feat: implement CLI model tier configuration

- Introduce model tiers (small, medium, large) for CLI tools
- Update agent.py to resolve model names based on tier
- Modify chore_implement.py and prompt.py to accept tier options
- Enhance README.md and overview.md with model tier documentation

Author: MarkusNeusinger

.env.example

@@ -41,3 +41,24 @@ PORT=8000
 
 # Anthropic API key for Claude
 # ANTHROPIC_API_KEY=sk-ant-...
+
+# ============================================================================
+# CLI Model Tier Configuration (optional)
+# ============================================================================
+# Override the default model for each tier per CLI tool.
+# Tiers: small (fast/cheap), medium (balanced), large (complex)
+
+# Claude Code models
+# CLI_MODEL_CLAUDE_SMALL=haiku
+# CLI_MODEL_CLAUDE_MEDIUM=sonnet
+# CLI_MODEL_CLAUDE_LARGE=opus
+
+# GitHub Copilot models
+# CLI_MODEL_COPILOT_SMALL=gpt-4o-mini
+# CLI_MODEL_COPILOT_MEDIUM=gpt-4o
+# CLI_MODEL_COPILOT_LARGE=o1
+
+# Gemini CLI models
+# CLI_MODEL_GEMINI_SMALL=gemini-2.0-flash
+# CLI_MODEL_GEMINI_MEDIUM=gemini-2.0-flash-thinking
+# CLI_MODEL_GEMINI_LARGE=gemini-2.5-pro

agentic/README.md

@@ -21,7 +21,29 @@ agentic/
 uv run agentic/workflows/chore_implement.py "your task description"
 
