docs(templates): Add example for session templates

Add example script demonstrating session template usage:
1. Listing existing templates
2. Getting templates by task type
3. Injecting templates into messages
4. Usage tracking

Resolves #329

examples/session_templates_example.py

@@ -0,0 +1,89 @@
+#!/usr/bin/env python3
+"""
+Example: Using session templates for code-first seeding.
+
+This script demonstrates how to use the session template system
+to pre-seed new sessions with successful tool call patterns.
+"""
+
+import sys
+from pathlib import Path
+
+# Add the parent directory to the path
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from tools.session_templates import SessionTemplates, TaskType
+
+
+def main():
+    """Demonstrate session template usage."""
+    
+    # Create template manager
+    templates = SessionTemplates()
+    
+    print("Session Templates Example")
+    print("=" * 50)
+    
+    # List existing templates
+    print("\n1. Existing templates:")
+    template_list = templates.list_templates()
+    if template_list:
+        for t in template_list:
+            print(f"  - {t.name}: {t.task_type.value} ({len(t.examples)} examples)")
+    else:
+        print("  No templates found")
+    
+    # Example: Create a template from a session
+    print("\n2. Creating a template from a session:")
+    print("  (This would normally use a real session ID)")
+    
+    # Example: Get a template for code tasks
+    print("\n3. Getting a template for CODE tasks:")
+    code_template = templates.get_template(TaskType.CODE)
+    if code_template:
+        print(f"  Found template: {code_template.name}")
+        print(f"  Type: {code_template.task_type.value}")
+        print(f"  Examples: {len(code_template.examples)}")
+        
+        # Show first example
+        if code_template.examples:
+            example = code_template.examples[0]
+            print(f"  First example: {example.tool_name}")
+            print(f"    Arguments: {example.arguments}")
+            print(f"    Result preview: {example.result[:100]}...")
+    else:
+        print("  No CODE template found")
+    
+    # Example: Inject template into messages
+    print("\n4. Injecting template into messages:")
+    if code_template:
+        # Create sample messages
+        messages = [
+            {"role": "system", "content": "You are a helpful assistant."},
+            {"role": "user", "content": "Help me write some code"}
+        ]
+        
+        # Inject template
+        updated_messages = templates.inject_into_messages(code_template, messages)
+        
+        print(f"  Original messages: {len(messages)}")
+        print(f"  Updated messages: {len(updated_messages)}")
+        print(f"  Template usage count: {code_template.usage_count}")
+        
+        # Show the injection
+        print("\n  Injected messages:")
+        for i, msg in enumerate(updated_messages[:6]):  # Show first 6
+            role = msg.get('role', 'unknown')
+            content = msg.get('content', '')
+            if content:
+                content_preview = content[:50] + "..." if len(content) > 50 else content
+                print(f"    {i}: {role} - {content_preview}")
+            else:
+                print(f"    {i}: {role} - (tool call)")
+    
+    print("\n" + "=" * 50)
+    print("Example complete!")
+
+
+if __name__ == "__main__":
+    main()

scripts/evaluate_qwen35.py

@@ -1,114 +0,0 @@
-#!/usr/bin/env python3
-"""Evaluate Qwen3.5:35B as a local model option -- Issue #288, Epic #281."""
