Deployment Guide

Installing & Running Tharior Remedai

Deploy the Tharior Remedai Agentic Gateway on your local machine, via Docker, or onto an enterprise Kubernetes cluster with KEDA and HPA operators.

1. Prerequisites

Python: Version 3.10+ (3.11 recommended)
Anvesh: Vector & Knowledge Graph Engine (runs in embedded or remote mode)
Kubernetes: (Optional) v1.28+ for HPA & KEDA deployments
Bifrost: Local or Docker Bifrost on port 8080 (inference gateway)
OpenRouter key (optional): Catalog & weekly pricing sync — provider keys for inference belong in Bifrost

2. Install & start (recommended)

Use the repo scripts (no Makefile required):

bash — Clone, install, run

git clone https://github.com/vaagatech/tharior-remedai.git
cd tharior-remedai

./scripts/install.sh          # venv, npm, Bifrost prefetch, .env template
./scripts/start_local.sh      # Bifrost :8080, API :8000, Web Desk :5173
./scripts/verify-platform-apis.sh

# Stop everything:
./scripts/stop_local.sh

3. Configure Bifrost & catalog

Open Bifrost UI at http://127.0.0.1:8080 and add providers (OpenRouter, Ollama, vLLM, etc.).
Create a Bifrost virtual key (sk-bf-…).
In Web Desk → Bifrost LLM Router, save URL, tenant, and virtual key.
In Settings / Model Catalog, set OpenRouter key for live tier matrix sync (inference still flows through Bifrost).

Simulated / offline data

If the gateway is down or catalog has not loaded, the UI shows an amber Simulated / offline data badge. Platform APIs return _simulated: false when served live from the gateway.

4. Docker Compose

bash — Full stack in containers

./scripts/docker-up.sh
# Docs static site: host port 8081 (Bifrost uses 8080)
./scripts/docker-down.sh

5. Deploying to Kubernetes (HPA & KEDA)

Deploy to Kubernetes with automated Horizontal Pod Autoscaling and KEDA event-driven scaling:

bash — Kubernetes Deployment

# 1. Apply Deployment & Config
kubectl apply -f k8s/configmap.yaml
kubectl apply -f k8s/deployment.yaml
kubectl apply -f k8s/service.yaml

# 2. Apply HPA (CPU/Memory scaling) & KEDA (Workload queue scaling)
kubectl apply -f k8s/hpa.yaml
kubectl apply -f k8s/keda-scaledobject.yaml
kubectl apply -f k8s/keda-scaledjob.yaml