 # Options
---model [haiku|sonnet|opus]    # Default: opus
+--model [small|medium|large]   # Default: large
 --cli [claude|copilot|gemini]  # Default: claude
 --working-dir PATH             # Default: current directory
+```
+
+## Model Tiers
+
+Model tiers abstract away CLI-specific model names, allowing the same command to work across different AI tools.
+
+| Tier | Purpose | Claude | Copilot | Gemini |
+|------|---------|--------|---------|--------|
+| small | Fast/cheap tasks | haiku | gpt-4o-mini | gemini-2.0-flash |
+| medium | Balanced tasks | sonnet | gpt-4o | gemini-2.0-flash-thinking |
+| large | Complex tasks | opus | o1 | gemini-2.5-pro |
+
+### Override Mappings
+
+Override the default model for any tier via environment variables:
+
+```bash
+# Use a specific Claude model for the "large" tier
+CLI_MODEL_CLAUDE_LARGE=claude-3-5-sonnet-20240620
+
+# Use a different Copilot model for the "medium" tier
+CLI_MODEL_COPILOT_MEDIUM=gpt-4-turbo
 ```

agentic/workflows/chore_implement.py

@@ -3,6 +3,7 @@
 # requires-python = ">=3.10"
 # dependencies = [
 #   "pydantic",
+#   "pydantic-settings",
 #   "python-dotenv",
 #   "click",
 #   "rich",
@@ -24,7 +25,7 @@
 
 Examples:
     # Run with specific model
-    ./agentic/workflows/chore_implement.py "Add logging to agent.py" --model sonnet
+    ./agentic/workflows/chore_implement.py "Add logging to agent.py" --model medium
 
     # Run from a different working directory
     ./agentic/workflows/chore_implement.py "Update documentation" --working-dir /path/to/project
@@ -130,9 +131,9 @@ def extract_plan_path(output: str) -> str:
 @click.argument("prompt", required=True)
 @click.option(
     "--model",
-    type=click.Choice(["haiku", "sonnet", "opus"]),
-    default="opus",
-    help="Claude model to use",
+    type=click.Choice(["small", "medium", "large"]),
+    default="large",
+    help="Model tier (maps to CLI-specific models)",
 )
 @click.option(
     "--working-dir",

agentic/workflows/modules/agent.py

@@ -13,6 +13,20 @@
 from pydantic import BaseModel
 from dotenv import load_dotenv
 
+# Add project root to path for imports
+_project_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))))
+if _project_root not in sys.path:
+    sys.path.insert(0, _project_root)
+
+# Import settings directly from the module file to avoid core/__init__.py
+# which has dependencies (PIL) not available in uv script environments
+import importlib.util
+_config_path = os.path.join(_project_root, "core", "config.py")
+_spec = importlib.util.spec_from_file_location("core_config", _config_path)
+_core_config = importlib.util.module_from_spec(_spec)
+_spec.loader.exec_module(_core_config)
+settings = _core_config.settings
+
 
 # Retry codes for Claude Code execution errors
 class RetryCode(str, Enum):
@@ -29,7 +43,7 @@ class AgentPromptRequest(BaseModel):
     prompt: str
     run_id: str
     agent_name: str = "ops"
-    model: Literal["haiku", "sonnet", "opus"] = "sonnet"
+    model: Literal["small", "medium", "large"] = "medium"
     cli: Literal["claude", "copilot", "gemini"] = "claude"
     dangerously_skip_permissions: bool = False
     output_file: str
@@ -50,7 +64,7 @@ class AgentTemplateRequest(BaseModel):
     slash_command: str
     args: List[str]
     run_id: str
-    model: Literal["haiku", "sonnet", "opus"] = "sonnet"
+    model: Literal["small", "medium", "large"] = "medium"
     cli: Literal["claude", "copilot", "gemini"] = "claude"
     working_dir: Optional[str] = None
 
@@ -485,12 +499,15 @@ def prompt_claude_code(request: AgentPromptRequest) -> AgentPromptResponse:
         if os.path.exists(potential_mcp_path):
             mcp_config_path = potential_mcp_path
 
+    # Resolve model tier to actual model name for the CLI
+    actual_model = settings.resolve_model(request.cli, request.model)
+
     # Build CLI-specific command
     cmd = build_cli_command(
         cli=request.cli,
         cli_path=cli_path,
         prompt=request.prompt,
-        model=request.model,
+        model=actual_model,
         dangerously_skip_permissions=request.dangerously_skip_permissions,
         mcp_config_path=mcp_config_path,
     )
@@ -707,7 +724,7 @@ def execute_template(request: AgentTemplateRequest) -> AgentPromptResponse:
             slash_command="/implement",
             args=["plan.md"],
             run_id="abc12345",
-            model="sonnet"  # Explicitly set model
+            model="medium"  # Model tier (small/medium/large)
         )
         response = execute_template(request)
     """

agentic/workflows/prompt.py

@@ -3,6 +3,7 @@
 # requires-python = ">=3.10"
 # dependencies = [
 #   "pydantic",
+#   "pydantic-settings",
 #   "python-dotenv",
 #   "click",
 #   "rich",
@@ -23,7 +24,7 @@
 
 Examples:
     # Run with specific model
-    ./adw_prompt.py "Explain this code" --model opus
+    ./adw_prompt.py "Explain this code" --model large
 
     # Run with custom output file
     ./adw_prompt.py "Create a FastAPI app" --output my_result.jsonl
@@ -71,9 +72,9 @@
 @click.argument("prompt", required=True)
 @click.option(
     "--model",
-    type=click.Choice(["haiku", "sonnet", "opus"]),
-    default="sonnet",
-    help="Claude model to use",
+    type=click.Choice(["small", "medium", "large"]),
+    default="medium",
+    help="Model tier (maps to CLI-specific models)",
 )
 @click.option(
     "--output",

core/config.py

@@ -97,6 +97,39 @@ class Settings(BaseSettings):
     claude_review_max_tokens: int = 2000
     """Maximum tokens for Claude review responses"""
 
+    # =============================================================================
+    # CLI MODEL TIER CONFIGURATION
+    # =============================================================================
+    # Model tiers (small/medium/large) map to CLI-specific model names.
+    # Override these to use different models for each tier.
+
+    cli_model_claude_small: str = "haiku"
+    """Claude model for 'small' tier (fast/cheap tasks)"""
+
+    cli_model_claude_medium: str = "sonnet"
+    """Claude model for 'medium' tier (balanced tasks)"""
+
+    cli_model_claude_large: str = "opus"
+    """Claude model for 'large' tier (complex tasks)"""
+
+    cli_model_copilot_small: str = "gpt-4o-mini"
+    """Copilot model for 'small' tier"""
+
+    cli_model_copilot_medium: str = "gpt-4o"
+    """Copilot model for 'medium' tier"""
+
+    cli_model_copilot_large: str = "o1"
+    """Copilot model for 'large' tier"""
+
+    cli_model_gemini_small: str = "gemini-2.0-flash"
+    """Gemini model for 'small' tier"""
+
+    cli_model_gemini_medium: str = "gemini-2.0-flash-thinking"
+    """Gemini model for 'medium' tier"""
+
+    cli_model_gemini_large: str = "gemini-2.5-pro"
+    """Gemini model for 'large' tier"""
+
     # =============================================================================
     # CACHE
     # =============================================================================
@@ -154,6 +187,36 @@ def has_google_cloud(self) -> bool:
         """Check if Google Cloud is configured."""
         return bool(self.google_application_credentials or self.instance_connection_name)
 
+    def resolve_model(self, cli: str, tier: str) -> str:
+        """Resolve a model tier (small/medium/large) to the actual model name for a CLI.
+
+        Args:
+            cli: The CLI tool name ("claude", "copilot", or "gemini")
+            tier: The model tier ("small", "medium", or "large")
+
+        Returns:
+            The actual model name for the specified CLI and tier.
+            If the tier is not recognized, returns the tier unchanged (pass-through).
+        """
+        mapping = {
+            "claude": {
+                "small": self.cli_model_claude_small,
+                "medium": self.cli_model_claude_medium,
+                "large": self.cli_model_claude_large,
+            },
+            "copilot": {
+                "small": self.cli_model_copilot_small,
+                "medium": self.cli_model_copilot_medium,
+                "large": self.cli_model_copilot_large,
+            },
+            "gemini": {
+                "small": self.cli_model_gemini_small,
+                "medium": self.cli_model_gemini_medium,
+                "large": self.cli_model_gemini_large,
+            },
+        }
+        return mapping.get(cli, {}).get(tier, tier)
+
 
 # Global settings instance
 # Import this from other modules: from core.config import settings

docs/workflows/overview.md

@@ -182,3 +182,41 @@ gh workflow run bulk-generate.yml -f specification_id=all -f library=matplotlib
 ```
 
