feat: migration for workflow relations and RAG indexing

feat: wire GraphRAG API handlers, activities, and database layer
chore: remove docker-compose (use k8s + CI/CD only)
2026-09-05 06:01:08 -07:00 · 2026-09-05 06:00:58 -07:00 · 2026-09-05 05:58:53 -07:00 · 2026-09-05 05:58:18 -07:00 · 2026-09-05 05:57:34 -07:00 · 2026-09-05 05:57:08 -07:00
22 changed files with 1817 additions and 726 deletions
@@ -0,0 +1,11 @@
+.git
+.gitignore
+*.md
+.env.local
+.env
+tests/
+*.test.go
+coverage/
+.DS_Store
+k8s/
+migrations/
@@ -0,0 +1,293 @@
+# Poimen Application Deployment
+
+## Overview
+
+Poimen is a unified application consisting of three services:
+- **poimen-memory**: Memory/Graph RAG service
+- **poimen-workflows**: Temporal orchestration + API
+- **poimen-frontend**: Next.js frontend
+
+All services are deployed together as a single application in the `poimen` namespace.
+
+## Local Development
+
+### Prerequisites
+- Docker
+- Docker Compose
+- Node.js 18+
+- Go 1.21+
+- Python 3.11+
+
+### Start Local Stack
+
+```bash
+docker-compose up -d
+```
+
+This starts:
+- PostgreSQL (memory + workflows DBs)
+- Redis (cache)
+- Temporal (workflow orchestration)
+- poimen-memory (8000)
+- poimen-workflows (8080)
+- poimen-workflows-worker
+- poimen-frontend (3000)
+
+### Access Services
+
+- Frontend: http://localhost:3000
+- Workflows API: http://localhost:8080
+- Memory API: http://localhost:8000
+- Temporal UI: http://localhost:8233
+
+### Stop Stack
+
+```bash
+docker-compose down
+```
+
+## Building & Pushing Images
+
+### Build All Services
+
+```bash
+./build-push.sh latest
+```
+
+Or specific services:
+
+```bash
+docker build -t forgejo.riotpiao.com/rock/poimen-memory:v1.0.0 ./memory
+docker push forgejo.riotpiao.com/rock/poimen-memory:v1.0.0
+```
+
+### Image Tagging Strategy
+
+- `latest`: Development/staging
+- `v1.0.0`, `v1.0.1`, etc.: Production releases
+- `main-{commit-hash}`: CI/CD automated builds
+
+## Kubernetes Deployment
+
+### Prerequisites
+
+- Kubernetes cluster (1.24+)
+- kubectl configured
+- Kustomize installed
+- Registry credentials configured
+
+### Deploy to Cluster
+
+```bash
+cd k8s
+./deploy.sh -a
+```
+
+Or with specific tags:
+
+```bash
+./deploy.sh -m v1.0.0 -w v1.0.0 -f v1.0.0
+```
+
+### Verify Deployment
+
+```bash
+kubectl get pods -n poimen
+kubectl get svc -n poimen
+kubectl logs -n poimen -l app=poimen-workflows
+```
+
+## Configuration
+
+### Environment Variables
+
+Configure in `k8s/poimen-application.yaml` under `spec.template.spec.env`:
+
+**Common:**
+- `TEMPORAL_HOST`: Temporal server (default: temporal:7233)
+- `DATABASE_URL`: PostgreSQL connection
+- `JWT_SECRET`: JWT signing key
+- `LOG_LEVEL`: debug|info|warn|error
+
+**Memory Service:**
+- `REDIS_URL`: Redis connection
+- `ELASTICSEARCH_URL`: Optional full-text search
+
+**Workflows Service:**
+- `MEMORY_SERVICE_URL`: Internal memory service URL
+
+**Frontend:**
+- `NEXT_PUBLIC_WORKFLOWS_API`: External workflows API
+- `NEXT_PUBLIC_MEMORY_API`: External memory API
+- `OAUTH_CLIENT_ID`, `OAUTH_CLIENT_SECRET`: Auth provider
+
+### Secrets
+
+Create secrets before deployment:
+
+```bash
+kubectl create secret generic poimen-db-credentials \
+  --from-literal=memory-url="postgresql://..." \
+  --from-literal=workflows-url="postgresql://..." \
+  -n poimen
+
+kubectl create secret generic poimen-secrets \
+  --from-literal=jwt-secret="..." \
+  --from-literal=oauth-client-id="..." \
+  --from-literal=oauth-client-secret="..." \
+  -n poimen
+```
+
+## Architecture
+
+```
+┌─────────────────────────────────────────────┐
+│             LoadBalancer Service            │
+│          poimen-frontend:80→3000            │
+└─────────────────┬───────────────────────────┘
+                  │
+          ┌───────┴────────┐
+          ▼                ▼
+   ┌──────────────┐  ┌──────────────┐
+   │   Frontend   │  │  Workflows   │
+   │   (3000)     │  │   API (8080) │
+   │   2 replicas │  │  2 replicas  │
+   └──────────────┘  └──────┬───────┘
+          │                 │
+          │           ┌─────┴─────┐
+          │           ▼           ▼
+          │      ┌─────────────────────┐
+          │      │  Temporal Cluster   │
+          │      │  (External)         │
+          │      └─────────────────────┘
+          │
+          └──────────────────┬──────────────┐
+                             ▼              ▼
+                    ┌──────────────┐  ┌──────────────┐
+                    │   Memory     │  │  PostgreSQL  │
+                    │  (8000)      │  │  (5432)      │
+                    │  1 replica   │  │              │
+                    └──────────────┘  └──────────────┘
+```
+
+## Scaling
+
+### Horizontal Scaling
+
+Adjust replicas in `k8s/poimen-application.yaml`:
+
+```yaml
+spec:
+  replicas: 3  # Increase this
+```
+
+Or patch:
+
+```bash
+kubectl patch deployment poimen-workflows -p '{"spec":{"replicas":3}}' -n poimen
+```
+
+### Resource Requests/Limits
+
+Add to container spec:
+
+```yaml
+resources:
+  requests:
+    cpu: 100m
+    memory: 256Mi
+  limits:
+    cpu: 500m
+    memory: 512Mi
+```
+
+## Monitoring & Logging
+
+### Check Status
+
+```bash
+kubectl get pods -n poimen -w
+kubectl describe pod <pod-name> -n poimen
+kubectl logs -n poimen -f -l app=poimen-workflows --all-containers=true
+```
+
+### Health Checks
+
+All services expose `/health` endpoint:
+
+```bash
+curl http://poimen-workflows:8080/health
+curl http://poimen-memory:8000/health
+curl http://poimen-frontend:3000/
+```
+
+## Updates & Rollbacks
+
+### Rolling Update
+
+```bash
+./deploy.sh -w v1.0.1
+```
+
+Kubernetes automatically rolls out with health checks.
+
+### View Rollout Status
+
+```bash
+kubectl rollout status deploy/poimen-workflows -n poimen
+```
+
+### Rollback
+
+```bash
+kubectl rollout undo deploy/poimen-workflows -n poimen
+```
+
+## Troubleshooting
+
+### Services Can't Connect
+
+Check service DNS:
+```bash
+kubectl run -it --rm debug --image=busybox --restart=Never -- nslookup poimen-workflows
+```
+
+### Database Migrations Failing
+
+```bash
+kubectl exec -it <workflows-pod> -n poimen -- \
+  ./workflows migrate up
+```
+
+### Temporal Worker Not Picking Up Activities
+
+Check worker logs:
+```bash
+kubectl logs -n poimen -l app=poimen-workflows --all-containers=true | grep -i activity
+```
+
+Verify activities registered in `cmd/worker/main.go`
+
+## CI/CD Integration
+
+### GitHub Actions Example
+
+```yaml
+name: Build & Push Poimen
+
+on:
+  push:
+    branches: [main]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - name: Build & Push
+        run: ./build-push.sh main-${{ github.sha }}
+```
+
+### Automatic Deployment
+
+Configure ArgoCD to watch `k8s/` directory for updates.
@@ -1,385 +1,22 @@
-# Multi-stage build for Poimen Temporal Worker
-# Stage 1: Builder - Compile Go binary and set up tools
-FROM golang:1.25-alpine AS builder
-
-WORKDIR /build
-
-# Install system dependencies (ast-grep, git, build essentials)
-RUN apk add --no-cache \
-    git \
-    curl \
-    wget \
-    ca-certificates \
-    gcc \
-    musl-dev \
-    bash \
-    && echo "[builder] System dependencies installed"
-
-# Install ast-grep CLI tool
-RUN curl -fsSL https://github.com/ast-grep/ast-grep/releases/download/0.24.0/sg-x86_64-unknown-linux-musl.tar.gz \
-    | tar xzf - -C /usr/local/bin \
-    && chmod +x /usr/local/bin/sg \
-    && sg --version \
-    && echo "[builder] ast-grep installed"
-
-# Install Node.js for pi CLI and browser-use
-RUN apk add --no-cache nodejs npm \
-    && echo "[builder] Node.js installed"
-
-# Install pi CLI globally
-RUN npm install -g @earendil-works/pi-coding-agent --unsafe-perm \
-    && pi --version \
-    && echo "[builder] pi CLI installed"
-
-# Install browser-use CLI for browser automation
-RUN npm install -g browser-use --unsafe-perm \
-    && browser-use --version \
-    && echo "[builder] browser-use CLI installed"
-
-# Set up pi home directory and skills
-RUN mkdir -p ~/.pi/agent/skills ~/.pi/agent/agents \
-    && echo "[builder] pi directories created"
-
-# Stage 2: Download pi skills (caveman & andrej karpathy)
-# Clone caveman skill from pi-agent repo
-RUN cd /tmp && git clone https://github.com/earendil-works/pi-agent.git pi-repo \
-    && mkdir -p ~/.pi/agent/skills/caveman \
-    && cp -r pi-repo/examples/skills/caveman/* ~/.pi/agent/skills/caveman/ 2>/dev/null || true \
-    && echo "[builder] caveman skill installed"
-
-# Create andrej karpathy skill manually (reference/training patterns)
-RUN mkdir -p ~/.pi/agent/skills/andrej-karpathy && cat > ~/.pi/agent/skills/andrej-karpathy/SKILL.md << 'EOF'
-# Andrej Karpathy LLM & AI Principles Skill
-
-Build neural networks and LLM systems with proven patterns from Andrej Karpathy.
-Topics: attention mechanisms, transformer training, inference optimization, edge cases.
-
-## Key Principles
-
-### 1. Simplicity First
- Start with minimal implementation
- Add complexity only when justified
- Test each component independently
- Use debugging tools effectively
-
-### 2. Neural Network Architecture
- Understand backward pass deeply
- Implement from scratch when possible
- Use visualization for debugging
- Profile before optimizing
-
-### 3. LLM Training Patterns
- Quality data > quantity
- Curriculum learning for complex tasks
- Loss landscape visualization
- Checkpoint strategy matters
-
-### 4. Inference Optimization
- Quantization without quality loss
- KV cache management
- Batch processing strategies
- Latency profiling
-
-### 5. Failure Analysis
- Log intermediate activations
- Check gradient flow
- Validate data pipeline
- Test edge cases explicitly
-
-## Usage in Poimen
-
-Apply when:
- Designing workflow stages (like training curricula)
- Optimizing inference (planner/judge/implementer prompts)
- Debugging convergence issues (retry patterns)
- Scaling to production (quantization patterns)
-
-## Resources
- github.com/karpathy/minGPT - Minimal GPT implementation
- youtube: "Neural Networks: Zero to Hero" series
- Papers: Attention Is All You Need, GPT series whitepapers
-EOF
-    && echo "[builder] andrej-karpathy skill created"
-
-# Create browser-use skill for web testing & automation
-RUN mkdir -p ~/.pi/agent/skills/browser-use && cat > ~/.pi/agent/skills/browser-use/SKILL.md << 'EOF'
-# browser-use: Browser Automation Skill
-
-Automate web browser interactions for testing, verification, and UI validation.
-Topics: headless browser control, visual testing, form automation, screenshot capture.
-
-## Key Capabilities
-
-### 1. Browser Control
- Launch headless Chrome/Firefox
- Navigate to URLs
- Wait for elements/navigation
- Handle popups/dialogs
-
-### 2. Interaction Patterns
- Click buttons/links
- Fill forms (text, dropdown, checkbox)
- Drag & drop
- Keyboard input
-
-### 3. Verification & Capture
- Screenshot capture
- Element inspection
- Accessibility checks
- Network monitoring
-
-### 4. Wait Strategies
- Wait for element visible
- Wait for navigation
- Wait for condition (custom JS)
- Timeout handling
-
-### 5. Error Recovery
- Retry failed actions
- Handle stale elements
- Browser crash recovery
- Memory leak prevention
-
-## Usage in Poimen Phases
-
-### Phase T2 (Implementation)
- Test generated UI code in real browser
- Verify visual layout matches spec
- Validate form inputs work correctly
-
-### Phase T3 (Verification)
- Visual regression testing
- Accessibility validation (ARIA, keyboard nav)
- Cross-browser verification
-
-### Phase T6 (Integration)
- End-to-end workflow testing
- External service integration testing
- User journey verification
-
-### Phase T9 (Release)
- Pre-release smoke tests
- Deployment verification
- Production canary testing
-
-## Example Workflows
-
-```bash
-# Launch browser and take screenshot
-browser-use screenshot "https://example.com" --file output.png
-
-# Fill form and submit
-browser-use interact "https://example.com" \
-  --click "#submit-btn" \
-  --type "#email" "[email protected]" \
-  --type "#password" "secretpass" \
-  --click ".submit"
-
-# Wait for dynamic content and extract data
-browser-use extract "https://example.com" \
-  --wait ".dynamic-content" \
-  --selector ".data-row" \
-  --output json
-
-# Accessibility audit
-browser-use audit "https://example.com" \
-  --check wcag2a \
-  --report a11y-report.html
-```
-
-## Integration with Poimen
-
-Pre-generated code can be tested:
-```bash
-# Generate code (T2)
-implementer_output = "function handleClick() { ... }"
-
-# Verify in browser (T3)
-browser-use interact "http://localhost:3000" \
-  --click ".test-button" \
-  --screenshot result.png
-
-# Compare with expected
-verify_visual_match(result.png, expected.png)
-```
-
-## Performance Notes
-
- Startup: ~2-5s per browser
- Action latency: 100-500ms per interaction
- Screenshot: 500ms-2s (depends on page size)
- Keep browser alive for batch operations (pool management)
-
-## Error Handling
-
- Transient: Network timeout → retry with backoff
- Permanent: Element not found → fail and log
- Flaky: Wait strategies → increase timeout gradually
- Memory: Reuse browser instances → kill after 10 uses
-
-## Resources
- docs.browseruse.com - Official documentation
- github.com/browser-use/browser-use - Source code
- Chrome DevTools Protocol - Advanced browser control
-EOF
-    && echo "[builder] browser-use skill created"
-
-# Copy Go source code
-COPY . /build/
-
-# Download Go dependencies
-RUN go mod download \
-    && echo "[builder] Go dependencies downloaded"
-
-# Build worker binary
-RUN CGO_ENABLED=1 GOOS=linux go build -o /build/worker ./cmd/worker \
-    && echo "[builder] Worker binary built"
-
-# Verify binary
-RUN file /build/worker && ls -lh /build/worker
-
-# Stage 3: Runtime - Minimal base image with runtime dependencies
-FROM alpine:3.20
-
-LABEL maintainer="Poimen Team"
-LABEL description="Poimen Temporal Worker with memory service, ast-grep, and browser automation"
+FROM golang:1.21-alpine as builder

