Local workspace quickstart

The fastest path to the unified Inferix stack (console + LensAI + TraceForge + RouteIQ + DriftWatch) is the multi-repo workspace orchestrator — not the legacy LensAI-only lensai-integration quickstart.

Prerequisites

• Docker Desktop with 6+ GB RAM allocated (--full profile)
• Git clones of sibling repos next to inferix/ (see inferix REPO_ARCHITECTURE.md)

Minimum siblings for ./run-local.sh:

• infra-ai-streaming
• inferix-ui
• distributed-flagd

Full profile also needs traceforge, fallback-chain, driftwatch-launch, and related libs (listed in the architecture doc).

One command

From the inferix/ repo in your workspace:

./run-local.sh

This prints the startup banner and execs the workspace orchestrator:

../run-inferix.sh --full

Equivalent from workspace root:

./run-inferix.sh --full          # all product slices
./run-inferix.sh --minimal       # LensAI + console only (~3.3 GB compose limits)
./run-inferix.sh --full --smoke  # full stack + smoke-e2e.sh

After startup

Seed demo data (inference events, routing_log, judge_scores, spans):

./inferix/scripts/seed-inferix-demo.sh

Validate the cross-product business path:

SEED=1 ./inferix/scripts/workflow-e2e.sh

macOS / RAM notes

On an 8 GB MacBook, prefer --minimal locally and run --full on CI or a 16 GB machine. Compose mem_limit sums are documented in inferix TESTING_STRATEGY.md.