenAI Chat
|
|
190
|
+
Completions support is deliberately bounded to zero-tool, one-shot visual
|
|
191
|
+
analysis and context compaction. These consumers select and pin one configured
|
|
192
|
+
model before the request; they do not convert protocols, vote on normal results,
|
|
193
|
+
or automatically escalate to another model after a completed answer.
|
|
194
|
+
|
|
181
195
|
## Workflows
|
|
182
196
|
|
|
183
197
|
The `workflow` command and `/workflows` slash command run multi-step task
|