flowstate-dojo-community-epicdm-flowstate-ai-gateway

Use this skill when a task touches packages/flowstate-ai-gateway, imports @epicdm/flowstate-ai-gateway, or needs to teach humans/agents how this package works.

Package identity:

• Package: @epicdm/flowstate-ai-gateway
• Repository: epic-flowstate-community
• Version authority: package.json (0.1.0)
• Dojo course: flowstate-dojo-community-epicdm-flowstate-ai-gateway

How to use the package:

• Purpose: Local containerized AI Gateway — tracks token usage + cost for every LLM call from FlowState agent sidecars.
• Start with package-local docs when present.
• Verify details against source and tests before changing behavior.
• Use the Dojo course lessons as the teaching path: setup, mental model, core workflow, integration, feature labs, maintenance.

Public surfaces discovered:

• AppDeps
• buildApp
• GatewayConfig
• loadConfig
• buildOpenaiKeyResolver
• SqliteLedgerStore

Composable feature skilllets:

• Build App: use feature labs for workflow, source anchor, and regression risk.
• Build Openai Key Resolver: use feature labs for workflow, source anchor, and regression risk.
• GET /healthz: use feature labs for workflow, source anchor, and regression risk.
• GET /usage: use feature labs for workflow, source anchor, and regression risk.
• Load Config: use feature labs for workflow, source anchor, and regression risk.
• POST /chat/completions: use feature labs for workflow, source anchor, and regression risk.
• POST /messages: use feature labs for workflow, source anchor, and regression risk.
• POST /responses: use feature labs for workflow, source anchor, and regression risk.
• POST /v1/chat/completions: use feature labs for workflow, source anchor, and regression risk.
• POST /v1/messages: use feature labs for workflow, source anchor, and regression risk.
• POST /v1/responses: use feature labs for workflow, source anchor, and regression risk.
• Sqlite Ledger Store: use feature labs for workflow, source anchor, and regression risk.

Agent operating rules:

• Treat package.json as the Dojo version authority.
• Do not turn feature inventory into user-facing lessons without workflow context.
• Update this Dojo package when package behavior, docs, examples, or feature inventory changes.
• Verify downstream packages when package use and connection metadata shows inbound dependencies or references.
• Never mark publish status passed without successful cloud publish and catalog lookup.

Use this Dojo skill with package human docs, package-local feature inventory, agent-skillset.md, and the generated course.json in this folder. Keep this manifest version aligned with package.json.