Configuration Reference

Consolidated environment variables, configuration patterns, and secret management for the LLM Platform ecosystem.

Linkable sections - READMEs can reference specific anchors (e.g., configuration-reference#gitlab-configuration).

---

Table of Contents

1. Ecosystem Configuration
2. Common Environment Variables
3. Service URLs
4. Database Configuration
5. Observability Stack
6. AI Provider Configuration
7. GitLab Configuration
8. GitHub Configuration
9. Security Configuration
10. Token Management
11. Kubernetes Configuration
12. Docker Compose Patterns
13. Project-Specific Templates
14. Environment Hierarchy

---

Ecosystem Configuration

The ecosystem configuration (~/.env.ecosystem or /Users/flux423/Sites/LLM/.env.ecosystem) provides shared settings across all platform projects.

Loading Ecosystem Config

All projects should source the ecosystem configuration at the start of their .env files:

# Source ecosystem configuration
source /Users/flux423/Sites/LLM/.env.ecosystem

# Local overrides are automatically loaded from .env.local

Core Ecosystem Variables

---

Common Environment Variables

Variables used across most or all services.

Application Runtime

API Configuration

WebSocket Configuration

GraphQL Configuration

---

Service URLs

Platform service endpoints using domain names.

Agent Services

URL Format Examples

# Development (local domain)
AGENT_PROTOCOL_URL=http://api.local.blueflyagents.com:3001
AGENT_TRACER_URL=http://tracer.local.blueflyagents.com:3002

# Production (Kubernetes DNS)
AGENT_PROTOCOL_URL=http://agent-protocol.ossa-agents.svc.cluster.local:3001

---

Database Configuration

PostgreSQL

Multiple PostgreSQL instances for service isolation:

URL Format:

POSTGRES_<SERVICE>_URL=postgresql://<user>:<password>@<host>:<port>/<database>

# Example
POSTGRES_LLM_URL=postgresql://postgres:postgres@postgres-llm-platform.local:5433/llm_platform

Individual Variables (alternative):

POSTGRES_HOST=localhost
POSTGRES_PORT=5432
POSTGRES_USER=postgres
POSTGRES_PASSWORD=secure-password
POSTGRES_DB=agent_tracer

Redis

URL Format:

REDIS_<SERVICE>_URL=redis://<host>:<port>

# With password
REDIS_<SERVICE>_URL=redis://:<password>@<host>:<port>

MongoDB

URL Format:

MONGODB_URL=mongodb://<host>:<port>/<database>

Vector Databases

Graph Databases

ClickHouse

For high-volume analytics and time-series data:

---

Observability Stack

OpenTelemetry

Phoenix AI Observability

Prometheus & Grafana

Additional Observability

RPC Tracing

---

AI Provider Configuration

OpenAI

Anthropic

Ollama (Local)

MLflow

Langflow

---

GitLab Configuration

Core GitLab Settings

GitLab Tokens

Bot Service Account Tokens

Kubernetes Agent Tokens

---

GitHub Configuration

For GitHub mirroring and cross-platform operations:

---

Security Configuration

Authentication

Security Settings

Rate Limiting

---

Token Management

Token Storage Location

All tokens are stored in ~/.tokens/ directory:

~/.tokens/
  gitlab            # GitLab personal access token
  gitlab-npm        # GitLab NPM registry token
  openai            # OpenAI API key
  anthropic         # Anthropic API key
  github            # GitHub personal access token
  posthog           # PostHog API key
  signoz            # SigNoz API key

Token Format

Each token file contains only the raw token value (no formatting):

# Read token in scripts
export GITLAB_TOKEN=$(cat ~/.tokens/gitlab)
export OPENAI_API_KEY=$(cat ~/.tokens/openai)

Token Prefixes

Security Rules

1. NEVER commit tokens to git repositories
2. NEVER expose tokens in terminal output or logs
3. ALWAYS use environment variables or token files
4. Rotate tokens quarterly or immediately if exposed
5. Use project access tokens with minimal permissions for automation

---

Kubernetes Configuration

In-Cluster Service DNS

Format: <service>.<namespace>.svc.cluster.local:<port>

# Observability namespace
NEO4J_URI=bolt://neo4j.observability.svc.cluster.local:7687
OTEL_EXPORTER_OTLP_ENDPOINT=http://otel-collector.observability.svc.cluster.local:4318

# Phoenix namespace
PHOENIX_ENDPOINT=http://phoenix.phoenix.svc.cluster.local:6006
PHOENIX_OTLP_GRPC=http://phoenix.phoenix.svc.cluster.local:4317

# OSSA-agents namespace
PROMETHEUS_URL=http://prometheus-service.ossa-agents.svc.cluster.local:9090
GRAFANA_URL=http://grafana-service.ossa-agents.svc.cluster.local:3000
QDRANT_URL=http://qdrant-service.ossa-agents.svc.cluster.local:6333

Ingress URLs (Browser Access)

NEO4J_BROWSER_URL=http://neo4j.local.bluefly.io
GRAFANA_BROWSER_URL=http://grafana.local.bluefly.io
PROMETHEUS_BROWSER_URL=http://prometheus.local.bluefly.io
QDRANT_BROWSER_URL=http://qdrant.local.bluefly.io

ConfigMap Pattern

apiVersion: v1
kind: ConfigMap
metadata:
  name: agent-config
  namespace: ossa-agents
data:
  NODE_ENV: "production"
  LOG_LEVEL: "info"
  METRICS_ENABLED: "true"
  # Database URLs (secrets referenced separately)
  POSTGRES_HOST: "postgres.ossa-agents.svc.cluster.local"
  POSTGRES_PORT: "5432"
  REDIS_HOST: "redis.ossa-agents.svc.cluster.local"
  REDIS_PORT: "6379"

