How deepgram/self-hosted-resources works
This repository is a resource collection for deploying and operating Deepgram in self-hosted environments, rather than an application server. Operators choose a Kubernetes Helm chart, Docker Compose, or Podman Compose deployment; the chart describes API, engine, billing, and licensing components. Separate resources support troubleshooting, monitoring, and load/latency benchmarking. The benchmark scripts exercise HTTP or WebSocket endpoints and collect or plot metrics. Deployment internals and monitoring wiring are mostly unsampled, so those connections are shown as unverified rather than asserted.
Main entry points: deepgram-self-hosted Kubernetes chart · k6tts.js TTS tests · common Service configurations · k6streaming.js Streaming tests
Components 22
Deployment resources
- Kubernetes chart
charts/deepgram-self-hosted - Docker Compose
docker - Podman Compose
podman - Service configurations
common
Self-hosted services
- API service
charts/deepgram-self-hosted/templates/api - Inference engine
charts/deepgram-self-hosted/templates/engine - Billing service
charts/deepgram-self-hosted/templates/billing - License proxy
charts/deepgram-self-hosted/templates/license-proxy - Model volumes
charts/deepgram-self-hosted/templates/volumes
Operations
- Troubleshooting tools
diagnostics - Monitoring resources
monitoring
Benchmarking
- Benchmark suites
benchmarking - HTTP batch tests
benchmarking/batch/k6batch.js - SITS tests
benchmarking/batch/sits/k6batch_sits.js - Throughput calculation
benchmarking/batch/sits/calculate_k6_throughput.py - TTS tests
benchmarking/tts/k6tts.js - Streaming tests
benchmarking/streaming/k6streaming.js - Streaming stages
benchmarking/streaming/streamingstages.js - Latency plots
benchmarking/tts/plot_tts_latency.py - Throughput plots
benchmarking/tts/plot_tts_throughput.py
Other components
- Operator
- Self-hosted API endpoint
Connections 19
- Operator → Kubernetes chart: deploys with
README.md - Operator → Docker Compose: deploys with
README.md - Operator → Podman Compose: deploys with
README.md - Operator → Troubleshooting tools: troubleshoots with
README.md - Kubernetes chart → API service: configures
- Kubernetes chart → Inference engine: configures
- Kubernetes chart → Billing service: configures
- Kubernetes chart → License proxy: configures
- Kubernetes chart → Model volumes: mounts
- Service configurations → API service: configures
- Service configurations → Inference engine: configures
- HTTP batch tests → Self-hosted API endpoint: posts audio
benchmarking/batch/k6batch.js - SITS tests → Self-hosted API endpoint: posts audio URL
benchmarking/batch/sits/k6batch_sits.js - TTS tests → Self-hosted API endpoint: posts text
benchmarking/tts/k6tts.js - Streaming tests → Self-hosted API endpoint: streams audio
benchmarking/streaming/k6streaming.js - Streaming tests → Streaming stages: imports stages
benchmarking/streaming/k6streaming.js - Throughput calculation → SITS tests: analyzes output
- Latency plots → TTS tests: plots metrics
- Throughput plots → TTS tests: plots metrics