@kodrunhq/opencode-autopilot 1.1.1 → 1.1.2

package/assets/commands/oc-configure.md

@@ -5,41 +5,110 @@ agent: autopilot
 Help the user configure opencode-autopilot by walking through the model
 assignment process interactively.
 
-Start by calling the oc_configure tool with subcommand "start". This returns:
-- Available models grouped by provider (from the user's OpenCode config)
-- The 8 agent groups with descriptions and recommendations
-- Current assignments if reconfiguring
-- Adversarial diversity rules
+## Step 1: Discover available models
 
-Then walk through each of the 8 agent groups in order (architects first,
-utilities last), explaining for each:
+Call the oc_configure tool with subcommand "start". This returns:
+- `availableModels`: a map of provider -> list of "provider/model" strings
+- `groups`: the 8 agent groups with descriptions and recommendations
+- `currentConfig`: existing assignments if reconfiguring
+- `diversityRules`: adversarial diversity constraints
 
-1. What the group does and which agents belong to it
-2. The model tier recommendation
+If `availableModels` is empty or has no entries, tell the user:
+"No models were discovered from your providers. Run `opencode models`
+in your terminal to see available models, then type them manually below."
+
+## Step 2: Build the complete model list
+
+Combine ALL models from ALL providers into a single numbered list.
+Every model the user has access to must appear. Do NOT filter, summarize,
+or show only "recommended" models. The user decides — you present options.
+
+Example (show ALL of them, not a subset). The IDs must match exactly
+what `availableModels` returns (provider prefix comes from the provider):
+```
+Available models:
+ 1. anthropic/claude-opus-4-6
+ 2. anthropic/claude-sonnet-4-6
+ 3. anthropic/claude-haiku-4-5
+ 4. openai/gpt-5.4
+ 5. openai/gpt-5.4-mini
+ 6. openai/gpt-5.4-codex
+ 7. google/gemini-3.1-pro
+ 8. google/gemini-3-flash
+ ...
+```
+
+## Step 3: Walk through each group
+
+For each of the 8 groups (architects first, utilities last):
+
+1. Explain what the group does and which agents belong to it
+2. Show the tier recommendation
 3. For adversarial groups (challengers, reviewers, red-team): explain WHY
-   model diversity matters and which group they're adversarial to
-4. List available models from the user's providers
+   model diversity matters and which group they are adversarial to
+4. Show the full numbered model list again whenever asking for selections
+
+### Collecting models for each group
+
+For each group, collect an ORDERED LIST of models (not just one):
 
-For each group, present the available models as a numbered list:
+```
+Group: Architects
+Pick models in priority order. The first is the primary; the rest are
+fallbacks tried in sequence when the primary is rate-limited or fails.
 
-Example format:
-  Primary model for Architects:
-  1. anthropic/claude-opus-4-6
-  2. openai/gpt-5.4
-  3. google/gemini-3.1-pro
+Enter model numbers separated by commas (e.g. 1,4,7):
+```
 
-  Pick a number (or type a model ID):
+- The FIRST number is the primary model
+- All subsequent numbers are fallbacks, tried in the order given
+- Minimum 1 model (the primary), recommend 2-3 total
+- Emphasize that fallbacks are the core feature: "When your primary model
+  hits a rate limit, the plugin automatically retries with the next model
+  in your fallback chain. More fallbacks = more resilience."
 
-If the user sends just a number, map it to the corresponding model.
-Then ask for optional fallbacks the same way (1-3 fallback models).
-Call oc_configure with subcommand "assign" for each group.
+Parse the user's response:
+- If they send numbers like "1,4,7": map to model IDs
+- If they send model IDs directly: use as-is
+- If they send a single number: that's the primary with no fallbacks
 
-If the assign response contains diversityWarnings, explain them
+Call oc_configure with subcommand "assign":
+- `group`: the group ID
+- `primary`: first model from the user's list
+- `fallbacks`: remaining models as comma-separated string
+
+### Diversity warnings
+
+If the assign response contains `diversityWarnings`, explain them
 conversationally. Strong warnings should be highlighted — the user can
 still proceed, but make the quality trade-off clear.
 
-After all 8 groups are assigned, call oc_configure with subcommand "commit"
-to persist the configuration.
+Example: "Heads up: Architects and Challengers both use Claude models.
+Challengers are supposed to critique Architect decisions — using the same
+model family means you get confirmation bias instead of genuine challenge.
+Consider picking a different family for one of them. Continue anyway?"
+
+## Step 4: Commit and verify
+
+After all 8 groups are assigned, call oc_configure with subcommand "commit".
+
+Then call oc_configure with subcommand "doctor" to verify health.
+
+Show a final summary table:
+
+```
+Group          | Primary                      | Fallbacks
+───────────────┼──────────────────────────────┼──────────────────────────
+Architects     | anthropic/claude-opus-4-6    | openai/gpt-5.4
+Challengers    | openai/gpt-5.4              | google/gemini-3.1-pro
+...
+```
+
+## Rules
 
-End by showing the final summary table and running the "doctor" subcommand
-to verify everything is healthy.
+- NEVER pre-select models for the user. Always present the full list.
+- NEVER skip fallback collection. Always ask for ordered model lists.
+- NEVER filter the model list to "recommended" models. Show everything.
+- If the user says "pick for me" or "use defaults", THEN you may suggest
+  assignments based on the tier recommendations and diversity rules, but
+  still show what you picked and ask for confirmation.

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@kodrunhq/opencode-autopilot",
-	"version": "1.1.1",
+	"version": "1.1.2",
 	"description": "Curated agents, skills, and commands for the OpenCode AI coding CLI — autonomous orchestrator, multi-agent code review, model fallback, and in-session asset creation tools.",
 	"main": "src/index.ts",
 	"keywords": [