 **Concurrency limit**: Max 3 parallel implementations globally.
+
+---
+
+## CLI Model Tiers
+
+The agentic workflows use abstract model tiers (`small`, `medium`, `large`) instead of CLI-specific model names. This allows the same command to work across different AI tools.
+
+| Tier | Purpose | Claude | Copilot | Gemini |
+|------|---------|--------|---------|--------|
+| small | Fast/cheap tasks | haiku | gpt-4o-mini | gemini-2.0-flash |
+| medium | Balanced tasks | sonnet | gpt-4o | gemini-2.0-flash-thinking |
+| large | Complex tasks | opus | o1 | gemini-2.5-pro |
+
+### Usage
+
+```bash
+# Use large tier (default for chore_implement)
+uv run agentic/workflows/chore_implement.py "Add feature X" --model large
+
+# Use medium tier with Copilot
+uv run agentic/workflows/prompt.py "Quick fix" --model medium --cli copilot
+```
+
+### Override Mappings
+
+Override the default model for any tier via environment variables:
+
+```bash
+# Linux/WSL
+export CLI_MODEL_CLAUDE_LARGE=claude-3-5-sonnet-20240620
+export CLI_MODEL_COPILOT_MEDIUM=gpt-4-turbo
+
+# Windows PowerShell
+$env:CLI_MODEL_CLAUDE_LARGE = "claude-3-5-sonnet-20240620"
+$env:CLI_MODEL_COPILOT_MEDIUM = "gpt-4-turbo"
+```
+
+All environment variable names follow the pattern: `CLI_MODEL_{CLI}_{TIER}`