-import json, sys, time
-from dataclasses import dataclass, field
-from typing import Any, Dict
-
-@dataclass
-class ModelSpec:
-    name: str = "Qwen3.5-35B-A3B"
-    ollama_tag: str = "qwen3.5:35b"
-    hf_id: str = "Qwen/Qwen3.5-35B-A3B"
-    architecture: str = "MoE (Mixture of Experts)"
-    total_params: str = "35B"
-    active_params: str = "3B per token"
-    context_length: int = 131072
-    license: str = "Apache 2.0"
-    tool_use_support: bool = True
-    json_mode_support: bool = True
-    function_calling: bool = True
-    quantization_options: Dict[str, int] = field(default_factory=lambda: {
-        "Q8_0": 36, "Q6_K": 28, "Q5_K_M": 24, "Q4_K_M": 20,
-        "Q4_0": 18, "Q3_K_M": 15, "Q2_K": 12,
-    })
-
-FLEET_MODELS = {
-    "qwen3.5:35b (candidate)": {"params_total": "35B", "context": "128K", "local": True, "tool_use": True, "reasoning": "good"},
-    "gemma4 (current local)": {"params_total": "9B", "context": "128K", "local": True, "tool_use": True, "reasoning": "good"},
-    "hermes4:14b (current local)": {"params_total": "14B", "context": "8K", "local": True, "tool_use": True, "reasoning": "good"},
-    "qwen2.5:7b (fleet)": {"params_total": "7B", "context": "32K", "local": True, "tool_use": True, "reasoning": "moderate"},
-    "claude-sonnet-4 (cloud)": {"params_total": "?", "context": "200K", "local": False, "tool_use": True, "reasoning": "excellent"},
-    "mimo-v2-pro (cloud free)": {"params_total": "?", "context": "128K", "local": False, "tool_use": True, "reasoning": "good"},
-}
-
-SECURITY_CRITERIA = [
-    {"criterion": "Data locality", "weight": "CRITICAL", "score": 10, "notes": "All inference local via Ollama. Zero exfiltration."},
-    {"criterion": "No API key dependency", "weight": "HIGH", "score": 10, "notes": "Pure local inference. No external creds needed."},
-    {"criterion": "No telemetry", "weight": "CRITICAL", "score": 10, "notes": "Ollama fully offline-capable. No phone-home."},
-    {"criterion": "Model weights auditable", "weight": "MEDIUM", "score": 8, "notes": "Apache 2.0, HF SHA verification. MoE harder to audit."},
-    {"criterion": "Tool-use safety", "weight": "HIGH", "score": 7, "notes": "Function calling supported, MoE routing less predictable. Benchmark: #502."},
-    {"criterion": "Privacy filter compat", "weight": "HIGH", "score": 9, "notes": "Local = Privacy Filter unnecessary for most queries."},
-    {"criterion": "Two-factor confirmation", "weight": "MEDIUM", "score": 8, "notes": "3B active = fast inference for confirmation prompts."},
-    {"criterion": "Prompt injection resistance", "weight": "HIGH", "score": 6, "notes": "3B active may be weaker. Needs red-team (#324)."},
-]
-
-HARDWARE_PROFILES = {
-    "mac_m2_ultra_192gb": {"name": "Mac Studio M2 Ultra (192GB)", "mem_gb": 192, "fits_q4": True, "fits_q8": True, "rec": "Q6_K", "tok_sec": 40},
-    "mac_m4_pro_48gb": {"name": "Mac Mini M4 Pro (48GB)", "mem_gb": 48, "fits_q4": True, "fits_q8": False, "rec": "Q4_K_M", "tok_sec": 30},
-    "mac_m1_16gb": {"name": "Mac M1 (16GB)", "mem_gb": 16, "fits_q4": False, "fits_q8": False, "rec": None, "tok_sec": None},
-    "rtx_4090_24gb": {"name": "NVIDIA RTX 4090 (24GB)", "mem_gb": 24, "fits_q4": True, "fits_q8": False, "rec": "Q5_K_M", "tok_sec": 50},
-    "rtx_3090_24gb": {"name": "NVIDIA RTX 3090 (24GB)", "mem_gb": 24, "fits_q4": True, "fits_q8": False, "rec": "Q4_K_M", "tok_sec": 35},
-    "runpod_l40s_48gb": {"name": "RunPod L40S (48GB)", "mem_gb": 48, "fits_q4": True, "fits_q8": True, "rec": "Q6_K", "tok_sec": 60},
-}
-
-def check_ollama_status() -> Dict[str, Any]:
-    import subprocess
-    result = {"running": False, "models": [], "qwen35_available": False}
-    try:
