chati-dev 3.0.4 → 3.0.6

package/framework/orchestrator/chati.md

@@ -70,7 +70,17 @@ Before determining state, check if the user passed a subcommand:
 -> First run. Go to Step 4 (New Project Setup)
 ```
 
-**If session.yaml has active project:**
+**If session.yaml has project BUT current_agent is empty AND all agents are "pending":**
+```
+-> Fresh install (installer created session.yaml but no agent has run yet).
+-> Skip directly to Step 4d (Route to First Agent).
+-> Use project.type from session.yaml (already set by installer).
+-> Use language from session.yaml (already set by installer).
+-> Do NOT ask project type or language again — installer already collected these.
+-> Do NOT present options — immediately activate the first agent.
+```
+
+**If session.yaml has active project (current_agent is set OR at least one agent is not "pending"):**
 ```
 -> Resume session. Go to Step 5 (Session Resume)
 ```
@@ -155,9 +165,18 @@ user_level_confidence: 0.0
 ```
 
 #### 3d. Route to First Agent
+
+This step is reached either from Step 4 (full setup) OR directly from Step 3 (fresh install).
+In fresh install case, project.type and language are already in session.yaml — use them directly.
+
 ```
-If greenfield -> Read chati.dev/agents/discover/greenfield-wu.md -> Activate
-If brownfield -> Read chati.dev/agents/discover/brownfield-wu.md -> Activate
+1. Update session.yaml: current_agent = greenfield-wu | brownfield-wu (based on project.type)
+2. Activate Session Lock (see Session Lock Protocol)
+3. If greenfield -> Read chati.dev/agents/discover/greenfield-wu.md -> Activate IMMEDIATELY
+   If brownfield -> Read chati.dev/agents/discover/brownfield-wu.md -> Activate IMMEDIATELY
+4. The agent starts its work right away — no "Continue with X?" prompt needed
+   For greenfield-wu: begin asking the user about their project vision
+   For brownfield-wu: begin analyzing the existing codebase
 ```
 
 ### Step 5: Session Resume

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "chati-dev",
-  "version": "3.0.4",
+  "version": "3.0.6",
   "description": "AI-Powered Multi-Agent Orchestration System — Structured vibe coding for Full Stack Development",
   "type": "module",
   "bin": {

package/src/installer/templates.js

@@ -195,11 +195,27 @@ When the orchestrator references Claude-specific files, use these equivalents:
 
 **NEVER create or reference CLAUDE.md, CLAUDE.local.md, or .claude/ directories.**
 
+## CRITICAL — Model Name Mapping
+
+The orchestrator and agents reference Claude model names. Replace them with Codex equivalents:
+
+| Orchestrator says | You use instead |
+|-------------------|-----------------|
+| \`opus\` | \`codex\` |
+| \`sonnet\` | \`mini\` |
+| \`haiku\` | \`mini\` |
+| \`/model opus\` | \`/model codex\` |
+| \`/model sonnet\` | \`/model mini\` |
+| \`/model haiku\` | \`/model mini\` |
+
+When the orchestrator displays a model recommendation (e.g. "Model recommendation: haiku"),
+translate the model name using the table above before showing it to the user.
+
 ## Load
 
 Read and execute the full orchestrator at \`chati.dev/orchestrator/chati.md\`.
 
-Apply the Provider Context Mapping above when following the orchestrator instructions.
+Apply the Provider Context Mapping AND Model Name Mapping above when following the orchestrator instructions.
 
 **Context to pass:**
 - \`.chati/session.yaml\` (session state — includes language)
@@ -234,8 +250,18 @@ When the orchestrator references Claude-specific files, use these equivalents:
 - .claude/commands/ -> .gemini/commands/
 NEVER create or reference CLAUDE.md, CLAUDE.local.md, or .claude/ directories.
 
+CRITICAL — Model Name Mapping:
+The orchestrator and agents reference Claude model names. Replace them with Gemini equivalents:
+- opus -> pro
+- sonnet -> flash
+- haiku -> flash
+- /model opus -> /model pro
+- /model sonnet -> /model flash
+- /model haiku -> /model flash
+When displaying model recommendations, translate names using the mapping above.
+
 Read and execute the full orchestrator at \`chati.dev/orchestrator/chati.md\`.
-Apply the Provider Context Mapping above when following instructions.
+Apply the Provider Context Mapping AND Model Name Mapping above when following instructions.
 
 Context to load:
 - \`.chati/session.yaml\` (session state — includes language)
@@ -288,11 +314,27 @@ When the orchestrator references Claude-specific files, use these equivalents:
 
 **NEVER create or reference CLAUDE.md, CLAUDE.local.md, or .claude/ directories.**
 
+## CRITICAL — Model Name Mapping
+
+The orchestrator and agents reference Claude model names. Replace them with Copilot equivalents:
+
+| Orchestrator says | You use instead |
+|-------------------|-----------------|
+| \`opus\` | \`claude-sonnet\` |
+| \`sonnet\` | \`claude-sonnet\` |
+| \`haiku\` | \`gpt-5\` |
+| \`/model opus\` | \`/model claude-sonnet\` |
+| \`/model sonnet\` | \`/model claude-sonnet\` |
+| \`/model haiku\` | \`/model gpt-5\` |
+
+When the orchestrator displays a model recommendation (e.g. "Model recommendation: haiku"),
+translate the model name using the table above before showing it to the user.
+
 ## Load
 
 Read and execute the full orchestrator at \`chati.dev/orchestrator/chati.md\`.
 
-Apply the Provider Context Mapping above when following the orchestrator instructions.
+Apply the Provider Context Mapping AND Model Name Mapping above when following the orchestrator instructions.
 
 **Context to pass:**
 - \`.chati/session.yaml\` (session state — includes language)