flowstate-ai-gateway Package Skill

Package: @epicdm/flowstate-ai-gateway Repository: epic-flowstate-community Path: packages/flowstate-ai-gateway Version: 0.1.0

Use this published Dojo skill as the package-level entrypoint. For detailed workflows, use the package-local skills listed below.

Primary Package Workflow

FlowState AI Gateway Deployment Integration

Use this skill for deploying or wiring the runtime gateway.

Source Map

• docker/Dockerfile.ai-gateway: image build.
• docker/docker-compose.ai.yml and compose templates: service env and network exposure.
• packages/gateway-routes/src/routes.ts: Kong exposure.
• packages/flowstate-cli/src/cli-commands/agent/ai-usage.ts: CLI usage client.
• src/server.ts and src/config.ts: runtime env behavior.

Deployment Model

• The service listens in-container and should not expose the LLM proxy directly on a host port.
• Host/operator access to /usage and /healthz should go through Kong when deployed.
• Sidecars use service-network URLs to send LLM traffic.
• Runtime env controls provider selection, credentials, ledger path, budgets, identity HMAC, and payload audit.

Guardrails

• Do not expose /v1/messages, /v1/chat/completions, or /v1/responses as a public unauthenticated proxy.
• Keep /usage and /healthz exposure aligned with Kong route policy.
• Do not put API keys or HMAC secrets in generated files that are intended for commit.
• Coordinate compose env changes with CLI instance templates and gateway routes.

Verification

yarn nx build @epicdm/flowstate-ai-gateway
yarn workspace @epicdm/flowstate-ai-gateway typecheck

For deployment changes, run the smallest local compose/smoke path that exercises the changed wiring.

Package Skill Inventory

Skill Inventory: @epicdm/flowstate-ai-gateway

Package path: packages/flowstate-ai-gateway.

Skillset

Composition Notes

• Use flowstate-ai-gateway-provider-routing together with flowstate-ai-gateway-trusted-identity for per-agent OpenAI routing.
• Use flowstate-ai-gateway-usage-budget-ledger before changing budget enforcement or /usage.
• Use flowstate-ai-gateway-deployment-integration before changing exposed routes or host ports.

Publication Readiness

Status: drafted, pending verification.

Before Dojo publication, rebuild package Dojo artifacts from this reviewed skillset.

Review Gate

Package Skill Review: @epicdm/flowstate-ai-gateway

Package path: packages/flowstate-ai-gateway.

Review Result

Status: drafted, source-backed, verified with one package-script lint blocker, pending publication.

The skillset is workflow-shaped around app/runtime operation, provider routing, usage/budget ledger behavior, trusted identity, deployment integration, and focused tests.

Checks

• Source review: completed from package review agent output, package metadata, root exports, app/config/server/key resolver/sqlite store files, Docker/Kong/CLI references, tests, and generated artifact drift.
• Skill inventory: completed.
• Skill files: completed.
• Verification: completed.
• Generated Dojo replacement: pending.
• Cloud Dojo publication: pending.

Verification Commands

Commands run:

yarn nx build @epicdm/flowstate-ai-gateway
yarn nx test @epicdm/flowstate-ai-gateway
yarn workspace @epicdm/flowstate-ai-gateway typecheck
yarn workspace @epicdm/flowstate-ai-gateway lint
yarn nx build @epicdm/flowstate-ai-gateway-core

Results:

• Skill file ASCII scan passed.
• Tests passed: 3 suites, 59 tests; Jest reported an open-handle warning after completion.
• Typecheck passed.
• Nx build passed for flowstate-ai-gateway and dependency flowstate-ai-gateway-core; Nx Cloud reported the existing unconnected-workspace 401.
• Lint did not run because the package script resolved eslint as missing: command not found: eslint.

Publication Gate

Do not publish existing .flowstate/dojo artifacts until they are rebuilt from reviewed package skills.

Start with this skill to understand the package purpose, then use the package-local focused skills for specific workflows. Keep this Dojo version aligned with packages/flowstate-ai-gateway/package.json. Do not treat generated feature lists as lessons; use workflow evidence from CODE-REVIEW.md, SKILL-INVENTORY.md, and REVIEW.md.