Skip to content

9.3. Atlas Documentation

Documentation index for Atlas.

1. Documentation structure

1.1. Quick Start guides

1.2. Service documentation

  • Backend API — always-on adaptive FastAPI service
  • Open WebUI — main chat UI
  • LiteLLM Gateway — always-on OpenAI-compatible front door for every LLM provider
  • Ollama (LiteLLM upstream) — local LLM engine modes (container CPU/GPU, localhost, none)
  • ComfyUI — image generation workflows
  • Weaviate — vector database
  • MinIO — S3-compatible artifact-tier object storage
  • n8n — workflow automation
  • SearxNG — privacy metasearch
  • Redis — cache/queue infrastructure
  • Local Deep Researcher — research/orchestration service
  • Multi2Vec CLIP — multimodal vectorization
  • STT Provider (Speech-to-Text) — pluggable: Speaches (Faster-Whisper, default), Parakeet-TDT, whisper.cpp
  • TTS Provider (Text-to-Speech) — pluggable: Speaches (Kokoro/Piper, default), Chatterbox (voice cloning)
  • Document Processor (Docling) — document processing
  • Supabase Ecosystem — database, auth, and storage services
  • Neo4j (Graph Database) — graph database service
  • Kong (API Gateway) — dynamic API gateway configuration
  • JupyterHub (Data Science IDE) — interactive Jupyter Lab environment
  • LightRAG — graph-augmented RAG server with WebUI + multimodal ingestion
  • OpenClaw (AI Agent) — AI agent for messaging platforms
  • Hermes Agent — programmable AI agent runtime (Nous Research)
  • Ray — distributed compute substrate (head + workers, opt-in via RAY_SOURCE)
  • Prometheus — observability scraper + TSDB with bundled node-exporter and cAdvisor (opt-in via PROMETHEUS_SOURCE)
  • Grafana — observability dashboards + unified alerting on top of Prometheus (opt-in via GRAFANA_SOURCE)
  • Apache Spark — standalone Spark cluster (5-container family: master + workers + history + Spark Connect gRPC sidecar + minio/mc bucket init) for batch / SQL / DataFrame workloads (opt-in via SPARK_SOURCE)
  • Apache Zeppelin — Spark-first notebook UI; Spark interpreter pre-configured, JDBC interpreter requires a one-time UI setup (opt-in via ZEPPELIN_SOURCE; gated on Spark)
  • Apache Airflow — code-defined DAG orchestrator (webserver + scheduler + dag-processor + init) with LiteLLM-wired LLM operators (opt-in via AIRFLOW_SOURCE)
  • TEI Reranker — Cross-encoder reranker (default mxbai-rerank-base-v1) for RAG quality lift
  • Cloudflared Tunnel — Cloudflare Tunnel for zero-config public TLS ingress fronting Kong (opt-in via CLOUDFLARED_SOURCE)
  • Backup / Restore — on-demand Postgres dump + volume tarballs to S3/MinIO with a restore path (opt-in via BACKUP_SOURCE)

1.3. Deployment guides

1.4. Contributors

  • Adding a service runbook — six-decision walkthrough + the regen + lint chain
  • Security policy — threat tiers, supported versions, responsible-disclosure address
  • External dependency contract ledger — durable record of consumed external API/CLI/config contract checks from maintenance passes

1.5. Architecture diagrams

  • Diagrams README — top-level diagram update workflow + the per-service auto-generation chain
  • Split architecture catalog — MkDocs-linked high-level architecture perspectives generated for the publishable docs site
  • The top-level diagram itself lives at diagrams/architecture.svg (embedded in the project README) and diagrams/architecture.html (standalone view)

1.6. Cross-service research (Phase B corpus)

  • Research corpus guide — layout, authoring rules, and the schema the validator enforces
  • Integration matrix — auto-generated index linking every service to its candidate integrations
  • Per-service rows — missing-pair integrations, candidate new services, per-service feature gaps
  • Candidate one-pagers — design notes per candidate service

1.7. Feature-track plans and specs

  • superpowers/plans + superpowers/specs — point-in-time implementation plans and specs for the larger 2026-05/06 feature tracks (consult when archaeology on a past track is needed; CHANGELOG entries link the relevant ones)

1.8. Numbering-policy notes

  • Generated research files keep schema-fixed headings such as ## Headline; see research/README.md for the explicit exemption.
  • Provider implementation notes under services/*/provider/ are operational backend-specific runbooks. They may keep compact unnumbered headings when numbering would make command-oriented maintenance notes harder to scan.
  • Conventional history/planning artifacts such as CHANGELOG, ROADMAP, and docs/plans/ may keep their established release-note or planning heading style when renumbering would obscure chronology.
  • Literal UI/output glyphs may appear only when the documentation is naming an actual terminal control, status marker, tree connector, or generated output. Do not use glyphs as decorative prose. Prefer words such as Warning: in explanatory text, and keep flow arrows or tree characters inside technical notation or literal examples.
  • Main README — project overview and quick start
  • ROADMAP — future development plans
  • CHANGELOG — release history and completed features

3. Getting help

If you can't find what you're looking for:

  1. Check the Troubleshooting Guide
  2. Search through the service-specific documentation
  3. Open an issue on GitHub if you need additional help

4. Contributing to documentation

  • Found a typo or error? Open a PR.
  • Missing information? Open an issue.
  • Before submitting documentation changes, run the single root-safe gate: make docs-check.