feat: implement real-time model performance tracking and advisory system

Author: sidrat2612

.gitignore

@@ -17,4 +17,5 @@ env/
 htmlcov/
 .mypy_cache/
 .ruff_cache/
+.routesmith/
 .DS_Store

CHANGELOG.md

@@ -2,7 +2,20 @@
 
 All notable changes to this project will be documented in this file.
 
-## [0.1.4] - Unreleased
+## [0.1.5] - Unreleased
+
+### Added
+- Real-time model performance tracking with rolling-window persistence.
+- `PerformanceTracker` class records per-model task outcomes, durations, and success rates.
+- `routesmith stats` CLI command to view model performance tables and advisory.
+- Performance advisory automatically injected into run results when models underperform.
+- `PerformanceTracker` exported from the public API.
+
+### Fixed
+- Gemini CLI adapter no longer returns a hardcoded fallback model when no environment is detected.
+- `SkillConfig.routing_preference` now normalizes input strings (case-insensitive, dash/space tolerant).
+
+## [0.1.4] - 2026-05-07
 
 ### Added
 - GitHub Release badge and direct release asset links in the README.

README.md

@@ -145,6 +145,14 @@ Start it with `routesmith serve-stdio`.
 
 This lets IDE extensions and agents call routesmith as a tool.
 
+## Performance Tracking
+
+routesmith records per-model task outcomes (duration, success/failure, capability class) across runs. Data is stored in `.routesmith/performance.json` and accumulates over time within a project.
+
+View stats with `routesmith stats`. Filter to a specific model with `routesmith stats --model claude-sonnet-4-6`. Clear tracked data with `routesmith stats --clear`.
+
+When a model's historical success rate drops below 70% or average latency exceeds 5 seconds, routesmith injects performance advisory messages into run results automatically.
+
 ## Install Configs for Hosts
 
 Generate host-specific configuration files:
@@ -197,7 +205,7 @@ Development setup:
 - [x] Cost-aware routing
 - [x] Python policy plugins
 - [x] Gemini CLI host adapter
-- [ ] Real-time model performance tracking
+- [x] Real-time model performance tracking
 - [ ] Additional host adapters
 
 ## License

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "routesmith"
-version = "0.1.4"
+version = "0.1.5"
 description = "Auto-route coding agent tasks to the best model in your IDE. Python library + MCP server for Claude Code, Codex, Gemini CLI, Copilot, Cursor, and Aider."
 readme = "README.md"
 license = "MIT"

src/routesmith/__init__.py

@@ -18,6 +18,7 @@
 )
 from routesmith.executor import Executor
 from routesmith.planner import Planner
+from routesmith.performance import PerformanceTracker
 from routesmith.policy_plugins import BasePolicyPlugin, PolicyPluginContext, PolicyPluginResult
 from routesmith.router import Router
 from routesmith.hosts.detector import detect_host, get_host_capabilities
@@ -64,6 +65,7 @@ def _read_version_from_pyproject() -> str:
     "PolicyPluginResult",
     "Executor",
     "Planner",
+    "PerformanceTracker",
     "Router",
 ]
 

src/routesmith/cli.py

@@ -238,6 +238,59 @@ def serve_stdio_cmd() -> None:
     run_stdio_server()
 
 
+@app.command("stats")
+def stats_cmd(
+    model: str | None = typer.Option(None, "--model", "-m", help="Filter stats by model name."),
+    clear: bool = typer.Option(False, "--clear", help="Clear all tracked performance data."),
+) -> None:
+    """Show real-time model performance statistics."""
+    from routesmith.performance import PerformanceTracker
+
+    tracker = PerformanceTracker()
+
+    if clear:
+        tracker.clear()
+        console.print("[green]Performance data cleared.[/green]")
+        return
+
+    stats = tracker.get_model_stats(model=model)
+    if not stats:
+        console.print("[dim]No performance data recorded yet. Run some tasks first.[/dim]")
+        return
+
+    table = Table(title="Model Performance")
+    table.add_column("Model", style="cyan")
+    table.add_column("Tasks", justify="right")
+    table.add_column("Success", justify="right", style="green")
+    table.add_column("Fail", justify="right", style="red")
+    table.add_column("Rate", justify="right")
+    table.add_column("Avg (ms)", justify="right")
+    table.add_column("Min (ms)", justify="right")
+    table.add_column("Max (ms)", justify="right")
+
+    for s in stats:
+        rate_color = "green" if s.success_rate >= 0.9 else ("yellow" if s.success_rate >= 0.7 else "red")
+        table.add_row(
+            s.model,
+            str(s.total_tasks),
+            str(s.successes),
+            str(s.failures),
+            f"[{rate_color}]{s.success_rate:.0%}[/{rate_color}]",
+            f"{s.avg_duration_ms:.1f}",
+            f"{s.min_duration_ms:.1f}",
+            f"{s.max_duration_ms:.1f}",
+        )
+
+    console.print(table)
+
+    # Show advisory if any
+    advisory = tracker.get_performance_advisory()
+    if advisory:
+        console.print()
+        for msg in advisory:
+            console.print(f"  [yellow]⚠[/yellow] {msg}")
+
+
 # Install sub-commands
 @install_app.callback(invoke_without_command=True)
 def install_default(

src/routesmith/executor.py

@@ -8,6 +8,7 @@
 from routesmith.config import load_config
 from routesmith.hosts.detector import detect_host, get_host_adapter
 from routesmith.metrics import RouteMetrics, compute_metrics
+from routesmith.performance import PerformanceTracker
 from routesmith.planner import Planner
 from routesmith.review import review_plan, review_results
 from routesmith.router import Router
@@ -27,6 +28,7 @@ class Executor:
     def __init__(self, config: SkillConfig | None = None) -> None:
         self.config = config or load_config()
         self.planner = Planner()
+        self.performance_tracker = PerformanceTracker()
 
     def explain(self, prompt: str) -> RoutePlan:
         """Explain the route plan without executing."""
@@ -117,6 +119,14 @@ def run(
             metrics=metrics.model_dump(),
         )
 
+        # Record performance data
+        self.performance_tracker.record_run(plan, task_results)
+
+        # Inject performance advisory
+        perf_advisory = self.performance_tracker.get_performance_advisory()
+        if perf_advisory:
+            result.advisory.extend(perf_advisory)
+
         # Persist route if configured
         if self.config.save_routes:
             save_route(plan, task_results, metrics, self.config.routes_dir)