 WORKDIR /app
+COPY go.mod go.sum ./
+RUN go mod download

-# Install runtime dependencies (including Chromium for browser-use)
-RUN apk add --no-cache \
-    ca-certificates \
-    git \
-    bash \
-    curl \
-    jq \
-    chromium \
-    chromium-chromedriver \
-    && echo "[runtime] Runtime dependencies installed"
+COPY . .
+RUN CGO_ENABLED=0 GOOS=linux go build -o workflows ./cmd/poimen/main.go

-# Install Node.js for pi CLI and browser-use
-RUN apk add --no-cache nodejs npm \
-    && echo "[runtime] Node.js installed"
+FROM alpine:latest

-# Install pi CLI in runtime image
-RUN npm install -g @earendil-works/pi-coding-agent --unsafe-perm \
-    && pi --version \
-    && echo "[runtime] pi CLI installed"
+RUN apk --no-cache add ca-certificates

-# Install browser-use CLI in runtime image
-RUN npm install -g browser-use --unsafe-perm \
-    && browser-use --version \
-    && echo "[runtime] browser-use CLI installed"
+WORKDIR /app
+COPY --from=builder /app/workflows .

-# Copy ast-grep binary from builder
-COPY --from=builder /usr/local/bin/sg /usr/local/bin/sg
-RUN chmod +x /usr/local/bin/sg && sg --version \
-    && echo "[runtime] ast-grep copied"
+EXPOSE 8080

-# Copy pi skills from builder
-COPY --from=builder /root/.pi /root/.pi
-RUN ls -la /root/.pi/agent/skills/ \
-    && echo "[runtime] pi skills configured"
+HEALTHCHECK --interval=10s --timeout=5s --start-period=10s --retries=3 \
+  CMD wget --no-verbose --tries=1 --spider http://localhost:8080/health || exit 1

-# Copy worker binary from builder
-COPY --from=builder /build/worker /app/worker
-RUN chmod +x /app/worker && file /app/worker \
-    && echo "[runtime] Worker binary copied"
-
-# Create app directory structure
-RUN mkdir -p /app/work /app/logs /app/screenshots \
-    && chmod 755 /app/work /app/logs /app/screenshots \
-    && echo "[runtime] App directories created"
-
-# Health check endpoint
-EXPOSE 8081
-
-# Worker task queue listener
-ENV TEMPORAL_NAMESPACE=poimen-harness \
-    TEMPORAL_HOSTPORT=temporal-frontend.temporal:7233 \
-    MEMORY_SERVICE_URL=http://memory-service.poimen:5000 \
-    MEMORY_SERVICE_TOKEN= \
-    ANTHROPIC_API_KEY= \
-    PI_SKILLS_PATH=/root/.pi/agent/skills \
-    AST_GREP_BIN=/usr/local/bin/sg \
-    BROWSER_USE_BIN=/usr/local/bin/browser-use \
-    CHROMIUM_BIN=/usr/bin/chromium-browser \
-    SCREENSHOTS_DIR=/app/screenshots
-
-# Entrypoint script with startup diagnostics
-COPY --chmod=755 << 'EOF' /app/entrypoint.sh
-#!/bin/bash
-set -e
-
-echo "[$(date)] ========== POIMEN WORKER STARTUP =========="
-echo "[$(date)] Container: $HOSTNAME"
-echo "[$(date)] Image: $(cat /etc/os-release | grep PRETTY_NAME | cut -d= -f2)"
-
-# Verify CLI tools
-echo "[$(date)] ✓ Checking CLI tools..."
-echo "  - Go version: $(go version 2>/dev/null || echo 'N/A')"
-echo "  - ast-grep: $(sg --version 2>&1 | head -1)"
-echo "  - pi: $(pi --version 2>&1 | head -1)"
-echo "  - browser-use: $(browser-use --version 2>&1 | head -1)"
-echo "  - chromium: $(chromium-browser --version 2>&1 || echo 'Not found')"
-echo "  - git: $(git --version)"
-echo "  - node: $(node --version)"
-echo "  - npm: $(npm --version)"
-
-# Verify pi skills
-echo "[$(date)] ✓ Checking pi skills..."
-if [ -d "$PI_SKILLS_PATH" ]; then
-    echo "  - Skills path: $PI_SKILLS_PATH"
-    ls -1 "$PI_SKILLS_PATH" | sed 's/^/    ✓ /'
-else
-    echo "  - WARNING: Skills path not found: $PI_SKILLS_PATH"
-fi
-
-# Verify browser tools
-echo "[$(date)] ✓ Checking browser automation tools..."
-echo "  - Chromium binary: $CHROMIUM_BIN"
-echo "  - Screenshots directory: $SCREENSHOTS_DIR"
-if [ -d "$SCREENSHOTS_DIR" ]; then
-    echo "  - Screenshots dir ready ($(du -sh $SCREENSHOTS_DIR 2>/dev/null | cut -f1 || echo '0B'))"
-fi
-
-# Check environment variables
-echo "[$(date)] ✓ Configuration loaded:"
-echo "  - TEMPORAL_NAMESPACE: $TEMPORAL_NAMESPACE"
-echo "  - TEMPORAL_HOSTPORT: $TEMPORAL_HOSTPORT"
-echo "  - MEMORY_SERVICE_URL: ${MEMORY_SERVICE_URL:-(not set)}"
-echo "  - PI_SKILLS_PATH: $PI_SKILLS_PATH"
-echo "  - CHROMIUM_BIN: $CHROMIUM_BIN"
-
-# Verify memory service connectivity (optional, non-blocking)
-if [ ! -z "$MEMORY_SERVICE_URL" ]; then
-    echo "[$(date)] ✓ Testing memory service connectivity..."
-    if curl -sf "$MEMORY_SERVICE_URL/health" > /dev/null 2>&1; then
-        echo "  - Memory service: HEALTHY"
-    else
-        echo "  - Memory service: UNREACHABLE (will retry in worker)"
-    fi
-fi
-
-# Test browser automation (optional, non-blocking)
-echo "[$(date)] ✓ Testing browser automation..."
-if command -v chromium-browser &> /dev/null && command -v browser-use &> /dev/null; then
-    echo "  - Chromium available: YES"
-    echo "  - browser-use available: YES"
-    echo "  - Browser automation: READY"
-else
-    echo "  - Browser automation: WARNING - missing dependencies"
-fi
-
-echo "[$(date)] ========== STARTING WORKER =========="
-exec /app/worker
-EOF
-
-RUN chmod +x /app/entrypoint.sh
-
-# Run worker with diagnostics
-ENTRYPOINT ["/app/entrypoint.sh"]
-
-# Health check
-HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
-    CMD curl -f http://localhost:8081/health || exit 1
+ENTRYPOINT ["./workflows"]
@@ -293,3 +293,62 @@ func knowledgeBaseData() string {
 	// For now, return empty - real implementation loads from file
 	return ""
 }
+
+// CanvasCompatibilityInput for Temporal activity
+type CanvasCompatibilityInput struct {
+	Nodes []db.WorkflowNode `json:"nodes"`
+	Edges []db.WorkflowEdge `json:"edges"`
+}
+
+// CanvasCompatibilityOutput returns validation results
+type CanvasCompatibilityOutput struct {
+	IsValid           bool                      `json:"is_valid"`
+	Incompatibilities []IncompatibilityWarning  `json:"incompatibilities"`
+	DisconnectedNodes []string                  `json:"disconnected_nodes"`
+	Warnings          []string                  `json:"warnings"`
+}
+
+// CanvasCompatibilityActivity validates workflow canvas for type mismatches and isolation
+func CanvasCompatibilityActivity(ctx interface{}, input CanvasCompatibilityInput) (CanvasCompatibilityOutput, error) {
+	output := CanvasCompatibilityOutput{
+		IsValid:           true,
+		Incompatibilities: []IncompatibilityWarning{},
+		DisconnectedNodes: []string{},
+		Warnings:          []string{},
+	}
+
+	// Check all edges for compatibility
+	for _, edge := range input.Edges {
+		var sourceNode, targetNode *db.WorkflowNode
+		for i := range input.Nodes {
+			if input.Nodes[i].ID == edge.Source {
+				sourceNode = &input.Nodes[i]
+			}
+			if input.Nodes[i].ID == edge.Target {
+				targetNode = &input.Nodes[i]
+			}
+		}
+
+		if sourceNode != nil && targetNode != nil {
+			if warning, err := ValidateConnection(sourceNode, targetNode); err != nil {
+				output.IsValid = false
+				output.Incompatibilities = append(output.Incompatibilities, warning)
+			}
+		}
+	}
+
+	// Find disconnected nodes
+	connected := make(map[string]bool)
+	for _, edge := range input.Edges {
+		connected[edge.Source] = true
+		connected[edge.Target] = true
+	}
+
+	for _, node := range input.Nodes {
+		if node.Type == "activity" && !connected[node.ID] {
+			output.DisconnectedNodes = append(output.DisconnectedNodes, node.ID)
+		}
+	}
+
+	return output, nil
+}
@@ -19,14 +19,34 @@ type CanvasReasonerInput struct {
 	AuthToken        string `json:"auth_token,omitempty"` // JWT for LLM calls
 }

