@compilr-dev/sdk 0.10.5 → 0.10.6

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.
@@ -174,18 +174,58 @@ Skills with issues (short description, empty body, name mismatch) show a ⚠ war
174
174
  // ── Models & Providers ──────────────────────────────────────────────────
175
175
  {
176
176
  id: 'models',
177
- title: 'Models and Providers',
178
- keywords: ['model', 'provider', 'claude', 'openai', 'gemini', 'ollama', 'llm'],
177
+ title: 'Models, Providers, and Model Switching',
178
+ keywords: [
179
+ 'model',
180
+ 'provider',
181
+ 'claude',
182
+ 'openai',
183
+ 'gemini',
184
+ 'ollama',
185
+ 'llm',
186
+ 'switch model',
187
+ 'change model',
188
+ 'hot swap',
189
+ 'tier',
190
+ 'fast',
191
+ 'balanced',
192
+ 'powerful',
193
+ 'settings model',
194
+ 'agent model',
195
+ 'per-agent model',
196
+ ],
179
197
  content: `Compilr Dev supports multiple LLM providers:
180
198
 
181
- - Claude (Anthropic) — Claude Opus, Sonnet, Haiku
199
+ - Claude (Anthropic) — Opus (powerful), Sonnet (balanced), Haiku (fast)
182
200
  - OpenAI — GPT-4o, GPT-4o-mini
183
201
  - Gemini (Google) — Gemini 2.5 Pro, Flash
184
202
  - Ollama — Local models (llama, codellama, etc.)
185
203
  - Together, Groq, Fireworks, Perplexity, OpenRouter
186
204
 
187
205
  Each provider requires an API key (except Ollama which runs locally).
188
- Models are organized by tier: premium, balanced, fast.`,
206
+
207
+ **Model tiers:**
208
+ - **Fast** — quick responses, lower cost (Haiku, GPT-4o-mini, Flash)
209
+ - **Balanced** — best mix of speed and capability, the default (Sonnet, GPT-4o)
210
+ - **Powerful** — best reasoning for complex tasks (Opus)
211
+
212
+ **Settings model vs agent model — important distinction:**
213
+ - The model in Settings is a **template for new agents**. Changing it does NOT affect existing agents. It only determines which model new agents get when you create them.
214
+ - Each agent has its **own model**. You can change an agent's model independently without affecting other agents.
215
+
216
+ **How to change an agent's model:**
217
+ - **Desktop:** Click the model name in the status bar (bottom right) to open the model picker for the active chat's agent. You can also change models from the Agents panel.
218
+ - **CLI:** Use /model to change the active agent's model.
219
+
220
+ **Same provider only (Level 1):** You can switch between models within the same provider (e.g., Claude Sonnet → Claude Opus). Switching providers (e.g., Claude → Gemini) is not yet supported.
221
+
222
+ **When to switch models:**
223
+ - Use a **powerful** model (Opus) for complex architecture decisions, debugging hard problems, or writing nuanced content
224
+ - Use the **balanced** model (Sonnet) for everyday coding, planning, and general tasks
225
+ - Use a **fast** model (Haiku) for quick questions, simple edits, or when cost matters
226
+ - Switch back after the hard task is done — no need to stay on the expensive model
227
+
228
+ **Cost awareness:** switching from Haiku to Opus mid-conversation can significantly increase cost. The model tier is shown in the status bar and agent cards so you always know what you're using.`,
189
229
  },
190
230
  // ── Context & Memory ────────────────────────────────────────────────────
191
231
  {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@compilr-dev/sdk",
3
- "version": "0.10.5",
3
+ "version": "0.10.6",
4
4
  "description": "Universal agent runtime for building AI-powered applications",
5
5
  "type": "module",
6
6
  "main": "dist/index.js",