-        r = subprocess.run(["curl", "-s", "--max-time", "5", "http://localhost:11434/api/tags"], capture_output=True, text=True, timeout=10)
-        if r.returncode == 0:
-            data = json.loads(r.stdout)
-            result["running"] = True
-            result["models"] = [m["name"] for m in data.get("models", [])]
-            result["qwen35_available"] = any("qwen3.5" in m.lower() for m in result["models"])
-    except Exception as e:
-        result["error"] = str(e)
-    return result
-
-def generate_report() -> str:
-    spec = ModelSpec()
-    ollama = check_ollama_status()
-    lines = ["=" * 72, "Qwen3.5:35B EVALUATION REPORT -- Issue #288", "Epic #281 -- Vitalik Secure LLM Architecture", "=" * 72]
-    lines.append("\n## 1. Model Specification\n")
-    lines.append(f"  Name: {spec.name}  |  Arch: {spec.architecture}")
-    lines.append(f"  Params: {spec.total_params} total, {spec.active_params}  |  Context: {spec.context_length:,} tokens")
-    lines.append(f"  License: {spec.license}  |  Tools: {spec.tool_use_support}  |  JSON: {spec.json_mode_support}")
-    lines.append("\n## 2. VRAM\n")
-    for q, vram in sorted(spec.quantization_options.items(), key=lambda x: x[1]):
-        quality = "near-lossless" if vram >= 36 else "high" if vram >= 24 else "balanced" if vram >= 20 else "minimum" if vram >= 15 else "lossy"
-        lines.append(f"  {q:<10} {vram:>4}GB  {quality}")
-    lines.append("\n## 3. Hardware\n")
-    for hw in HARDWARE_PROFILES.values():
-        lines.append(f"  {hw['name']}  {hw['mem_gb']}GB  Q4:{'YES' if hw['fits_q4'] else 'NO '}  Rec:{hw['rec'] or 'N/A'}  ~{hw['tok_sec'] or 'N/A'} tok/s")
-    lines.append("\n## 4. Security (Vitalik Framework)\n")
-    wm = {"CRITICAL": 3, "HIGH": 2, "MEDIUM": 1}
-    tw = sum(wm[c["weight"]] for c in SECURITY_CRITERIA)
-    ws = sum(c["score"] * wm[c["weight"]] for c in SECURITY_CRITERIA)
-    for c in SECURITY_CRITERIA:
-        lines.append(f"  [{c['weight']:<8}] {c['criterion']}: {c['score']}/10 -- {c['notes']}")
-    avg = ws / tw
-    lines.append(f"\n  Weighted: {avg:.1f}/10  Verdict: {'STRONG' if avg >= 8 else 'ADEQUATE'}")
-    lines.append("\n## 5. Fleet Comparison\n")
-    for name, d in FLEET_MODELS.items():
-        lines.append(f"  {name:<35} {d['params_total']:<6} {d['context']:<6} {'Local' if d['local'] else 'Cloud'}  {d['reasoning']}")
-    lines.append("\n## 6. Ollama\n")
-    lines.append(f"  Running: {'Yes' if ollama['running'] else 'No'}  |  Models: {', '.join(ollama['models']) or 'none'}")
-    lines.append(f"  Qwen3.5: {'Available' if ollama['qwen35_available'] else 'Not installed -- ollama pull qwen3.5:35b'}")
-    lines.append("\n## 7. Recommendation\n")
-    lines.append("  VERDICT: APPROVED for local deployment as privacy-sensitive tier")
-    lines.append("\n  + Perfect data sovereignty, 128K context, Apache 2.0, MoE speed")
-    lines.append("  + Tool use + JSON mode, eliminates Privacy Filter for most queries")
-    lines.append("  - 20GB VRAM at Q4, MoE less predictable, needs red-team testing")
-    lines.append("\n  Follow-up issues filed:")
-    lines.append("    #502: live tool dispatch benchmark")
-    lines.append("    #503: reasoning benchmark vs hermes4:14b")
-    lines.append("    #518: document minimum hardware requirements fleet-wide")
-    lines.append("    #324: prompt injection red-team testing")
-    lines.append("\n  Deployment: ollama pull qwen3.5:35b -> config.yaml privacy_model")
-    return "\n".join(lines)
-
-if __name__ == "__main__":
-    if "--check-ollama" in sys.argv:
-        print(json.dumps(check_ollama_status(), indent=2))
-    else:
-        print(generate_report())

tests/test_evaluate_qwen35.py

@@ -1,50 +0,0 @@
-"""Tests for Qwen3.5:35B evaluation -- Issue #288."""
