feat: integrate local LLM API (homelab-frontend) + Pi skills

Replace Anthropic client with OpenAI-compatible client targeting https://api.riotpiao.com.
Configure models: reasoning (Planner/Judge), ornith:35b (Implementer).
Add health check on startup.
Add Pi provider support for skill preparation (--pi-provider=local-llm).

Files changed:
- action/llm/client.go: OpenAI-compatible HTTP client + HealthCheck()
- action/llm/client_test.go: Unit tests for model validation & health
- cmd/starter/main.go: Health check before workflow, local model defaults
- statemachine/types.go: PiProvider field for OrchestratorInput

Models:
- Planner: reasoning (smart decisions)
- Judge: reasoning (quality review)
- Implementer: ornith:35b (cheap execution)

Skills: pi clone-or-fetch --provider=local-llm with 504 timeout learning.

Verification: go build ./cmd/starter ./cmd/worker ./action/llm ✓
Tests: go test -v ./action/llm ✓ (all passing)

cmd/starter/main.go

@@ -9,6 +9,7 @@ import (
 	"time"
 
 	"go.temporal.io/sdk/client"
+	"github.com/rockliang/poimen/workflows/action/llm"
 	"github.com/rockliang/poimen/workflows/internal/config"
 	"github.com/rockliang/poimen/workflows/internal/health"
 	"github.com/rockliang/poimen/workflows/internal/logging"
@@ -21,9 +22,10 @@ func main() {
 		remoteURL        = flag.String("remote", "", "remote URL")
 		milestone        = flag.String("milestone", "T0", "milestone ID")
 		dryRun           = flag.Bool("dry-run", false, "disable git push/merge")
-		plannerModel     = flag.String("planner-model", "ornith", "planner model ID")
-		judgeModel       = flag.String("judge-model", "ornith", "judge model ID")
-		implementerModel = flag.String("implementer-model", "claude-sonnet-5", "implementer model ID")
+		plannerModel     = flag.String("planner-model", "reasoning", "planner model ID (local-llm)")
+		judgeModel       = flag.String("judge-model", "reasoning", "judge model ID (local-llm)")
+		implementerModel = flag.String("implementer-model", "ornith:35b", "implementer model ID (local-llm ornith)")
+		piProvider       = flag.String("pi-provider", "local-llm", "pi provider name for skills (local-llm)")
 		healthCheck      = flag.Bool("health", false, "check health and exit")
 	)
 	flag.Parse()
@@ -78,6 +80,7 @@ func main() {
 		Milestone:          *milestone,
 		DryRun:             *dryRun,
 		MaxCyclesBeforeCAN: 100,
+		PiProvider:         *piProvider,
 		Config: statemachine.OrchestratorConfig{
 			SystemPrompt: "You are an expert software developer orchestrating multi-agent work.",
 			Skills:       []statemachine.SkillRef{},
@@ -109,6 +112,17 @@ func main() {
 		},
 	}
 
+	// Health check: verify local LLM API is reachable
+	logging.Info("checking local LLM API connectivity", logging.String("url", "https://api.riotpiao.com"))
+	llmClient, err := llm.NewClient()
+	if err != nil {
+		logging.Fatal("failed to create LLM client", logging.Err(err))
+	}
+	if err := llmClient.HealthCheck(context.Background()); err != nil {
+		logging.Fatal("local LLM API health check failed", logging.Err(err), logging.String("hint", "ensure homelab-frontend gateway is running and accessible"))
+	}
+	logging.Info("local LLM API is reachable", logging.String("planner-model", *plannerModel), logging.String("judge-model", *judgeModel), logging.String("implementer-model", *implementerModel))
+
 	// Start workflow
 	workflowID := "orch-" + strings.ReplaceAll(*repoPath, "/", "-")
 	logging.Info("starting orchestrator workflow", logging.String("workflowID", workflowID), logging.String("repo", *repoPath))