+// RelationWording describes semantic meaning of an edge
+type RelationWording struct {
+	Verb               string `json:"verb"`                 // outputs, inputs, depends-on, etc
+	SourceOutput       string `json:"source_output"`        // What source produces
+	TargetInput        string `json:"target_input"`         // What target requires
+	ConnectionType     string `json:"connection_type"`      // direct-map, requires-transformer, conditional
+	Confidence         float64 `json:"confidence"`          // 0.0-1.0
+	SemanticMatch      string `json:"semantic_match"`       // Human-readable explanation
+	TransformerNeeded  string `json:"transformer_needed,omitempty"` // If transformation required
+}
+
+// EdgeWithWording pairs an edge with its semantic description
+type EdgeWithWording struct {
+	Source           string          `json:"source"`
+	Target           string          `json:"target"`
+	RelationType     string          `json:"relation_type"`      // data-flow, dependency, conditional, parallel
+	RelationLabel    string          `json:"relation_label"`     // e.g., "CloneRepo outputs path → AnalyzeCode requires path"
+	RelationWording  RelationWording `json:"relation_wording"`
+}
+
 // CanvasReasonerOutput returns suggested edges and reasoning
 type CanvasReasonerOutput struct {
-	SuggestedEdges        []db.WorkflowEdge             `json:"suggested_edges"`         // New edges to add
-	RemovedEdges          []db.WorkflowEdge             `json:"removed_edges,omitempty"` // Edges to remove (if redesign)
+	SuggestedEdges        []EdgeWithWording             `json:"suggested_edges"`         // Edges with wording
+	RemovedEdges          []db.WorkflowEdge             `json:"removed_edges,omitempty"` // Edges to remove
 	Reasoning             string                        `json:"reasoning"`               // LLM explanation
 	Confidence            float64                       `json:"confidence"`              // 0.0-1.0
-	IncompatibleEdges     []IncompatibilityWarning      `json:"incompatible_edges,omitempty"` // Edges that can't be created
-	DisconnectedNodes     []string                      `json:"disconnected_nodes,omitempty"` // Nodes with no connections
+	IncompatibleEdges     []IncompatibilityWarning      `json:"incompatible_edges,omitempty"` // Can't connect
+	DisconnectedNodes     []string                      `json:"disconnected_nodes,omitempty"` // No connections
 	UserAlerts            []string                      `json:"user_alerts,omitempty"`        // Human-readable warnings
 }

@@ -48,23 +68,34 @@ func CanvasReasonerActivity(ctx context.Context, in CanvasReasonerInput) (Canvas
 	nodeDesc := buildNodeDescriptions(in.Nodes)
 	edgeDesc := buildEdgeDescriptions(in.Edges)

-	// Create prompt for LLM reasoning with compatibility guidance
-	systemPrompt := `You are a workflow automation expert. Analyze the following activities and suggest logical connections (edges) between them based on:
-1. Activity input/output compatibility (CRITICAL - only connect if outputs match inputs)
-2. Logical execution order and data flow
-3. Required dependencies
-4. Common workflow patterns
+	// Create prompt for LLM reasoning with relation wording
+	systemPrompt := `You are a workflow automation expert. Analyze activities and suggest logical connections with semantic descriptions.

-IMPORTANT: Only suggest edges where:
- Source activity has outputs (check "outputs" fields)
- Target activity has inputs (check "inputs" fields)
- Data types are compatible (string→string, object→object, etc)
- Connection makes semantic sense (don't connect a notifier to an analyzer)
+CRITICAL RULES:
+1. Only suggest edges where outputs→inputs match
+2. Provide relation wording: verb, source_output, target_input
+3. Assess connection confidence (0.0-1.0)
+4. Flag type mismatches that need transformers

-Respond with JSON containing:
+Respond with JSON:
 {
-  "edges": [{"source": "node-1", "target": "node-2"}, ...],
-  "reasoning": "explanation of why these connections make sense and any type mismatches noted",
+  "edges": [
+    {
+      "source": "node-1",
+      "target": "node-2",
+      "relation_type": "data-flow|dependency|conditional|parallel",
+      "relation_label": "Node1 outputs X → Node2 requires X",
+      "relation_wording": {
+        "verb": "outputs|depends-on|triggers|etc",
+        "source_output": "field_name (type): description",
+        "target_input": "field_name (type, required?): description",
+        "connection_type": "direct-map|requires-transformer|conditional",
+        "confidence": 0.95,
+        "semantic_match": "Explanation of why this makes sense"
+      }
+    }
+  ],
+  "reasoning": "Overall workflow structure explanation",
  "confidence": 0.85
 }`

@@ -0,0 +1,96 @@
+package action
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"github.com/rockliang/poimen/workflows/pkg/db"
+)
+
+// CanvasWithRelationsData combines canvas nodes/edges with relation wording
+type CanvasWithRelationsData struct {
+	WorkflowID  string              `json:"workflow_id"`
+	Version     int                 `json:"version"`
+	Nodes       []db.WorkflowNode   `json:"nodes"`
+	Edges       []db.WorkflowEdge   `json:"edges"`
+	Relations   []EdgeWithWording   `json:"relations"`
+	UpdatedAt   string              `json:"updated_at"`
+}
+
+// FetchCanvasRelationsInput parameters
+type FetchCanvasRelationsInput struct {
+	WorkflowID string `json:"workflow_id"`
+	Version    int    `json:"version"`
+}
+
+// FetchCanvasRelationsActivity fetches canvas + relations from DB
+func FetchCanvasRelationsActivity(ctx context.Context, input FetchCanvasRelationsInput) (CanvasWithRelationsData, error) {
+	logger := newActivityLogger(ctx)
+	output := CanvasWithRelationsData{
+		WorkflowID: input.WorkflowID,
+		Version:    input.Version,
+		Nodes:      []db.WorkflowNode{},
+		Edges:      []db.WorkflowEdge{},
+		Relations:  []EdgeWithWording{},
+	}
+
+	logger.logf("info", "Fetching canvas relations: %s v%d", input.WorkflowID, input.Version)
+
+	// Get database client from context or activity manager
+	dbClient, ok := ctx.Value("db_client").(*db.Client)
+	if !ok {
+		return output, fmt.Errorf("database client not in context")
+	}
+
+	// Fetch workflow
+	workflow, err := dbClient.GetWorkflow(ctx, input.WorkflowID)
+	if err != nil {
+		return output, fmt.Errorf("failed to get workflow: %w", err)
+	}
+
+	// Parse canvas nodes and edges
+	var nodes []db.WorkflowNode
+	if err := json.Unmarshal([]byte(workflow.Nodes), &nodes); err != nil {
+		return output, fmt.Errorf("failed to parse nodes: %w", err)
+	}
+
+	var edges []db.WorkflowEdge
+	if err := json.Unmarshal([]byte(workflow.Edges), &edges); err != nil {
+		return output, fmt.Errorf("failed to parse edges: %w", err)
+	}
+
+	output.Nodes = nodes
+	output.Edges = edges
+	output.UpdatedAt = workflow.UpdatedAt.String()
+
+	// Fetch workflow relations
+	relations, err := dbClient.GetWorkflowRelations(ctx, input.WorkflowID, input.Version)
+	if err != nil {
+		// Relations may not exist for old canvases - this is OK
+		logger.logf("warn", "Failed to fetch relations: %v", err)
+		return output, nil
+	}
+
+	// Map to EdgeWithWording
+	for _, rel := range relations {
+		edge := EdgeWithWording{
+			ID:             rel.ID,
+			Source:         rel.SourceNodeID,
+			Target:         rel.TargetNodeID,
+			RelationType:   rel.RelationType,
+			RelationLabel:  rel.Label,
+			CreatedAt:      rel.CreatedAt.String(),
+		}
+
+		// Parse relation wording JSON
+		if err := json.Unmarshal(rel.RelationWording, &edge.RelationWording); err != nil {
+			logger.logf("warn", "Failed to parse relation wording: %v", err)
+		}
+
+		output.Relations = append(output.Relations, edge)
+	}
+
+	logger.logf("info", "Fetched %d nodes, %d edges, %d relations", len(output.Nodes), len(output.Edges), len(output.Relations))
+	return output, nil
+}
@@ -0,0 +1,78 @@
+package action
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"time"
+
+	"github.com/rockliang/poimen/workflows/pkg/db"
+)
+
+// IndexGraphRAGInput sends workflow relations to GraphRAG for indexing
+type IndexGraphRAGInput struct {
+	WorkflowID string            `json:"workflow_id"`
+	Version    int               `json:"version"`
+	Nodes      []db.WorkflowNode `json:"nodes"`
+	Relations  []EdgeWithWording `json:"relations"`
+}
+
+// IndexGraphRAGOutput confirms indexing status
+type IndexGraphRAGOutput struct {
+	WorkflowID       string `json:"workflow_id"`
+	Version          int    `json:"version"`
+	IndexedEntities  int    `json:"indexed_entities"`
+	IndexedEdges     int    `json:"indexed_edges"`
+	Status           string `json:"status"`
+	GraphRAGChecksum string `json:"graph_rag_checksum"`
+	IndexedAt        string `json:"indexed_at"`
+}
+
+// IndexGraphRAGActivity indexes workflow canvas to GraphRAG (stub for now)
+func IndexGraphRAGActivity(ctx context.Context, input IndexGraphRAGInput) (IndexGraphRAGOutput, error) {
+	output := IndexGraphRAGOutput{
+		WorkflowID:      input.WorkflowID,
+		Version:         input.Version,
+		Status:          "indexed",
+		IndexedEntities: len(input.Nodes),
+		IndexedEdges:    len(input.Relations),
+		IndexedAt:       time.Now().UTC().Format(time.RFC3339),
+	}
+
+	// Stub implementation - actual GraphRAG indexing would happen here
+	// For now, just return success
+	return output, nil
+}
+
+// QueryGraphRAGRelationsInput for direct relation discovery
+type QueryGraphRAGRelationsInput struct {
+	WorkflowID string                 `json:"workflow_id"`
+	Version    int                    `json:"version"`
+	Query      string                 `json:"query"`
+	TopK       int                    `json:"top_k"`
+	Filters    map[string]interface{} `json:"filters,omitempty"`
+}
+
+// QueryGraphRAGRelationsOutput returns discovered relations
+type QueryGraphRAGRelationsOutput struct {
+	Query       string            `json:"query"`
+	Results     []EdgeWithWording `json:"results"`
+	TotalCount  int               `json:"total_count"`
+	ExecutionMs int64             `json:"execution_time_ms"`
+}
+
+// QueryGraphRAGRelationsActivity queries GraphRAG for relation patterns (stub)
+func QueryGraphRAGRelationsActivity(ctx context.Context, input QueryGraphRAGRelationsInput) (QueryGraphRAGRelationsOutput, error) {
+	output := QueryGraphRAGRelationsOutput{
+		Query:      input.Query,
+		Results:    []EdgeWithWording{},
+		TotalCount: 0,
+	}
+
+	// Stub implementation - actual GraphRAG querying would happen here
+	return output, nil
+}
@@ -0,0 +1,133 @@
+package action
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"time"
+)
+
+// GraphRAGQueryInput for Memory System endpoint
+type GraphRAGQueryInput struct {
+	WorkflowID      string                    `json:"workflow_id"`
+	Query           string                    `json:"query"`
+	SearchType      string                    `json:"search_type"`
+	RelationType    string                    `json:"relation_type"`
+	ConfidenceFloor float64                   `json:"confidence_floor"`
+	TopK            int                       `json:"top_k"`
+	RankingProfile  string                    `json:"ranking_profile"`
+	Canvas          CanvasWithRelationsData   `json:"canvas"`
+}
+
+// GraphRAGQueryOutput from Memory System
+type GraphRAGQueryOutput struct {
+	WorkflowID   string             `json:"workflow_id"`
+	Query        string             `json:"query"`
+	Edges        []EdgeWithWording  `json:"edges"`
+	Paths        []QueryPathData    `json:"paths"`
+	TotalCount   int                `json:"total_count"`
+	HasMore      bool               `json:"has_more"`
+	ExecutionMs  int64              `json:"execution_time_ms"`
+}
+
+type QueryPathData struct {
+	SourceID   string   `json:"source_id"`
+	TargetID   string   `json:"target_id"`
+	Distance   int      `json:"distance"`
+	PathCount  int      `json:"path_count"`
+	NodeIDs    []string `json:"node_ids"`
+	Confidence float64  `json:"total_confidence"`
+}
+
+// QueryGraphRAGActivity queries Memory System for semantic relations
+func QueryGraphRAGActivity(ctx context.Context, input GraphRAGQueryInput) (GraphRAGQueryOutput, error) {
+	logger := newActivityLogger(ctx)
+	output := GraphRAGQueryOutput{
+		WorkflowID: input.WorkflowID,
+		Query:      input.Query,
+		Edges:      []EdgeWithWording{},
+		Paths:      []QueryPathData{},
+	}
+
+	logger.logf("info", "Querying GraphRAG: %s", input.Query)
+
+	// Get Memory Service URL from env
+	memoryURL := os.Getenv("MEMORY_SERVICE_URL")
+	if memoryURL == "" {
+		memoryURL = "http://localhost:8000"
+	}
+
+	// Build payload for Memory System
+	payload := map[string]interface{}{
+		"workflow_id":      input.WorkflowID,
+		"query":            input.Query,
+		"search_type":      input.SearchType,
+		"relation_type":    input.RelationType,
+		"confidence_floor": input.ConfidenceFloor,
+		"top_k":            input.TopK,
+		"ranking_profile":  input.RankingProfile,
+		"canvas_nodes":     input.Canvas.Nodes,
+		"canvas_edges":     input.Canvas.Edges,
+		"relations":        input.Canvas.Relations,
+	}
+
+	reqBody, err := json.Marshal(payload)
+	if err != nil {
+		return output, fmt.Errorf("failed to marshal payload: %w", err)
+	}
+
+	// Call Memory System unified query endpoint
+	req, err := http.NewRequestWithContext(
+		ctx,
+		"POST",
+		memoryURL+"/workflows/query",
+		bytes.NewReader(reqBody),
+	)
+	if err != nil {
+		return output, fmt.Errorf("failed to create request: %w", err)
+	}
+
+	req.Header.Set("Content-Type", "application/json")
+	if token := ctx.Value("jwt_token"); token != nil {
+		req.Header.Set("Authorization", fmt.Sprintf("Bearer %v", token))
+	}
+
+	startTime := time.Now()
+	client := &http.Client{Timeout: 30 * time.Second}
+	resp, err := client.Do(req)
+	if err != nil {
+		return output, fmt.Errorf("failed to call Memory Service: %w", err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != 200 {
+		body, _ := io.ReadAll(resp.Body)
+		return output, fmt.Errorf("Memory Service returned %d: %s", resp.StatusCode, string(body))
+	}
+
+	// Parse response
+	var graphResp struct {
+		Edges      []EdgeWithWording `json:"edges"`
+		Paths      []QueryPathData   `json:"paths"`
+		TotalCount int               `json:"total_count"`
+		HasMore    bool              `json:"has_more"`
+	}
+
+	if err := json.NewDecoder(resp.Body).Decode(&graphResp); err != nil {
+		return output, fmt.Errorf("failed to decode response: %w", err)
+	}
+
+	output.Edges = graphResp.Edges
+	output.Paths = graphResp.Paths
+	output.TotalCount = graphResp.TotalCount
+	output.HasMore = graphResp.HasMore
+	output.ExecutionMs = time.Since(startTime).Milliseconds()
+
+	logger.logf("info", "GraphRAG returned %d edges, %d paths in %dms", 
+		len(output.Edges), len(output.Paths), output.ExecutionMs)
+	return output, nil
+}
@@ -52,6 +52,7 @@ func main() {
 	w.RegisterWorkflow(statemachine.TaskUnitWorkflow)
 	w.RegisterWorkflow(statemachine.TestWorkflow)
 	w.RegisterWorkflow(statemachine.RoutingWorkflow)
+	w.RegisterWorkflow(statemachine.WorkflowGraphQuery)

