Skip to content

System Context — learn-ai

Generated 2026-06-24 13:33 UTC · c4gen dev

The widest view: learn-ai and every external actor and system it exchanges data with. Edges shown are curated and code-verified; raw graph-derived candidates are listed under Dependencies & Cycles.

Interactive

Drag to pan, scroll to zoom. Click the learn-ai box to drill into its container view.

External systems & peers

System Role
Keycloak (SSO) OAuth2/OIDC identity provider (realm olapps). APISIX runs the openid-connect flow against it; APISIX_USERDATA maps claims to Django users via middleware.
HashiCorp Vault Secrets (LEARN_ACCESS_TOKEN, proxy/provider keys) injected at runtime.
LLM Provider (OpenAI-compatible / Bedrock) Backing completion models behind LiteLLM — OpenAI (default openai/gpt-4o-mini) and AWS Bedrock (Claude 3.5 Sonnet). Streamed token responses.
LangSmith LangChain tracing/observability. Each chat turn is wrapped in a langsmith trace when LANGSMITH_API_KEY is set.
MIT Learn API Course & learning-resource discovery platform. learn-ai's agent tools call its vector-search, content-file, video-transcript, and tutor problem-set endpoints (bearer LEARN_ACCESS_TOKEN). In MIT Learn's own map, learn-ai is modeled as the external /ai/* peer — the mirror of this edge.
PostHog Product analytics + LLM token tracking. The LangChain CallbackHandler emits generation/token events per turn.
OpenTelemetry Collector OTLP traces from mitol-django-observability when OPENTELEMETRY_ENDPOINT is set (off by default).