Secret Reference Pattern

apiVersion: v1
kind: Pod
spec:
  containers:
    - name: app
      envFrom:
        - configMapRef:
            name: agent-config
        - secretRef:
            name: agent-secrets
      env:
        - name: POSTGRES_PASSWORD
          valueFrom:
            secretKeyRef:
              name: postgres-credentials
              key: password

---

Docker Compose Patterns

Network Configuration

networks:
  llm_front:
    external: true
    name: llm_front
  llm_back:
    external: true
    name: llm_back
  internal:
    driver: bridge

Service with Token Mount

services:
  api:
    environment:
      - NODE_ENV=production
      - PORT=3003
    volumes:
      - ~/.tokens:/root/.tokens:ro  # Mount tokens read-only

Database Connection Pattern

services:
  app:
    environment:
      - POSTGRES_URL=postgresql://user:password@postgres:5432/db
      - REDIS_URL=redis://redis:6379
      - MONGODB_URL=mongodb://mongo:27017/db
    depends_on:
      - postgres
      - redis
      - mongo

Health Check Pattern

services:
  api:
    healthcheck:
      test: ['CMD', 'wget', '--quiet', '--tries=1', '--spider', 'http://localhost:3000/health']
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s

---

Project-Specific Templates

Drupal Module Template

Source: .env.example.drupal

# Source ecosystem configuration
source /Users/flux423/Sites/LLM/.env.ecosystem

# Project Configuration
export PROJECT_NAME=my_module
export PROJECT_VERSION=11.x-1.0
export PROJECT_TYPE=drupal

# GitLab CI Component
export GITLAB_CI_COMPONENT="${GITLAB_COMPONENTS_REPO}/drupal/module@${GITLAB_COMPONENTS_VERSION}"

# Drupal Configuration
export DRUPAL_VERSION="${DRUPAL_VERSION}"  # From ecosystem (11.x)
export PHP_VERSION="${PHP_VERSION}"        # From ecosystem (8.3)

# Code Quality
export PHPCS_STANDARD=Drupal
export ENABLE_PHPSTAN=true

# Testing
export ENABLE_TESTS=true
export ENABLE_COVERAGE=true
export COVERAGE_THRESHOLD=70

NPM Package Template

Source: .env.example.npm

# Source ecosystem configuration
source /Users/flux423/Sites/LLM/.env.ecosystem

# Project Configuration
export PROJECT_NAME=agent-brain
export PROJECT_VERSION=0.1.0
export PROJECT_TYPE=npm

# GitLab CI Component
export GITLAB_CI_COMPONENT="${GITLAB_COMPONENTS_REPO}/workflow/golden@${GITLAB_COMPONENTS_VERSION}"

# Build Configuration
export NODE_VERSION=20
export ENABLE_TESTS=true
export ENABLE_COVERAGE=true
export COVERAGE_THRESHOLD=70

# NPM Publishing
export NPM_PUBLISH_REGISTRY="${NPM_REGISTRY}"
export NPM_PACKAGE_SCOPE="${NPM_SCOPE}"

ML Model Template

Source: .env.example.model

# Source ecosystem configuration
source /Users/flux423/Sites/LLM/.env.ecosystem

# Project Configuration
export PROJECT_NAME=llm-platform_model
export PROJECT_VERSION=0.1.0
export PROJECT_TYPE=model

# Model Registry
export MODEL_REGISTRY_URL="${GITLAB_HOST}/${GITLAB_GROUP}"
export MODEL_NAME=llm-platform-model
export MODEL_VERSION=1.0.0

# ML Configuration
export MLFLOW_TRACKING_URI="${MLFLOW_TRACKING_URI}"
export MLFLOW_EXPERIMENT_NAME="${PROJECT_NAME}"

# Vector Database
export QDRANT_HOST="${QDRANT_HOST}"
export QDRANT_PORT="${QDRANT_PORT}"
export QDRANT_COLLECTION="${PROJECT_NAME}_embeddings"

# Training
export TRAINING_DATA_PATH="./data"
export MODEL_OUTPUT_PATH="./models/trained"

---

Environment Hierarchy

Configuration sources are loaded in order (later sources override earlier):

1. .env.ecosystem     (shared platform defaults)
2. .env.local         (local overrides, gitignored)
3. .env               (project defaults, committed)
4. .env.production    (production overrides)
5. .env.k8s           (Kubernetes-specific)
6. Environment vars   (runtime overrides)

File Locations

Loading Pattern

#!/bin/bash
# Load configuration in correct order

# 1. Ecosystem defaults
[ -f "/Users/flux423/Sites/LLM/.env.ecosystem" ] && source "/Users/flux423/Sites/LLM/.env.ecosystem"

# 2. Local overrides (secrets)
[ -f "/Users/flux423/Sites/LLM/.env.local" ] && source "/Users/flux423/Sites/LLM/.env.local"

# 3. Project defaults
[ -f ".env" ] && source ".env"

# 4. Environment-specific
[ -f ".env.${NODE_ENV}" ] && source ".env.${NODE_ENV}"

---

Feature Flags & Thresholds

Observability Features

Alert Thresholds

Data Retention

Resource Limits

---

Integration Configuration

Discord

MCP Registry

OpenStandardAgents.org

Tailscale

---

Migration Settings

For repository migration operations:

---

Advanced Settings

Flash Attention (ML)

Tree of Thought / Graph of Thought

Worktree Configuration

---

See Also

• Resource URLs - External service URLs and documentation links
• Security and Compliance - Security policies and practices
• Deployment and Operations - Deployment guides
• Getting Started - Initial setup guide