 	// Register all activities
 	w.RegisterActivity(action.CloneRepoActivity)
@@ -91,6 +92,13 @@ func main() {
 	// Memory activities
 	w.RegisterActivity(action.RetrieveMemoryActivity)

+	// GraphRAG activities
+	w.RegisterActivity(action.FetchCanvasRelationsActivity)
+	w.RegisterActivity(action.QueryGraphRAGActivity)
+	w.RegisterActivity(action.CanvasReasonerActivity)
+	w.RegisterActivity(action.IndexGraphRAGActivity)
+	w.RegisterActivity(action.CanvasCompatibilityActivity)
+
 	// Initialize health checker
 	healthChecker := health.NewChecker(c)
 	healthHandler := health.NewHandler(healthChecker)
@@ -119,7 +127,7 @@ func main() {
 	// Run worker in a goroutine
 	workerErrChan := make(chan error, 1)
 	go func() {
-		logging.Info("starting worker on queue", logging.String("queue", "poimen-taskqueue"))
+		logging.Info("starting worker", logging.String("queue", "poimen"))
 		if err := w.Run(worker.InterruptCh()); err != nil {
 			workerErrChan <- err
 		}
@@ -1,322 +0,0 @@
-# Canvas Reasoner: Auto-Inferring Workflow Connections
-
-## Overview
-
-The **CanvasReasonerActivity** uses LLM reasoning to automatically suggest connections between workflow activities when users drop new nodes onto the canvas. It analyzes input/output compatibility and detects connection problems.
-
-## Connection Logic
-
-### How It Works
-
-1. **Analyze Node Schemas**
-   - Get each activity's input/output fields from knowledge base
-   - Activities are classified as:
-     - **Generators**: No inputs, has outputs (e.g., API call, trigger)
-     - **Processors**: Has inputs and outputs (e.g., analyze code, security scan)
-     - **Sinks/Terminals**: Has inputs, no outputs (e.g., notification, approval)
-
-2. **LLM Reasoning**
-   - Pass all nodes + their schemas to reasoning model
-   - Ask LLM to suggest edges based on:
-     - Type compatibility (string→string, object→object)
-     - Logical execution order
-     - Data flow requirements
-     - Common workflow patterns
-
-3. **Validate Suggestions**
-   - Check all suggested edges exist in node map
-   - Skip self-loops
-   - Remove duplicates
-
-4. **Compatibility Checking**
-   - For each suggested edge: `source → target`
-   - Verify source produces outputs
-   - Verify target accepts inputs
-   - Check output/input type compatibility
-   - Flag incompatible connections
-
-5. **Identify Issues**
-   - Collect all incompatible edges
-   - Identify disconnected nodes (no edges in/out)
-   - Generate user alerts for problems
-
-## Connection Impossibility Detection
-
-### Why Connections Fail
-
-1. **Missing Outputs**
-   ```
-   NotifyStatusActivity → AnalyzeCodeActivity
-   ⚠️ NotifyStatusActivity produces no outputs
-   Reason: Notification is terminal activity (sink)
-   Solution: Add an intermediate processor that has outputs
-   ```
-
-2. **Missing Inputs**
-   ```
-   CloneRepoActivity → ApproveWorkflowActivity
-   ⚠️ ApproveWorkflowActivity accepts no inputs
-   Reason: Approval is a terminal activity (sink)
-   Solution: ApproveWorkflowActivity only works as final step
-   ```
-
-3. **Type Mismatch**
-   ```
-   LLMInferenceActivity (output: string) → DeploymentPreCheckActivity (input: object)
-   ⚠️ String output cannot satisfy object input requirement
-   Reason: Incompatible data types
-   Solution: Use LLM transformation node to convert string→object
-   ```
-
-4. **Semantic Incompatibility**
-   ```
-   NotifyStatusActivity → CloneRepoActivity
-   ⚠️ No logical connection between these activities
-   Reason: Notification cannot be input to clone operation
-   Solution: Ensure data flow makes semantic sense
-   ```
-
-### Incompatibility Data Structure
-
-```json
-{
-  "incompatible_edges": [
-    {
-      "source": "node-1",
-      "target": "node-2",
-      "reason": "Source activity produces no outputs",
-      "source_needs": "any output",
-      "target_needs": "path, depth",
-      "suggestion": "Use LLM transformation to map outputs to inputs"
-    }
-  ],
-  "disconnected_nodes": ["node-5", "node-8"],
-  "user_alerts": [
-    "⚠️  node-1 → node-2: Source activity produces no outputs. Use LLM transformation to map outputs to inputs",
-    "🔌 Node 'NotifyStatus-1' has no connections. Consider adding edges or removing it."
-  ]
-}
-```
-
-## User Alerts
-
-### Alert Types
-
-1. **Incompatibility Warnings** (⚠️)
-   ```
-   ⚠️  source → target: reason. suggestion.
-   ```
-   - Highlighted in red on canvas
-   - Shows in error sidebar
-   - Prevents workflow execution until fixed
-
-2. **Disconnection Warnings** (🔌)
-   ```
-   🔌 Node 'label' has no connections. Consider adding edges or removing it.
-   ```
-   - Highlighted in yellow
-   - Nodes with no input/output edges
-   - May be valid (first step, last step) or indicate design error
-
-3. **Type Mismatch Info** (ℹ️)
-   ```
-   ℹ️  To connect source → target, use transformer to map: {source_outputs} → {target_inputs}
-   ```
-   - Suggestion to use intermediate LLM node
-   - Provides mapping information
-
-## Frontend Integration
-
-### Canvas UI Feedback
-
-When CanvasReasonerActivity returns incompatibilities:
-
-1. **Visual Markers**
-   - Incompatible suggested edges: ❌ red dashed line (don't auto-add)
-   - Disconnected nodes: ⚠️ yellow border
-
-2. **Sidebar Alerts**
-   ```
-   🚨 Connection Issues (3)
-   
-   ⚠️  CloneRepo → ApproveWorkflow
-      Reason: ApproveWorkflow is terminal (no outputs)
-      Suggestion: Place ApproveWorkflow at end of workflow
-   
-   ⚠️  LLMInference → DeploymentPreCheck
-      Reason: Type mismatch (string ≠ object)
-      Suggestion: Add LLM transformation node
-   
-   🔌 SecurityScan-1 has no incoming edges
-      Suggestion: Connect CloneRepo → SecurityScan
-   ```
-
-3. **User Actions**
-   - ✅ Accept suggestions (green edges)
-   - ❌ Reject incompatible edges
-   - 🔧 Add transformer nodes
-   - 🗑️ Remove disconnected nodes
-
-### API Response Example
-
-```json
-{
-  "suggested_edges": [
-    {"source": "clone-1", "target": "analyze-1"},
-    {"source": "analyze-1", "target": "security-1"},
-    {"source": "security-1", "target": "report-1"}
-  ],
-  "reasoning": "Standard code review workflow: clone → analyze → scan → report",
-  "confidence": 0.92,
-  "incompatible_edges": [
-    {
-      "source": "report-1",
-      "target": "approve-1",
-      "reason": "ReportGenerator has no outputs (terminal activity)",
-      "suggestion": "ApproveWorkflow can only be a final step"
-    }
-  ],
-  "disconnected_nodes": [],
-  "user_alerts": [
-    "⚠️  report-1 → approve-1: ReportGenerator has no outputs (terminal activity). ApproveWorkflow can only be a final step"
-  ]
-}
-```
-
-## Knowledge Base Schema
-
-Each activity in `activity_knowledge_base.json` defines:
-
-```json
-{
-  "name": "CloneRepoActivity",
-  "inputs": {
-    "repo": {"type": "string", "required": true},
-    "branch": {"type": "string", "required": false}
-  },
-  "outputs": {
-    "path": {"type": "string"},
-    "commit": {"type": "string"}
-  }
-}
-```
-
-### Classification Rules
-
- **Generator** (0 inputs): trigger, API call, schedule
- **Processor** (1+ inputs, 1+ outputs): analysis, transformation, scan
- **Sink** (1+ inputs, 0 outputs): notification, approval, archive
- **Bypass** (0 inputs, 0 outputs): rare - usually error
-
-## Common Patterns
-
-### ✅ Valid Chains
-
-```
-CloneRepo → Analyze → SecurityScan → Report
-(generator) → (processor) → (processor) → (sink)
-```
-
-```
-Trigger → LLMInference → Decision → (Branch: Notify OR Approve)
-(gen) → (processor) → (processor) → (sink)
-```
-
-### ❌ Invalid Chains
-
-```
-Notify → CloneRepo  ❌
-(sink) → (generator) - backward flow
-
-CloneRepo → CloneRepo → Analyze  ❌
-self-loop - no benefit
-
-Analyze → Approve → Notify  ❌
-Approve is terminal (sink), can't output to Notify
-```
-
-## Edge Cases
-
-### Multiple Outputs → Single Input
-```
-SecurityScan → Report
-SecurityScan outputs: [issues, metrics, severity]
-Report inputs: [report_data]
-
-LLM must infer: bundle all outputs into single report_data object
-Confidence: 0.7 (requires transformation)
-```
-
-### Terminal Activities
- **ApproveWorkflowActivity**: Must be last (blocks workflow)
- **NotifyStatusActivity**: Can be mid-workflow (async notify)
- **ArchiveResultsActivity**: Should be last (persistence)
-
-### Data Transformation
-When source outputs don't match target inputs:
-
-```python
-# User can insert transformer node:
-LLMInference → [LLMTransformer] → DeploymentPreCheck
-
-# Transformer:
-# - Input: LLMInference.output (string)
-# - Output: DeploymentPreCheck.requirements (object)
-# - Action: Call LLM to convert format
-```
-
-## Testing Incompatibility Detection
-
-### Test Case 1: Terminal Activity as Source
-```go
-source := db.WorkflowNode{ID: "n1", Type: "notify-status", Label: "Notify"}
-target := db.WorkflowNode{ID: "n2", Type: "clone-repo", Label: "Clone"}
-
-warnings := CheckConnectionCompatibility(source, target)
-// Should warn: NotifyStatusActivity produces no outputs
-```
-
-### Test Case 2: Type Mismatch
-```go
-source := db.WorkflowNode{ID: "n1", Type: "llm-inference", ...}
-target := db.WorkflowNode{ID: "n2", Type: "deployment-check", ...}
-
-warnings := CheckConnectionCompatibility(source, target)
-// Should warn: string output ≠ object input
-```
-
-### Test Case 3: Disconnected Node
-```go
-nodes := []db.WorkflowNode{n1, n2, n3}
-edges := []db.WorkflowEdge{{Source: "n1", Target: "n2"}}
-
-disconnected := IdentifyDisconnectedNodes(nodes, edges)
-// Should return ["n3"]
-```
-
-## Future Enhancements
-
-1. **Automatic Transformer Insertion**
-   - Detect incompatibilities
-   - Auto-suggest LLM transformer nodes
-   - Chain transformers if needed
-
-2. **Confidence Scoring**
-   - Increase when types match perfectly
-   - Decrease for semantic mismatches
-   - Factor in activity dependencies
-
-3. **Learning from History**
-   - Track successful workflows
-   - Remember user edits to suggestions
-   - Improve LLM prompts over time
-
-4. **Multi-Path Analysis**
-   - Suggest multiple connection topologies
-   - Show cost/efficiency of each
-   - Rank by execution time/cost
-
-5. **Dry-Run Validation**
-   - Execute suggested workflow in simulation
-   - Catch runtime errors early
-   - Show data flow through each node
@@ -74,6 +74,22 @@ func (s *Server) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 			s.api.ExecuteWorkflow(w, r, parts[2])
 		}