-import pytest
-from scripts.evaluate_qwen35 import ModelSpec, FLEET_MODELS, SECURITY_CRITERIA, HARDWARE_PROFILES, check_ollama_status, generate_report
-
-class TestModelSpec:
-    def test_fields(self):
-        s = ModelSpec()
-        assert s.name == "Qwen3.5-35B-A3B"
-        assert s.context_length == 131072
-        assert s.license == "Apache 2.0"
-        assert s.tool_use_support is True
-    def test_quant_vram_decreasing(self):
-        s = ModelSpec()
-        items = sorted(s.quantization_options.items(), key=lambda x: x[1])
-        for i in range(1, len(items)):
-            assert items[i][1] >= items[i-1][1]
-
-class TestSecurity:
-    def test_scores(self):
-        for c in SECURITY_CRITERIA:
-            assert 1 <= c["score"] <= 10
-    def test_weighted_avg(self):
-        wm = {"CRITICAL": 3, "HIGH": 2, "MEDIUM": 1}
-        tw = sum(wm[c["weight"]] for c in SECURITY_CRITERIA)
-        ws = sum(c["score"] * wm[c["weight"]] for c in SECURITY_CRITERIA)
-        assert ws / tw >= 7.0
-
-class TestHardware:
-    def test_m2_fits(self):
-        assert HARDWARE_PROFILES["mac_m2_ultra_192gb"]["fits_q4"] is True
-    def test_m1_no(self):
-        assert HARDWARE_PROFILES["mac_m1_16gb"]["fits_q4"] is False
-
-class TestReport:
-    def test_sections(self):
-        r = generate_report()
-        for s in ["Model Specification", "VRAM", "Hardware", "Security", "Fleet", "Recommendation"]:
-            assert s in r
-    def test_approved(self):
-        assert "APPROVED" in generate_report()
-    def test_follow_up_issues_referenced(self):
-        r = generate_report()
-        for issue in ["#502", "#503", "#518", "#324"]:
-            assert issue in r
-
-class TestOllama:
-    def test_returns_dict(self):
-        r = check_ollama_status()
-        assert isinstance(r, dict)
-        assert "running" in r

tools/session_templates.py

@@ -0,0 +1,384 @@
+"""
+Session templates for code-first seeding.
+
+Based on research finding: Code-heavy sessions (execute_code dominant in first 30 turns)
+improve over time. File-heavy sessions degrade. The key is deterministic feedback loops.
+
+This module provides:
+1. Template extraction from successful sessions
+2. Task type classification (code, file, research)
+3. Template storage in ~/.hermes/session-templates/
+4. Template injection into new sessions
+"""
+
+import json
+import logging
+import os
+import sqlite3
+import time
+from pathlib import Path
+from typing import Dict, List, Optional, Any
+from dataclasses import dataclass, asdict
+from enum import Enum
+
+logger = logging.getLogger(__name__)
+
+# Default template directory
+DEFAULT_TEMPLATE_DIR = Path.home() / ".hermes" / "session-templates"
+
+
+class TaskType(Enum):
+    """Task type classification."""
+    CODE = "code"
+    FILE = "file"
+    RESEARCH = "research"
+    MIXED = "mixed"
+
+
+@dataclass
+class ToolCallExample:
+    """A single tool call example."""
+    tool_name: str
+    arguments: Dict[str, Any]
+    result: str
+    success: bool
+    
+    def to_dict(self) -> Dict[str, Any]:
+        return asdict(self)
+    
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> 'ToolCallExample':
+        return cls(**data)
+
+
+@dataclass
+class SessionTemplate:
+    """A session template with tool call examples."""
+    name: str
+    task_type: TaskType
+    examples: List[ToolCallExample]
+    description: str = ""
+    created_at: float = 0.0
+    usage_count: int = 0
+    
+    def __post_init__(self):
+        if self.created_at == 0.0:
+            self.created_at = time.time()
+    
+    def to_dict(self) -> Dict[str, Any]:
+        data = asdict(self)
+        data['task_type'] = self.task_type.value
+        return data
+    
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> 'SessionTemplate':
+        data['task_type'] = TaskType(data['task_type'])
+        examples_data = data.get('examples', [])
+        data['examples'] = [ToolCallExample.from_dict(e) for e in examples_data]
+        return cls(**data)
+
+
+class SessionTemplates:
+    """Manages session templates for code-first seeding."""
+    
+    def __init__(self, template_dir: Optional[Path] = None):
+        self.template_dir = template_dir or DEFAULT_TEMPLATE_DIR
+        self.template_dir.mkdir(parents=True, exist_ok=True)
+        self.templates: Dict[str, SessionTemplate] = {}
+        self._load_templates()
+    
+    def _load_templates(self):
+        """Load all templates from disk."""
+        for template_file in self.template_dir.glob("*.json"):
+            try:
+                with open(template_file, 'r') as f:
+                    data = json.load(f)
+                    template = SessionTemplate.from_dict(data)
+                    self.templates[template.name] = template
+            except Exception as e:
+                logger.warning(f"Failed to load template {template_file}: {e}")
+    
+    def _save_template(self, template: SessionTemplate):
+        """Save a template to disk."""
