Adapters overview¶

Adapters are optional telemetry / integration workers. They are not landing pages. Users do not open an adapter as the AMS home screen; modules and platforms do.

When to install an adapter¶

Install an adapter when you need AMS to observe or bridge an upstream product:

Adapter chart	Upstream / purpose
`ams-helm-dify-adapter`	Dify agent platform activity → AMS OTEL path
`ams-helm-ema-adapter`	EMA integration telemetry
`ams-helm-librechat-adapter`	LibreChat activity → AMS OTEL path
`ams-helm-lyzer-adapter`	Lyzer integration telemetry
`ams-helm-mscopilot-adapter`	Microsoft Copilot integration telemetry

You may install none, one, or several adapters independently of how many UI modules you chose.

How adapters fit the data path¶

flowchart LR
  platform[Platform e.g. Dify or LibreChat]
  langfuse[Langfuse optional]
  adapter[AMS adapter]
  otel[OTEL / Observability]
  platform --> langfuse
  langfuse --> adapter
  platform --> adapter
  adapter --> otel

A common Dify-oriented flow:

Users build and run agents in Dify.
Traces / evaluation data often land in Langfuse.
The Dify adapter reads that signal and exports into AMS OpenTelemetry / observability.
Operators inspect the result in Grafana (or related tools) from the observability stack.

LibreChat follows a similar idea: chat surface → adapter → OTEL. Exact env vars differ per chart — use each adapter page and release values.

Shared prerequisites¶

Observability stack installed and OTLP (or chart-specific) endpoints known
Network path from adapter namespace to the upstream API (Dify, Langfuse, Copilot, …)
Secrets for upstream API keys stored outside Git

General Helm pattern¶

helm upgrade --install ams-<adapter> ./ams-helm-<adapter>-adapter \
  --namespace <adapter-namespace> \
  --create-namespace \
  -f providers/azure/release/values.yaml

See the individual adapter pages for chart names and notes.