+	// GraphRAG query endpoint
+	case strings.HasSuffix(path, "/query") && method == http.MethodPost:
+		// POST /workflows/{id}/query
+		parts := strings.Split(path, "/")
+		if len(parts) >= 4 && parts[1] == "workflows" && parts[3] == "query" {
+			s.api.QueryWorkflowGraph(w, r, parts[2])
+		}
+
+	// Relation versions endpoint
+	case strings.Contains(path, "/relations/") && strings.Contains(path, "/versions") && method == http.MethodGet:
+		// GET /workflows/{id}/relations/{edge_id}/versions
+		parts := strings.Split(path, "/")
+		if len(parts) >= 6 && parts[1] == "workflows" && parts[3] == "relations" && parts[5] == "versions" {
+			s.api.GetWorkflowRelationVersions(w, r, parts[2], parts[4])
+		}
+
 	// Execution endpoints
 	case strings.HasPrefix(path, "/executions/") && method == http.MethodGet:
 		id := extractID(path, "/executions/")
@@ -587,3 +587,118 @@ func (api *WorkflowAPI) nodesToWorkflowSpec(wf *WorkflowResponse, inputs map[str

 	return spec
 }
+
+// QueryWorkflowGraph handles POST /workflows/{id}/query
+func (api *WorkflowAPI) QueryWorkflowGraph(w http.ResponseWriter, r *http.Request, workflowID string) {
+	ctx, cancel := context.WithTimeout(r.Context(), 60*time.Second)
+	defer cancel()
+
+	var req QueryWorkflowGraphRequest
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		http.Error(w, "Invalid request body", http.StatusBadRequest)
+		return
+	}
+
+	// Defaults
+	if req.SearchType == "" {
+		req.SearchType = "edges"
+	}
+	if req.ConfidenceFloor == 0 {
+		req.ConfidenceFloor = 0.5
+	}
+	if req.TopK == 0 {
+		req.TopK = 10
+	}
+	if req.MaxPathDepth == 0 {
+		req.MaxPathDepth = 3
+	}
+	if req.RankingProfile == "" {
+		req.RankingProfile = "default"
+	}
+
+	// Get latest version if not specified
+	if req.Version == 0 {
+		wf, err := api.db.GetWorkflow(ctx, workflowID)
+		if err != nil {
+			http.Error(w, "Workflow not found", http.StatusNotFound)
+			return
+		}
+		req.Version = wf.Version
+	}
+
+	// Call temporal workflow
+	run, err := api.temporalClient.ExecuteWorkflow(
+		ctx,
+		client.StartWorkflowOptions{
+			ID:        fmt.Sprintf("graph-query-%s-v%d", workflowID, req.Version),
+			TaskQueue: "poimen",
+		},
+		"WorkflowGraphQuery",
+		map[string]interface{}{
+			"workflow_id":       workflowID,
+			"query":             req.Query,
+			"search_type":       req.SearchType,
+			"relation_type":     req.RelationType,
+			"version":           req.Version,
+			"confidence_floor":  req.ConfidenceFloor,
+			"top_k":             req.TopK,
+			"find_paths":        req.FindPaths,
+			"target_node_id":    req.TargetNodeID,
+			"max_path_depth":    req.MaxPathDepth,
+			"ranking_profile":   req.RankingProfile,
+			"include_reasoning": req.IncludeReasoning,
+		},
+	)
+	if err != nil {
+		api.logger.Printf("Failed to start workflow: %v", err)
+		http.Error(w, "Failed to start query workflow", http.StatusInternalServerError)
+		return
+	}
+
+	var result map[string]interface{}
+	if err := run.Get(ctx, &result); err != nil {
+		api.logger.Printf("Workflow execution failed: %v", err)
+		http.Error(w, "Query execution failed", http.StatusInternalServerError)
+		return
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	json.NewEncoder(w).Encode(result)
+}
+
+// QueryWorkflowGraphRequest matches frontend payload
+type QueryWorkflowGraphRequest struct {
+	Query            string  `json:"query"`
+	SearchType       string  `json:"search_type"`
+	RelationType     string  `json:"relation_type"`
+	Version          int     `json:"version"`
+	ConfidenceFloor  float64 `json:"confidence_floor"`
+	TopK             int     `json:"top_k"`
+	FindPaths        bool    `json:"find_paths"`
+	TargetNodeID     string  `json:"target_node_id"`
+	MaxPathDepth     int     `json:"max_path_depth"`
+	RankingProfile   string  `json:"ranking_profile"`
+	IncludeReasoning bool    `json:"include_reasoning"`
+}
+
+// GetWorkflowRelationVersions handles GET /workflows/{id}/relations/{edge_id}/versions
+func (api *WorkflowAPI) GetWorkflowRelationVersions(w http.ResponseWriter, r *http.Request, workflowID, edgeID string) {
+	ctx, cancel := context.WithTimeout(r.Context(), 10*time.Second)
+	defer cancel()
+
+	// Query relation versions from DB
+	versions, err := api.db.GetRelationVersions(ctx, workflowID, edgeID)
+	if err != nil {
+		api.logger.Printf("Failed to get relation versions: %v", err)
+		http.Error(w, "Failed to fetch relation versions", http.StatusInternalServerError)
+		return
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	json.NewEncoder(w).Encode(map[string]interface{}{
+		"workflow_id": workflowID,
+		"edge_id":     edgeID,
+		"versions":    versions,
+		"total_count": len(versions),
+	})
+}
@@ -9,6 +9,6 @@ metadata:
    app.kubernetes.io/name: poimen
    app.kubernetes.io/component: orchestrator
 data:
-  GIT_COMMIT: "8cfa23e5d"  # Updated automatically by CI/CD
+  GIT_COMMIT: "84b4ca120"  # Updated automatically by CI/CD
  GIT_BRANCH: "main"
  DEPLOYMENT_DATE: "2026-09-05"
@@ -4,15 +4,19 @@ kind: Kustomization
 namespace: poimen

 resources:
-  - worker-deployment.yaml
-  - configmap.yaml
+  - poimen-application.yaml

 commonLabels:
  app.kubernetes.io/name: poimen
  app.kubernetes.io/component: worker

