Skip to content

README.md

Latest commit

 

History

History
228 lines (169 loc) · 8.19 KB

README.md

File metadata and controls

228 lines (169 loc) · 8.19 KB

Langflow Agent Template

A template for building, testing, and deploying Langflow-based AI agents on OpenShift. Provides identical stacks for local development and cluster deployment, with CLI tooling to move flows between environments.

Architecture

LOCAL (podman-compose)                    CLUSTER (Helm + OpenShift)
┌──────────────────────────┐              ┌──────────────────────────┐
│  Langflow UI  :7860      │              │  Langflow (Deployment)   │
│         │                │              │         │                │
│         ▼                │   flows/     │         ▼                │
│  Ollama :11434           │ ◄──────────► │  vLLM + KServe           │
│  (local LLM)             │   (Git)      │  (GPU model serving)     │
│         │                │              │         │                │
│         ▼                │              │         ▼                │
│  PostgreSQL :5432        │              │  PostgreSQL (Red Hat)    │
│         │                │              │         │                │
│         ▼                │              │         ▼                │
│  Langfuse :3000          │              │  Langfuse (Deployment)   │
│  (tracing)               │              │  (tracing)               │
└──────────────────────────┘              └──────────────────────────┘

The flows/ directory is the Git-trackable hub. Flows are exported as portable JSON files that can be synced between local and cluster environments.

Project Structure

langflow-agent-template/
├── scripts/
│   └── agentctl                   # CLI tool — all operations go through this
│
├── local/
│   ├── podman-compose.yml         # Langflow + PostgreSQL + Ollama + Langfuse
│   ├── init-db.sh                 # Creates langfuse database on first boot
│   └── .env.example               # Environment variables
│
├── helm/langflow-agent/
│   ├── Chart.yaml                 # Helm chart metadata
│   ├── values.yaml                # All configurable values
│   └── templates/
│       ├── langflow.yaml          # Langflow Deployment + Service + Route
│       ├── postgresql.yaml        # PostgreSQL Deployment + Secret + PVC
│       ├── langfuse.yaml          # Langfuse Deployment + Service + Route
│       ├── langfuse-secret.yaml   # Shared credentials (Langfuse <-> Langflow)
│       └── charts/model-serving/  # vLLM + KServe subchart (disabled by default)
│
└── flows/                         # Portable flow JSON files
    ├── demo-flow-v1.json
    └── ...

Prerequisites

  • Local: Podman + podman-compose (auto-installed by agentctl on macOS/Linux if missing)
  • Cluster: oc CLI + helm CLI + access to an OpenShift cluster with RHOAI installed

Setup

Add agentctl to your PATH:

export PATH="$PATH:$(pwd)/scripts"

Start Podman Machine

The Podman machine must be running before using agentctl local-up. If it's not already started:

podman machine start

To check its status:

podman machine list

Cluster Login

To get the oc login command for your cluster:

  1. Open the OpenShift web console in your browser
  2. Click your username in the top-right corner
  3. Click Copy login command
  4. Click Display Token
  5. Copy the oc login command and run it in your terminal:
oc login --token=sha256~XXXX --server=https://api.your-cluster.example.com:6443

Workflows

Local-First

Build and test your agent locally, then push to the cluster.

# 1. Start local environment (Langflow + PostgreSQL + Ollama + Langfuse)
agentctl local-up

# 2. Build your agent flow in the Langflow UI
open http://localhost:7860

# 3. Run the flow and verify traces in Langfuse
open http://localhost:3000    # Login: admin@langflow.local / admin123

# 4. Save flows from local Langflow to the flows/ directory
agentctl flows save

# 5. Commit flows to Git
git add flows/ && git commit -m "Add agent flow"

# 6. Deploy the full stack to OpenShift
oc login https://your-cluster:6443
agentctl deploy

# 7. Push flows to the cluster Langflow instance
agentctl flows push

# 8. Open the flow in cluster Langflow and verify it works
#    (URL printed by agentctl deploy)

# Cleanup
agentctl destroy              # Remove cluster resources
agentctl local-down           # Stop local environment

Cloud-First

Build your agent on the cluster, then pull to local for iteration.

# 1. Deploy the full stack to OpenShift
oc login https://your-cluster:6443
agentctl deploy

# 2. Build your agent flow in the cluster Langflow UI
#    (URL printed by agentctl deploy)

# 3. Run the flow and verify traces in cluster Langfuse

# 4. Pull flows from cluster Langflow to the flows/ directory
agentctl flows pull

# 5. Commit flows to Git
git add flows/ && git commit -m "Add agent flow"

# 6. Start local environment for iteration
agentctl local-up

# 7. Load flows into local Langflow
agentctl flows load

# 8. Open the flow in local Langflow
open http://localhost:7860

# Cleanup
agentctl destroy              # Remove cluster resources
agentctl local-down           # Stop local environment

CLI Reference

All operations go through agentctl:

Command Description
agentctl local-up Start local dev environment
agentctl local-down [--force] Stop local environment (--force cleans stuck containers)
agentctl deploy [--namespace ns] Deploy full stack to OpenShift
agentctl destroy [--namespace ns] Remove all cluster resources
agentctl flows save Local Langflow → flows/ directory
agentctl flows load flows/ directory → Local Langflow
agentctl flows pull [-n ns] Cluster Langflow → flows/ directory
agentctl flows push [-n ns] flows/ directory → Cluster Langflow
agentctl build <flow.json> [registry] [tag] Build flow into a container image and push to registry
agentctl list [--all-namespaces] List deployed agents
agentctl status <name> [-n ns] Show agent status and metadata

Configuration

Helm Values

Key values in helm/langflow-agent/values.yaml:

Parameter Description Default
langflow.image Langflow container image langflowai/langflow:1.7.1
langflow.replicas Number of Langflow replicas 1
langfuse.enabled Deploy Langfuse for tracing true
modelServing.enabled Deploy vLLM + KServe false
modelServing.modelName Model to serve meta-llama/Llama-3.1-8B-Instruct
modelServing.gpu.count GPUs for model serving 1

Deploy with Model Serving

agentctl deploy --no-model-serving    # Without GPU model serving (default)

To enable model serving, set modelServing.enabled: true in values.yaml. Requires a GPU node and KServe/RHOAI on the cluster.

Default Credentials

Service Username Password
Langflow (auto-login)
Langfuse (local) admin@langflow.local admin123
Langfuse (cluster) admin@langflow.local admin123
PostgreSQL langflow langflow

Services & Ports (Local)

Service URL
Langflow UI http://localhost:7860
Langfuse http://localhost:3000
Ollama API http://localhost:11434

Notes

  • Flows pulled from the cluster may contain model components pointing to cluster-internal URLs. When loading these locally, update the model endpoint in the Langflow UI to point to Ollama (http://ollama:11434/v1).
  • flows save and flows pull download all flows including Langflow's built-in starter templates. Only user-created flows are relevant for version control.
  • Langfuse auto-provisioning (org, project, API keys) only runs on first database creation. If you need to reset, remove the PostgreSQL volume and restart.