Quick start

The runnable path today is LensAI: clone the integration repo, build and start the stack, run smoke, open Grafana. About ten minutes on a machine with Docker.

Unified stack (recommended for Inferix console + closed loop)

Clone the inferix orchestrator repo with sibling product repos in one workspace, then run the canonical entrypoint from the workspace root:

# workspace layout: inferix/, inferix-lensai-streaming/, inferix-ui/, …
./run-inferix.sh --full          # build + start full profile
./run-inferix.sh --full --smoke  # optional E2E smoke after startup

• Console: http://localhost:3003
• Grafana (LensAI): http://localhost:3000
• Script source of truth: inferix/run-inferix.sh (workspace root run-inferix.sh delegates to the same compose files)

Legacy LensAI-only quickstart below remains valid for Grafana-first exploration.

1 · Clone lensai-integration

Primary path uses lensai-integration. Streaming and eBPF sources live in sibling repos linked from the suite map.

git clone https://github.com/AkshantVats/lensai-integration.git
cd lensai-integration

• lensai-integration — unified Compose quickstart (this path)
• infra-ai-streaming — Rust ingest, Go consumer, distributed-flagd, Helm
• inferix — suite overview and product repo map

2 · Build, up, smoke

make build    # builds app images from source (~5 min first run)
make up       # starts Redpanda, ClickHouse, ingest, flagd, Grafana, …
make smoke    # POSTs a synthetic event; exits 0 when ClickHouse has it

Smoke hits ingest on :8080, waits for the consumer flush, then queries ClickHouse. Override hosts with INGEST_HOST / INGEST_PORT / CH_HOST if needed — see the integration README.

3 · Open Grafana

Open http://localhost:3000 (admin / admin). The LensAI — Overview dashboard auto-provisions.

• Ingest HTTP — :8080 (events into the pipeline)
• distributed-flagd — :8081 (HTTP) / :50051 (gRPC)
• Grafana — :3000
• Prometheus — :9090

What you just ran

eBPF tracer (mock by default) → Rust ingest (HTTP)
        → Redpanda → Go consumer → ClickHouse → Grafana
              ↕
    distributed-flagd (etcd-backed)

Real eBPF needs Linux 5.8+ and CAP_BPF — set EBPF_MOCK=false and uncomment privileged mounts in the Compose file. Not supported on macOS/Windows.

Deeper deploy (Compose profiles, Helm, k3d): Install & self-host. Maturity of every product: Status & maturity.