-secretGenerator:
-  - name: poimen-secrets
-    envs:
-      - secrets.env
-    behavior: create
+images:
+  - name: forgejo.riotpiao.com/rock/poimen-memory
+    newName: forgejo.riotpiao.com/rock/poimen-memory
+    newTag: latest
+  - name: forgejo.riotpiao.com/rock/poimen-workflows
+    newName: forgejo.riotpiao.com/rock/poimen-workflows
+    newTag: latest
+  - name: forgejo.riotpiao.com/rock/poimen-frontend
+    newName: forgejo.riotpiao.com/rock/poimen-frontend
+    newTag: latest
@@ -0,0 +1,283 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: poimen
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: poimen-registry
+  namespace: poimen
+type: kubernetes.io/dockercfg
+data:
+  .dockercfg: eyJmb3JnZWpvLnJpb3RwaWFvLmNvbSI6eyJhdXRoIjoiYmFzZTY0LWVuY29kZWQtY3JlZGVudGlhbHMifX0=
+---
+# Poimen Memory Service
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: poimen-memory
+  namespace: poimen
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: poimen-memory
+  template:
+    metadata:
+      labels:
+        app: poimen-memory
+    spec:
+      imagePullSecrets:
+        - name: poimen-registry
+      containers:
+        - name: memory
+          image: forgejo.riotpiao.com/rock/poimen-memory:latest
+          imagePullPolicy: Always
+          ports:
+            - containerPort: 8000
+          env:
+            - name: DATABASE_URL
+              valueFrom:
+                secretKeyRef:
+                  name: poimen-db-credentials
+                  key: memory-url
+            - name: REDIS_URL
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: redis-url
+            - name: JWT_SECRET
+              valueFrom:
+                secretKeyRef:
+                  name: poimen-secrets
+                  key: jwt-secret
+          livenessProbe:
+            httpGet:
+              path: /health
+              port: 8000
+            initialDelaySeconds: 10
+            periodSeconds: 10
+          readinessProbe:
+            httpGet:
+              path: /ready
+              port: 8000
+            initialDelaySeconds: 5
+            periodSeconds: 5
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: poimen-memory
+  namespace: poimen
+spec:
+  selector:
+    app: poimen-memory
+  ports:
+    - port: 8000
+      targetPort: 8000
+  type: ClusterIP
+---
+# Poimen Workflows Service
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: poimen-workflows
+  namespace: poimen
+spec:
+  replicas: 2
+  selector:
+    matchLabels:
+      app: poimen-workflows
+  template:
+    metadata:
+      labels:
+        app: poimen-workflows
+    spec:
+      imagePullSecrets:
+        - name: poimen-registry
+      containers:
+        - name: workflows-server
+          image: forgejo.riotpiao.com/rock/poimen-workflows:latest
+          imagePullPolicy: Always
+          command: ["/app/workflows", "server"]
+          ports:
+            - containerPort: 8080
+          env:
+            - name: DATABASE_URL
+              valueFrom:
+                secretKeyRef:
+                  name: poimen-db-credentials
+                  key: workflows-url
+            - name: TEMPORAL_HOST
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: temporal-host
+            - name: MEMORY_SERVICE_URL
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: memory-service-url
+            - name: JWT_SECRET
+              valueFrom:
+                secretKeyRef:
+                  name: poimen-secrets
+                  key: jwt-secret
+          livenessProbe:
+            httpGet:
+              path: /health
+              port: 8080
+            initialDelaySeconds: 10
+            periodSeconds: 10
+          readinessProbe:
+            httpGet:
+              path: /ready
+              port: 8080
+            initialDelaySeconds: 5
+            periodSeconds: 5
+        - name: workflows-worker
+          image: forgejo.riotpiao.com/rock/poimen-workflows:latest
+          imagePullPolicy: Always
+          command: ["/app/workflows", "worker"]
+          env:
+            - name: DATABASE_URL
+              valueFrom:
+                secretKeyRef:
+                  name: poimen-db-credentials
+                  key: workflows-url
+            - name: TEMPORAL_HOST
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: temporal-host
+            - name: MEMORY_SERVICE_URL
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: memory-service-url
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: poimen-workflows
+  namespace: poimen
+spec:
+  selector:
+    app: poimen-workflows
+  ports:
+    - port: 8080
+      targetPort: 8080
+  type: ClusterIP
+---
+# Poimen Frontend Service
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: poimen-frontend
+  namespace: poimen
+spec:
+  replicas: 2
+  selector:
+    matchLabels:
+      app: poimen-frontend
+  template:
+    metadata:
+      labels:
+        app: poimen-frontend
+    spec:
+      imagePullSecrets:
+        - name: poimen-registry
+      containers:
+        - name: frontend
+          image: forgejo.riotpiao.com/rock/poimen-frontend:latest
+          imagePullPolicy: Always
+          ports:
+            - containerPort: 3000
+          env:
+            - name: NEXT_PUBLIC_WORKFLOWS_API
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: workflows-api-url
+            - name: NEXT_PUBLIC_MEMORY_API
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: memory-api-url
+            - name: NEXT_PUBLIC_AUTH_URL
+              valueFrom:
+                configMapKeyRef:
+                  name: poimen-config
+                  key: auth-url
+            - name: OAUTH_CLIENT_ID
+              valueFrom:
+                secretKeyRef:
+                  name: poimen-secrets
+                  key: oauth-client-id
+            - name: OAUTH_CLIENT_SECRET
+              valueFrom:
+                secretKeyRef:
+                  name: poimen-secrets
+                  key: oauth-client-secret
+          livenessProbe:
+            httpGet:
+              path: /
+              port: 3000
+            initialDelaySeconds: 10
+            periodSeconds: 10
+          readinessProbe:
+            httpGet:
+              path: /
+              port: 3000
+            initialDelaySeconds: 5
+            periodSeconds: 5
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: poimen-frontend
+  namespace: poimen
+spec:
+  selector:
+    app: poimen-frontend
+  ports:
+    - port: 80
+      targetPort: 3000
+  type: LoadBalancer
+---
+# ConfigMap for shared configuration
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: poimen-config
+  namespace: poimen
+data:
+  temporal-host: "temporal:7233"
+  redis-url: "redis://redis:6379"
+  workflows-api-url: "http://poimen-workflows:8080"
+  memory-api-url: "http://poimen-memory:8000"
+  auth-url: "https://auth.riotpiao.com"
+  memory-service-url: "http://poimen-memory:8000"
+---
+# Secrets placeholder - should be created separately
+apiVersion: v1
+kind: Secret
+metadata:
+  name: poimen-db-credentials
+  namespace: poimen
+type: Opaque
+stringData:
+  memory-url: "postgresql://user:pass@postgres:5432/poimen_memory"
+  workflows-url: "postgresql://user:pass@postgres:5432/poimen_workflows"
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: poimen-secrets
+  namespace: poimen
+type: Opaque
+stringData:
+  jwt-secret: "your-jwt-secret-here"
+  oauth-client-id: "your-oauth-client-id"
+  oauth-client-secret: "your-oauth-client-secret"
@@ -13,7 +13,7 @@ spec:
      labels:
        app: poimen-worker
      annotations:
-        git-commit: "8cfa23e5d"  # ✅ Updated on each push, triggers rolling restart
+        git-commit: "84b4ca120"  # ✅ Updated on each push, triggers rolling restart
        deployment-date: "2026-09-05"
    spec:
      containers:
@@ -0,0 +1,192 @@
+-- Poimen Workflows Schema
+-- Tables: workflows, workflow_executions, execution_logs, workflow_memory_links
+
+CREATE EXTENSION IF NOT EXISTS "uuid-ossp";
+CREATE EXTENSION IF NOT EXISTS "vector";
+
+-- Workflows (canvas definitions)
+CREATE TABLE IF NOT EXISTS workflows (
+  id UUID PRIMARY KEY DEFAULT uuid_generate_v4(),
+  customer_id TEXT NOT NULL,
+  name TEXT NOT NULL,
+  description TEXT,
+  status TEXT NOT NULL CHECK (status IN ('draft', 'active', 'archived')) DEFAULT 'draft',
+  version INT NOT NULL DEFAULT 1,
+  
+  -- Canvas data
+  nodes JSONB NOT NULL DEFAULT '[]'::jsonb,  -- WorkflowNode[]
+  edges JSONB NOT NULL DEFAULT '[]'::jsonb,  -- WorkflowEdge[]
+  
+  -- Metadata
+  created_by TEXT NOT NULL,
+  created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  updated_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  last_executed_at TIMESTAMPTZ,
+  
+  CONSTRAINT workflow_name_per_customer UNIQUE (customer_id, name)
+);
+
+CREATE INDEX idx_workflows_customer ON workflows(customer_id);
+CREATE INDEX idx_workflows_status ON workflows(status);
+CREATE INDEX idx_workflows_created_at ON workflows(created_at DESC);
+
+-- Workflow executions (runs triggered by user)
+CREATE TABLE IF NOT EXISTS workflow_executions (
+  id UUID PRIMARY KEY DEFAULT uuid_generate_v4(),
+  workflow_id UUID NOT NULL REFERENCES workflows(id) ON DELETE CASCADE,
+  customer_id TEXT NOT NULL,
+  
+  -- Temporal details
+  temporal_id TEXT NOT NULL UNIQUE,  -- Temporal workflow execution ID
+  status TEXT NOT NULL CHECK (status IN ('pending', 'running', 'success', 'failed', 'cancelled')) DEFAULT 'pending',
+  
+  -- Input/Output
+  inputs JSONB NOT NULL,
+  outputs JSONB,
+  
+  -- Timing
+  started_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  completed_at TIMESTAMPTZ,
+  duration_ms INT,
+  
+  -- Error tracking
+  error_message TEXT,
+  error_count INT DEFAULT 0,
+  
+  CONSTRAINT duration_when_completed CHECK (
+    (status IN ('success', 'failed') AND completed_at IS NOT NULL) OR
+    (status IN ('pending', 'running', 'cancelled'))
+  )
+);
+
+CREATE INDEX idx_executions_workflow ON workflow_executions(workflow_id);
+CREATE INDEX idx_executions_customer ON workflow_executions(customer_id);
+CREATE INDEX idx_executions_status ON workflow_executions(status);
+CREATE INDEX idx_executions_temporal_id ON workflow_executions(temporal_id);
+CREATE INDEX idx_executions_started_at ON workflow_executions(started_at DESC);
+
+-- Execution logs (detailed activity logs)
+CREATE TABLE IF NOT EXISTS execution_logs (
+  id BIGSERIAL PRIMARY KEY,
+  execution_id UUID NOT NULL REFERENCES workflow_executions(id) ON DELETE CASCADE,
+  
+  -- Node/Activity info
+  node_id TEXT NOT NULL,  -- "activity-123" from canvas
+  activity_name TEXT NOT NULL,  -- "CloneRepo", "AnalyzeCode", etc.
+  
+  -- Log entry
+  level TEXT NOT NULL CHECK (level IN ('info', 'warn', 'error', 'debug')),
+  message TEXT NOT NULL,
+  metadata JSONB,  -- Arbitrary structured data (duration, result, etc.)
+  
+  -- Timing
+  logged_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  
+  CONSTRAINT log_order UNIQUE (execution_id, logged_at, id)
+);
+
+CREATE INDEX idx_logs_execution ON execution_logs(execution_id);
+CREATE INDEX idx_logs_node ON execution_logs(execution_id, node_id);
+CREATE INDEX idx_logs_level ON execution_logs(level);
+CREATE INDEX idx_logs_logged_at ON execution_logs(logged_at DESC);
+
+-- Memory links (connect executions to memory/lessons learned)
+CREATE TABLE IF NOT EXISTS workflow_memory_links (
+  execution_id UUID NOT NULL REFERENCES workflow_executions(id) ON DELETE CASCADE,
+  memory_node_sha TEXT NOT NULL,  -- SHA256 from memory.memory_node
+  relationship TEXT NOT NULL CHECK (relationship IN ('generated', 'used', 'learned', 'failed_on')),
+  
+  -- Context
+  created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  notes TEXT,
+  
+  PRIMARY KEY (execution_id, memory_node_sha, relationship)
+);
+
+CREATE INDEX idx_memory_links_memory_node ON workflow_memory_links(memory_node_sha);
+CREATE INDEX idx_memory_links_execution ON workflow_memory_links(execution_id);
+
+-- Activity execution trace (detailed per-activity metrics)