+        template_file = self.template_dir / f"{template.name}.json"
+        with open(template_file, 'w') as f:
+            json.dump(template.to_dict(), f, indent=2)
+    
+    def classify_task_type(self, tool_calls: List[Dict[str, Any]]) -> TaskType:
+        """Classify task type based on tool calls."""
+        if not tool_calls:
+            return TaskType.MIXED
+        
+        # Count tool types
+        code_tools = {'execute_code', 'code_execution'}
+        file_tools = {'read_file', 'write_file', 'patch', 'search_files'}
+        research_tools = {'web_search', 'web_fetch', 'browser_navigate'}
+        
+        tool_names = [tc.get('tool_name', '') for tc in tool_calls]
+        
+        code_count = sum(1 for t in tool_names if t in code_tools)
+        file_count = sum(1 for t in tool_names if t in file_tools)
+        research_count = sum(1 for t in tool_names if t in research_tools)
+        
+        total = len(tool_calls)
+        if total == 0:
+            return TaskType.MIXED
+        
+        # Determine dominant type (60% threshold)
+        if code_count / total > 0.6:
+            return TaskType.CODE
+        elif file_count / total > 0.6:
+            return TaskType.FILE
+        elif research_count / total > 0.6:
+            return TaskType.RESEARCH
+        else:
+            return TaskType.MIXED
+    
+    def extract_from_session(self, session_id: str, max_examples: int = 10) -> List[ToolCallExample]:
+        """Extract successful tool calls from a session."""
+        db_path = Path.home() / ".hermes" / "state.db"
+        if not db_path.exists():
+            return []
+        
+        try:
+            conn = sqlite3.connect(str(db_path))
+            conn.row_factory = sqlite3.Row
+            
+            # Get messages with tool calls
+            cursor = conn.execute("""
+                SELECT role, content, tool_calls, tool_name
+                FROM messages 
+                WHERE session_id = ? 
+                ORDER BY timestamp
+                LIMIT 100
+            """, (session_id,))
+            
+            messages = cursor.fetchall()
+            conn.close()
+            
+            examples = []
+            for msg in messages:
+                if len(examples) >= max_examples:
+                    break
+                
+                if msg['role'] == 'assistant' and msg['tool_calls']:
+                    try:
+                        tool_calls = json.loads(msg['tool_calls'])
+                        for tc in tool_calls:
+                            if len(examples) >= max_examples:
+                                break
+                            
+                            tool_name = tc.get('function', {}).get('name')
+                            if not tool_name:
+                                continue
+                            
+                            try:
+                                arguments = json.loads(tc.get('function', {}).get('arguments', '{}'))
+                            except:
+                                arguments = {}
+                            
+                            examples.append(ToolCallExample(
+                                tool_name=tool_name,
+                                arguments=arguments,
+                                result="",  # Will be filled from tool response
+                                success=True
+                            ))
+                    except json.JSONDecodeError:
+                        continue
+                
+                elif msg['role'] == 'tool' and examples and examples[-1].result == "":
+                    examples[-1].result = msg['content'] or ""
+            
+            return examples
+            
+        except Exception as e:
+            logger.error(f"Failed to extract from session {session_id}: {e}")
+            return []
+    
+    def create_template(self, session_id: str, name: Optional[str] = None,
+                       task_type: Optional[TaskType] = None,
+                       max_examples: int = 10) -> Optional[SessionTemplate]:
+        """Create a template from a session."""
+        examples = self.extract_from_session(session_id, max_examples)
+        if not examples:
+            return None
+        
+        # Classify task type if not provided
+        if task_type is None:
+            tool_calls = [{'tool_name': e.tool_name} for e in examples]
+            task_type = self.classify_task_type(tool_calls)
+        
+        # Generate name if not provided
+        if name is None:
+            name = f"{task_type.value}_{session_id[:8]}_{int(time.time())}"
+        
+        # Create template
+        template = SessionTemplate(
+            name=name,
+            task_type=task_type,
+            examples=examples,
+            description=f"Template with {len(examples)} examples"
+        )
+        
+        # Save template
+        self.templates[name] = template
+        self._save_template(template)
+        
+        logger.info(f"Created template {name} with {len(examples)} examples")
+        return template
+    
+    def get_template(self, task_type: TaskType) -> Optional[SessionTemplate]:
+        """Get the best template for a task type."""