+CREATE TABLE IF NOT EXISTS activity_traces (
+  id BIGSERIAL PRIMARY KEY,
+  execution_id UUID NOT NULL REFERENCES workflow_executions(id) ON DELETE CASCADE,
+  node_id TEXT NOT NULL,
+  
+  -- Activity details
+  activity_name TEXT NOT NULL,
+  parameters JSONB NOT NULL,
+  result JSONB,
+  
+  -- Timing
+  started_at TIMESTAMPTZ NOT NULL,
+  completed_at TIMESTAMPTZ,
+  duration_ms INT,
+  
+  -- Retry info
+  attempt INT DEFAULT 1,
+  retry_reason TEXT,
+  
+  -- Status
+  status TEXT NOT NULL CHECK (status IN ('running', 'success', 'failed', 'skipped')),
+  error_message TEXT
+);
+
+CREATE INDEX idx_traces_execution ON activity_traces(execution_id);
+CREATE INDEX idx_traces_activity ON activity_traces(activity_name);
+CREATE INDEX idx_traces_status ON activity_traces(status);
+CREATE INDEX idx_traces_started_at ON activity_traces(started_at DESC);
+
+-- Workflow stats (materialized for fast dashboard queries)
+CREATE TABLE IF NOT EXISTS workflow_stats (
+  workflow_id UUID PRIMARY KEY REFERENCES workflows(id) ON DELETE CASCADE,
+  customer_id TEXT NOT NULL,
+  
+  total_runs INT DEFAULT 0,
+  successful_runs INT DEFAULT 0,
+  failed_runs INT DEFAULT 0,
+  
+  avg_duration_ms NUMERIC,
+  min_duration_ms INT,
+  max_duration_ms INT,
+  
+  last_30d_runs INT DEFAULT 0,
+  last_30d_success_rate NUMERIC,
+  
+  updated_at TIMESTAMPTZ DEFAULT now()
+);
+
+CREATE INDEX idx_stats_customer ON workflow_stats(customer_id);
+
+-- View: Recent executions with workflow context
+CREATE OR REPLACE VIEW v_recent_executions AS
+SELECT 
+  we.id,
+  we.workflow_id,
+  w.name as workflow_name,
+  we.customer_id,
+  we.status,
+  we.started_at,
+  we.completed_at,
+  we.duration_ms,
+  we.error_message,
+  (SELECT COUNT(*) FROM execution_logs WHERE execution_id = we.id) as log_count,
+  (SELECT COUNT(*) FROM activity_traces WHERE execution_id = we.id) as activity_count
+FROM workflow_executions we
+JOIN workflows w ON we.workflow_id = w.id
+ORDER BY we.started_at DESC;
+
+-- View: Execution timeline (for state machine visualization)
+CREATE OR REPLACE VIEW v_execution_timeline AS
+SELECT 
+  el.execution_id,
+  el.logged_at,
+  el.node_id,
+  el.activity_name,
+  el.level,
+  el.message,
+  at.duration_ms as activity_duration,
+  at.status as activity_status
+FROM execution_logs el
+LEFT JOIN activity_traces at ON el.execution_id = at.execution_id 
+  AND el.node_id = at.node_id
+ORDER BY el.execution_id, el.logged_at;
@@ -0,0 +1,190 @@
+-- Poimen Workflows schema
+-- Tables: workflows, workflow_executions, execution_logs, activity_traces, workflow_stats, workflow_memory_links
+-- Integrates with temporal workflow orchestrator and memory service
+
+-- Workflows (canvas definitions with JSONB nodes/edges)
+CREATE TABLE IF NOT EXISTS workflows (
+  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+  customer_id TEXT NOT NULL,
+  name TEXT NOT NULL,
+  description TEXT,
+  status TEXT NOT NULL CHECK (status IN ('draft', 'active', 'archived')) DEFAULT 'draft',
+  version INT NOT NULL DEFAULT 1,
+  
+  -- Canvas data (React Flow format)
+  nodes JSONB NOT NULL DEFAULT '[]'::jsonb,  -- WorkflowNode[]
+  edges JSONB NOT NULL DEFAULT '[]'::jsonb,  -- WorkflowEdge[]
+  
+  -- Metadata
+  created_by TEXT NOT NULL,
+  created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  updated_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  last_executed_at TIMESTAMPTZ,
+  
+  CONSTRAINT workflow_name_per_customer UNIQUE (customer_id, name)
+);
+
+CREATE INDEX idx_workflows_customer ON workflows(customer_id);
+CREATE INDEX idx_workflows_status ON workflows(status);
+CREATE INDEX idx_workflows_created_at ON workflows(created_at DESC);
+
+-- Workflow executions (runs triggered by user)
+CREATE TABLE IF NOT EXISTS workflow_executions (
+  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+  workflow_id UUID NOT NULL REFERENCES workflows(id) ON DELETE CASCADE,
+  customer_id TEXT NOT NULL,
+  
+  -- Temporal details
+  temporal_id TEXT NOT NULL UNIQUE,  -- Temporal workflow execution ID
+  status TEXT NOT NULL CHECK (status IN ('pending', 'running', 'success', 'failed', 'cancelled')) DEFAULT 'pending',
+  
+  -- Input/Output
+  inputs JSONB NOT NULL,
+  outputs JSONB,
+  
+  -- Timing
+  started_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  completed_at TIMESTAMPTZ,
+  duration_ms INT,
+  
+  -- Error tracking
+  error_message TEXT,
+  error_count INT DEFAULT 0,
+  
+  CONSTRAINT duration_when_completed CHECK (
+    (status IN ('success', 'failed') AND completed_at IS NOT NULL) OR
+    (status IN ('pending', 'running', 'cancelled'))
+  )
+);
+
+CREATE INDEX idx_executions_workflow ON workflow_executions(workflow_id);
+CREATE INDEX idx_executions_customer ON workflow_executions(customer_id);
+CREATE INDEX idx_executions_status ON workflow_executions(status);
+CREATE INDEX idx_executions_temporal_id ON workflow_executions(temporal_id);
+CREATE INDEX idx_executions_started_at ON workflow_executions(started_at DESC);
+
+-- Execution logs (detailed activity logs)
+CREATE TABLE IF NOT EXISTS execution_logs (
+  id BIGSERIAL PRIMARY KEY,
+  execution_id UUID NOT NULL REFERENCES workflow_executions(id) ON DELETE CASCADE,
+  
+  -- Node/Activity info
+  node_id TEXT NOT NULL,  -- "activity-123" from canvas
+  activity_name TEXT NOT NULL,  -- "CloneRepo", "AnalyzeCode", etc.
+  
+  -- Log entry
+  level TEXT NOT NULL CHECK (level IN ('info', 'warn', 'error', 'debug')),
+  message TEXT NOT NULL,
+  metadata JSONB,  -- Arbitrary structured data (duration, result, etc.)
+  
+  -- Timing
+  logged_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  
+  CONSTRAINT log_order UNIQUE (execution_id, logged_at, id)
+);
+
+CREATE INDEX idx_logs_execution ON execution_logs(execution_id);
+CREATE INDEX idx_logs_node ON execution_logs(execution_id, node_id);
+CREATE INDEX idx_logs_level ON execution_logs(level);
+CREATE INDEX idx_logs_logged_at ON execution_logs(logged_at DESC);
+
+-- Activity execution trace (detailed per-activity metrics)
+CREATE TABLE IF NOT EXISTS activity_traces (
+  id BIGSERIAL PRIMARY KEY,
+  execution_id UUID NOT NULL REFERENCES workflow_executions(id) ON DELETE CASCADE,
+  node_id TEXT NOT NULL,
+  
+  -- Activity details
+  activity_name TEXT NOT NULL,
+  parameters JSONB NOT NULL,
+  result JSONB,
+  
+  -- Timing
+  started_at TIMESTAMPTZ NOT NULL,
+  completed_at TIMESTAMPTZ,
+  duration_ms INT,
+  
+  -- Retry info
+  attempt INT DEFAULT 1,
+  retry_reason TEXT,
+  
+  -- Status
+  status TEXT NOT NULL CHECK (status IN ('running', 'success', 'failed', 'skipped')),
+  error_message TEXT
+);
+
+CREATE INDEX idx_traces_execution ON activity_traces(execution_id);
+CREATE INDEX idx_traces_activity ON activity_traces(activity_name);
+CREATE INDEX idx_traces_status ON activity_traces(status);
+CREATE INDEX idx_traces_started_at ON activity_traces(started_at DESC);
+
+-- Workflow stats (materialized for fast dashboard queries)
+CREATE TABLE IF NOT EXISTS workflow_stats (
+  workflow_id UUID PRIMARY KEY REFERENCES workflows(id) ON DELETE CASCADE,
+  customer_id TEXT NOT NULL,
+  
+  total_runs INT DEFAULT 0,
+  successful_runs INT DEFAULT 0,
+  failed_runs INT DEFAULT 0,
+  
+  avg_duration_ms NUMERIC,
+  min_duration_ms INT,
+  max_duration_ms INT,
+  
+  last_30d_runs INT DEFAULT 0,
+  last_30d_success_rate NUMERIC,
+  
+  updated_at TIMESTAMPTZ DEFAULT now()
+);
+
+CREATE INDEX idx_stats_customer ON workflow_stats(customer_id);
+
+-- Memory links (connect executions to memory/lessons learned)
+CREATE TABLE IF NOT EXISTS workflow_memory_links (
+  execution_id UUID NOT NULL REFERENCES workflow_executions(id) ON DELETE CASCADE,
+  memory_node_sha TEXT NOT NULL REFERENCES memory_node(sha256) ON DELETE CASCADE,
+  relationship TEXT NOT NULL CHECK (relationship IN ('generated', 'used', 'learned', 'failed_on')),
+  
+  -- Context
+  created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  notes TEXT,
+  
+  PRIMARY KEY (execution_id, memory_node_sha, relationship)
+);
+
+CREATE INDEX idx_memory_links_memory_node ON workflow_memory_links(memory_node_sha);
+CREATE INDEX idx_memory_links_execution ON workflow_memory_links(execution_id);
+
+-- View: Recent executions with workflow context
+CREATE OR REPLACE VIEW v_recent_executions AS
+SELECT 
+  we.id,
+  we.workflow_id,
+  w.name as workflow_name,
+  we.customer_id,
+  we.status,
+  we.started_at,
+  we.completed_at,
+  we.duration_ms,
+  we.error_message,
+  (SELECT COUNT(*) FROM execution_logs WHERE execution_id = we.id) as log_count,
+  (SELECT COUNT(*) FROM activity_traces WHERE execution_id = we.id) as activity_count
+FROM workflow_executions we
+JOIN workflows w ON we.workflow_id = w.id
+ORDER BY we.started_at DESC;
+
+-- View: Execution timeline (for state machine visualization)
+CREATE OR REPLACE VIEW v_execution_timeline AS
+SELECT 
+  el.execution_id,
+  el.logged_at,
+  el.node_id,
+  el.activity_name,
+  el.level,
+  el.message,
+  at.duration_ms as activity_duration,
+  at.status as activity_status
+FROM execution_logs el
+LEFT JOIN activity_traces at ON el.execution_id = at.execution_id 
+  AND el.node_id = at.node_id
+ORDER BY el.execution_id, el.logged_at;
@@ -0,0 +1,55 @@
+-- Workflow Relations Schema
+-- Stores semantic relations between workflow canvas nodes with versioning support
+
+-- Table for storing workflow relations with relation wording
+CREATE TABLE IF NOT EXISTS workflow_relations (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    workflow_id VARCHAR(255) NOT NULL,
+    version INTEGER NOT NULL,
+    source_node_id VARCHAR(255) NOT NULL,
+    target_node_id VARCHAR(255) NOT NULL,
+    relation_type VARCHAR(50) NOT NULL, -- data-flow, dependency, conditional, parallel
+    label TEXT NOT NULL,
+    relation_wording JSONB NOT NULL, -- {verb, source_output, target_input, connection_type, confidence, semantic_match}
+    metadata JSONB,
+    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    FOREIGN KEY (workflow_id) REFERENCES workflows(id) ON DELETE CASCADE,
+    UNIQUE(workflow_id, source_node_id, target_node_id, version)
+);
+
+-- Table for versioned history of relation changes
+CREATE TABLE IF NOT EXISTS workflow_relation_versions (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    workflow_id VARCHAR(255) NOT NULL,
+    edge_id VARCHAR(255) NOT NULL,
+    version_num INTEGER NOT NULL,
+    operation VARCHAR(10) NOT NULL, -- CREATE, UPDATE, DELETE
+    snapshot JSONB NOT NULL,
+    changed_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    changed_by VARCHAR(255),
+    fields_changed JSONB, -- ["field1", "field2"]
+    FOREIGN KEY (workflow_id) REFERENCES workflows(id) ON DELETE CASCADE,
+    UNIQUE(workflow_id, edge_id, version_num)
+);
+
+-- Table for GraphRAG indexing metadata
+CREATE TABLE IF NOT EXISTS workflow_rag_index (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    workflow_id VARCHAR(255) NOT NULL UNIQUE,
+    version INTEGER NOT NULL,
+    indexed_status VARCHAR(20) NOT NULL DEFAULT 'pending', -- pending, indexed, partial, failed
+    embedding_model VARCHAR(100),
+    last_indexed_at TIMESTAMP,
+    index_metadata JSONB,
+    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    FOREIGN KEY (workflow_id) REFERENCES workflows(id) ON DELETE CASCADE
+);
+
+-- Indexes for performance
+CREATE INDEX idx_workflow_relations_workflow_version ON workflow_relations(workflow_id, version);
+CREATE INDEX idx_workflow_relations_type ON workflow_relations(relation_type);
+CREATE INDEX idx_workflow_relations_nodes ON workflow_relations(source_node_id, target_node_id);
+CREATE INDEX idx_workflow_relation_versions_workflow_edge ON workflow_relation_versions(workflow_id, edge_id);
+CREATE INDEX idx_workflow_relation_versions_operation ON workflow_relation_versions(operation);
+CREATE INDEX idx_workflow_rag_index_status ON workflow_rag_index(indexed_status);
@@ -454,3 +454,82 @@ func (db *DB) FetchActivityTraces(ctx context.Context, executionID string) ([]Ac

 	return traces, rows.Err()
 }
+
+// GetWorkflowRelations retrieves all relations for a workflow version
+func (db *DB) GetWorkflowRelations(ctx context.Context, workflowID string, version int) ([]WorkflowRelation, error) {
+	var relations []WorkflowRelation
+
+	query := `
+		SELECT id, workflow_id, version, source_node_id, target_node_id, 
+		       relation_type, label, relation_wording, metadata, created_at
+		FROM workflow_relations
+		WHERE workflow_id = $1 AND version = $2
+		ORDER BY created_at DESC
+	`
+
+	rows, err := db.conn.QueryContext(ctx, query, workflowID, version)
+	if err != nil {
+		return nil, fmt.Errorf("failed to query relations: %w", err)
+	}
+	defer rows.Close()
+
+	for rows.Next() {
+		var rel WorkflowRelation
+		if err := rows.Scan(
+			&rel.ID,
+			&rel.WorkflowID,
+			&rel.Version,
+			&rel.SourceNodeID,
+			&rel.TargetNodeID,
+			&rel.RelationType,
+			&rel.Label,
+			&rel.RelationWording,
+			&rel.Metadata,
+			&rel.CreatedAt,
+		); err != nil {
+			return nil, fmt.Errorf("failed to scan relation: %w", err)
+		}
+		relations = append(relations, rel)
+	}
+
+	return relations, rows.Err()
+}
+
+// GetRelationVersions retrieves version history for a specific relation
+func (db *DB) GetRelationVersions(ctx context.Context, workflowID string, edgeID string) ([]WorkflowRelationVersion, error) {
+	var versions []WorkflowRelationVersion
+
+	query := `
+		SELECT id, workflow_id, edge_id, version_num, operation, snapshot, 
+		       changed_at, changed_by, fields_changed
+		FROM workflow_relation_versions
+		WHERE workflow_id = $1 AND edge_id = $2
+		ORDER BY version_num ASC
+	`
+
+	rows, err := db.conn.QueryContext(ctx, query, workflowID, edgeID)
+	if err != nil {
+		return nil, fmt.Errorf("failed to query relation versions: %w", err)
+	}
+	defer rows.Close()
+
+	for rows.Next() {
+		var v WorkflowRelationVersion
+		if err := rows.Scan(
+			&v.ID,
+			&v.WorkflowID,
+			&v.EdgeID,
+			&v.VersionNum,
+			&v.Operation,
+			&v.Snapshot,
+			&v.ChangedAt,
+			&v.ChangedBy,
+			&v.FieldsChanged,
+		); err != nil {
+			return nil, fmt.Errorf("failed to scan version: %w", err)
+		}
+		versions = append(versions, v)
+	}
+
+	return versions, rows.Err()
+}
@@ -111,3 +111,30 @@ type WorkflowMemoryLink struct {
 	CreatedAt     time.Time `db:"created_at"`
 	Notes         string    `db:"notes"`
 }
+
+// WorkflowRelation represents a semantic relation between two canvas nodes
+type WorkflowRelation struct {
+	ID               string    `db:"id"`
+	WorkflowID       string    `db:"workflow_id"`
+	Version          int       `db:"version"`
+	SourceNodeID     string    `db:"source_node_id"`
+	TargetNodeID     string    `db:"target_node_id"`
+	RelationType     string    `db:"relation_type"` // "data-flow", "dependency", "conditional"
+	Label            string    `db:"label"`         // Human-readable relation description
+	RelationWording  []byte    `db:"relation_wording"`  // JSONB with verb, outputs, inputs, confidence
+	Metadata         []byte    `db:"metadata"`      // JSONB for extensibility
+	CreatedAt        time.Time `db:"created_at"`
+}
+
+// WorkflowRelationVersion represents versioned history of relation changes
+type WorkflowRelationVersion struct {
+	ID            string    `db:"id"`
+	WorkflowID    string    `db:"workflow_id"`
+	EdgeID        string    `db:"edge_id"`
+	VersionNum    int       `db:"version_num"`
+	Operation     string    `db:"operation"` // "CREATE", "UPDATE", "DELETE"
+	Snapshot      []byte    `db:"snapshot"` // JSONB full state at this version
+	ChangedAt     time.Time `db:"changed_at"`
+	ChangedBy     string    `db:"changed_by"`
+	FieldsChanged []byte    `db:"fields_changed"` // JSONB array of changed field names
+}
@@ -0,0 +1,106 @@
+package statemachine
+
+import (
+	"context"
+	"time"
+
+	"go.temporal.io/sdk/workflow"
+	"github.com/rockliang/poimen/workflows/action"
+	"github.com/rockliang/poimen/workflows/pkg/db"
+)
+
+type WorkflowGraphQueryInput struct {
+	WorkflowID      string  `json:"workflow_id"`
+	Query           string  `json:"query"`
+	SearchType      string  `json:"search_type"`
+	RelationType    string  `json:"relation_type"`
+	Version         int     `json:"version"`
+	ConfidenceFloor float64 `json:"confidence_floor"`
+	TopK            int     `json:"top_k"`
+	FindPaths       bool    `json:"find_paths"`
+	TargetNodeID    string  `json:"target_node_id"`
+	MaxPathDepth    int     `json:"max_path_depth"`
+	RankingProfile  string  `json:"ranking_profile"`
+	IncludeReasoning bool   `json:"include_reasoning"`
+}
+
+type WorkflowGraphQueryOutput struct {
+	WorkflowID      string                `json:"workflow_id"`
+	Query           string                `json:"query"`
+	Version         int                   `json:"version"`
+	ExecutionTimeMs int64                 `json:"execution_time_ms"`
+	Results         []action.EdgeWithWording `json:"results"`
+	Paths           []QueryPath           `json:"paths"`
+	TotalCount      int                   `json:"total_count"`
+	HasMore         bool                  `json:"has_more"`
+	RankingProfile  string                `json:"ranking_profile"`
+}
+
+type QueryPath struct {
+	SourceID   string   `json:"source_id"`
+	TargetID   string   `json:"target_id"`
+	Distance   int      `json:"distance"`
+	PathCount  int      `json:"path_count"`
+	NodeIDs    []string `json:"node_ids"`
+	Confidence float64  `json:"total_confidence"`
+}
+
+func WorkflowGraphQuery(ctx workflow.Context, input WorkflowGraphQueryInput) (WorkflowGraphQueryOutput, error) {
+	startTime := time.Now()
+	output := WorkflowGraphQueryOutput{
+		WorkflowID:     input.WorkflowID,
+		Query:          input.Query,
+		Version:        input.Version,
+		RankingProfile: input.RankingProfile,
+		Results:        []action.EdgeWithWording{},
+		Paths:          []QueryPath{},
+	}
+
+	opts := workflow.ActivityOptions{
+		StartToCloseTimeout: 120 * time.Second,
+		RetryPolicy: &workflow.RetryPolicy{
+			InitialInterval:    2 * time.Second,
+			BackoffCoefficient: 2.0,
+			MaxInterval:        10 * time.Second,
+			MaxAttempts:        3,
+		},
+	}
+	ctx = workflow.WithActivityOptions(ctx, opts)
+
+	// Fetch canvas + relations
+	var canvasData action.CanvasWithRelationsData
+	err := workflow.ExecuteActivity(ctx, action.FetchCanvasRelationsActivity,
+		action.FetchCanvasRelationsInput{
+			WorkflowID: input.WorkflowID,
+			Version:    input.Version,
+		},
+	).Get(ctx, &canvasData)
+	if err != nil {
+		return output, err
+	}
+
+	// Query Memory System via unified endpoint
+	var graphResults action.GraphRAGQueryOutput
+	err = workflow.ExecuteActivity(ctx, action.QueryGraphRAGActivity,
+		action.GraphRAGQueryInput{
+			WorkflowID:      input.WorkflowID,
+			Query:           input.Query,
+			SearchType:      input.SearchType,
+			RelationType:    input.RelationType,
+			ConfidenceFloor: input.ConfidenceFloor,
+			TopK:            input.TopK,
+			RankingProfile:  input.RankingProfile,
+			Canvas:          canvasData,
+		},
+	).Get(ctx, &graphResults)
+	if err != nil {
+		return output, err
+	}
+
+	output.Results = graphResults.Edges
+	output.TotalCount = graphResults.TotalCount
+	output.HasMore = graphResults.HasMore
+
+	output.ExecutionTimeMs = time.Since(startTime).Milliseconds()
+	return output, nil
+}
Author	SHA1	Message	Date
Test	58237cc1ff	feat: migration for workflow relations and RAG indexing ci / test (push) Failing after 2m9s Details	2026-09-05 06:01:08 -07:00
Test	5c0eb2b66a	feat: wire GraphRAG API handlers, activities, and database layer ci / test (push) Failing after 2m9s Details	2026-09-05 06:00:58 -07:00
Test	8474d7e494	chore: remove docker-compose (use k8s + CI/CD only) ci / test (push) Failing after 2m15s Details	2026-09-05 05:58:53 -07:00
Test	447951daca	feat: GraphRAG query API handlers and activities ci / test (push) Failing after 2m6s Details	2026-09-05 05:58:18 -07:00
Test	da44923c5c	docs: deployment guide for unified Poimen application ci / test (push) Failing after 2m7s Details	2026-09-05 05:57:34 -07:00
Test	70a9b9a2ab	feat: unified Poimen application with k8s + docker-compose infrastructure ci / test (push) Failing after 2m4s Details	2026-09-05 05:57:08 -07:00
Test	e01dad4e8c	feat: GraphRAG query workflow and indexing ci / test (push) Failing after 2m5s Details	2026-09-05 05:52:56 -07:00
Test	84b4ca120f	feat: add relation wording schema ci / test (push) Failing after 2m24s Details	2026-09-05 05:45:47 -07:00
Test	a461e9799a	docs: temporal + graph RAG integration with unified query	2026-09-05 05:45:21 -07:00