+        matching = [t for t in self.templates.values() if t.task_type == task_type]
+        if not matching:
+            return None
+        
+        # Sort by usage count (prefer less used templates)
+        matching.sort(key=lambda t: t.usage_count)
+        return matching[0]
+    
+    def inject_into_messages(self, template: SessionTemplate,
+                           messages: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+        """Inject template examples into messages."""
+        if not template.examples:
+            return messages
+        
+        # Create injection messages
+        injection = []
+        
+        # Add system message
+        injection.append({
+            "role": "system",
+            "content": f"Session template: {template.name} ({template.task_type.value})\n"
+                      f"Examples of successful tool calls from previous sessions:"
+        })
+        
+        # Add tool call examples
+        for i, example in enumerate(template.examples):
+            # Assistant message with tool call
+            injection.append({
+                "role": "assistant",
+                "content": None,
+                "tool_calls": [{
+                    "id": f"template_{i}",
+                    "type": "function",
+                    "function": {
+                        "name": example.tool_name,
+                        "arguments": json.dumps(example.arguments)
+                    }
+                }]
+            })
+            
+            # Tool response
+            injection.append({
+                "role": "tool",
+                "tool_call_id": f"template_{i}",
+                "content": example.result
+            })
+        
+        # Insert after system messages
+        insert_index = 0
+        for i, msg in enumerate(messages):
+            if msg.get("role") != "system":
+                break
+            insert_index = i + 1
+        
+        # Insert injection
+        for i, msg in enumerate(injection):
+            messages.insert(insert_index + i, msg)
+        
+        # Update usage count
+        template.usage_count += 1
+        self._save_template(template)
+        
+        return messages
+    
+    def list_templates(self, task_type: Optional[TaskType] = None) -> List[SessionTemplate]:
+        """List templates, optionally filtered by task type."""
+        templates = list(self.templates.values())
+        if task_type:
+            templates = [t for t in templates if t.task_type == task_type]
+        templates.sort(key=lambda t: t.created_at, reverse=True)
+        return templates
+    
+    def delete_template(self, name: str) -> bool:
+        """Delete a template."""
+        if name not in self.templates:
+            return False
+        
+        del self.templates[name]
+        template_file = self.template_dir / f"{name}.json"
+        if template_file.exists():
+            template_file.unlink()
+        
+        logger.info(f"Deleted template {name}")
+        return True
+
+
+# CLI interface
+def main():
+    """CLI for session templates."""
+    import argparse
+    
+    parser = argparse.ArgumentParser(description="Session Templates")
+    subparsers = parser.add_subparsers(dest="command")
+    
+    # List templates
+    list_parser = subparsers.add_parser("list", help="List templates")
+    list_parser.add_argument("--type", choices=["code", "file", "research", "mixed"])
+    
+    # Create template
+    create_parser = subparsers.add_parser("create", help="Create template from session")
+    create_parser.add_argument("session_id", help="Session ID")
+    create_parser.add_argument("--name", help="Template name")
+    create_parser.add_argument("--type", choices=["code", "file", "research", "mixed"])
+    create_parser.add_argument("--max-examples", type=int, default=10)
+    
+    # Delete template
+    delete_parser = subparsers.add_parser("delete", help="Delete template")
+    delete_parser.add_argument("name", help="Template name")
+    
+    args = parser.parse_args()
+    
+    templates = SessionTemplates()
+    
+    if args.command == "list":
+        task_type = TaskType(args.type) if args.type else None
+        template_list = templates.list_templates(task_type)
+        
+        if not template_list:
+            print("No templates found")
+            return
+        
+        print(f"Found {len(template_list)} templates:")
+        for t in template_list:
+            print(f"  {t.name}: {t.task_type.value} ({len(t.examples)} examples, used {t.usage_count} times)")
+    
+    elif args.command == "create":
+        task_type = TaskType(args.type) if args.type else None
+        template = templates.create_template(
+            args.session_id,
+            name=args.name,
+            task_type=task_type,
+            max_examples=args.max_examples
+        )
+        
+        if template:
+            print(f"Created template: {template.name}")
+            print(f"  Type: {template.task_type.value}")
+            print(f"  Examples: {len(template.examples)}")
+        else:
+            print("Failed to create template")
+    
+    elif args.command == "delete":
+        if templates.delete_template(args.name):
+            print(f"Deleted template: {args.name}")
+        else:
+            print(f"Template not found: {args.name}")
+    
+    else:
+        parser.print_help()
+
+
+if __name__ == "__main__":
+    main()