Lumen

Type a one-sentence learning goal — an AI orchestrator builds you a private course in ~50 seconds, a RAG tutor with citations teaches it, and you can audit every agent decision it made.

_{Custom multi-agent system, no LangChain · public evals with the weak scores kept in · live in production}

Live demo · Eval results · Architecture · MCP server

_{Real production recording (Groq Llama 3.3 70B). Intake trimmed 6×, the ~50 s build 16× — the brief and the finished course are real-time.
Try it yourself: the one-click demo pre-fills demo@lumen.test / Demo!2026 and drops you into the tutor (free-tier box — give a cold page a few seconds).}

A learner-owned, two-role e-learning platform — every signed-in user runs the whole loop themselves; admin only moderates and configures. The product is the loop; the point of the repo is the agentic system underneath it.

Shipped to production as 2.0.0-two-role (CHANGELOG) — built as a gated waterfall: requirements → design → 6 ADRs → seven build streams, each cleared a Codex challenge, an independent Claude review, and a live in-browser walk before merge.

Every item below is on production today, with the code one click away.

The tutor picks per-turn among five sub-agents in tutor_subagents/ — retriever, web_searcher, code_runner, quiz_generator, concept_explainer — under a hard cap on tool-call rounds (streaming variant). The authoring side runs a six-stage pipeline — researcher → outliner → critic → reviser → lesson-drafter → final-critic — in authoring_orchestrator.py, capped at six revise/critic calls.

_{Production recording at 2× — the retriever fires (latency on-screen), then the answer streams.}

Retrieval is scoped per course and routed through a single ACL clause (visibility.py, ADR-0029) so private and cloned courses never leak chunks. Embeddings via Cloudflare Workers AI (bge-small-en-v1.5, 384-dim) into pgvector; answers cite specific lesson chunks.

Each LLM call logs prompt/completion tokens, USD cost, latency, and outcome to the llm_calls table (llm_call_log.py); each agent step lands in agent_tracer.py. Learners get a per-turn "show me how you got this" drill-down — planner steps, tool calls, retrieval audits with similarity scores; authors get a step-by-step build replay.

Tutor-turn trace	Authoring build replay

Three golden suites (30-item tutor, 10 authoring, 10 ingest) under evals/, judged 0–5 per axis, plus adversarial probes. A 3-item smoke gates every PR (workflow); results are public at /eval.

The point isn't the scores — it's the harness: LLM-as-judge applied honestly to one strong subsystem and two early ones, every number reproducible with make eval suite=… and smoke-gated in CI.

The weak scores are published whole, on purpose. Methodology · raw reports.

app/mcp/ exposes nine tools (catalog, RAG tutor, FSRS reviews, AI authoring, ingest) over stdio + HTTP with OAuth client-credentials, published as io.github.ahmedEid1/lumen. Write tools gate on the can_author capability; URL ingest stays admin-only + flag-gated (ADR-0025).

Wiring it into an MCP client takes one JSON block:

// claude_desktop_config.json
{
  "mcpServers": {
    "lumen": {
      "command": "python",
      "args": ["-m", "app.mcp", "--transport", "stdio"],
      "env": {
        "LUMEN_MCP_AUTH_TOKEN": "<client-secret from `make mcp-token`>",
        "DATABASE_URL": "postgresql+asyncpg://lumen:lumen@localhost:5432/lumen"
      }
    }
  }
}

Per-credential 256-bit DEKs wrapped by a versioned server KEK (secrets_crypto.py); decryption only inside the dispatch path — never in logs, traces, exports, or admin views. A prod boot guard refuses to start with stored credentials but no real KEK (ADR-0027). Request-count quotas close the $0-BYOK bypass of the dollar budget guard.

The two-role rebuild migrated student/instructor → user live: widen accepted roles → backfill → drop old values only after access tokens drained their 15-min TTL. The Alembic chain enforces one phase per run, and migration-level evidence gates refuse to tighten constraints over un-backfilled rows — one of them caught a real prod data condition (85 legacy chunks with NULL embedding model) at deploy time.

The short version: one FastAPI service owns all invariants, the agents are plain service-layer code, and every LLM call crosses one provider seam and one cost meter.

flowchart LR
    user([User · authors + learns])
    admin([Admin · moderates + config])

    subgraph App[Application]
      web[Next.js 15 · RSC]
      api[FastAPI · Python 3.13<br/>capability-based authz<br/>central is_publicly_listed authorizer]
      worker[Celery worker + beat]
    end

    subgraph Agents[Agent layer]
      define[Goal intake → brief]
      authoring[Authoring orchestrator<br/>researcher → … → final-critic]
      tutor[Tutor orchestrator<br/>+ 5 sub-agents]
    end

    subgraph Data[Data]
      pg[(Postgres 17 + pgvector)]
      redis[(Redis 7)]
      s3[(MinIO S3)]
    end

    subgraph LLM[Swappable LLM layer]
      provider{LLM_PROVIDER dispatch}
      groq[Groq · Llama 3.3 70B]
      byok[BYOK allowlist<br/>OpenAI · Anthropic · Mistral · Groq]
    end

    subgraph Eval[Eval loop]
      golden[(Golden datasets)]
      judge[LLM-as-judge 0–5]
      meter[llm_calls<br/>tokens · cost · latency]
    end

    user --> web --> api --> pg
    admin --> web
    api --> redis & s3 & worker
    api --> define --> authoring
    api --> tutor --> pg
    mcp[MCP server · 9 tools] --> api
    authoring & tutor --> provider
    provider -.platform.-> groq
    provider -.per-user.-> byok
    api --> meter
    golden --> judge --> provider

Provider-agnostic by env var — the live demo runs Groq's free tier; users dispatch on their own allowlisted keys. Every call crosses the cost meter, so budgets, quotas, and observability behave identically across providers. Full topology: docs/architecture.md.

Stack: Python 3.13 · FastAPI · async SQLAlchemy 2 · Alembic · Celery — Next.js 15 · React 19 · TypeScript 5 · Tailwind 4 · TanStack Query — PostgreSQL 17 (pgvector + tsvector) · Redis 7 · MinIO — Docker Compose · GitHub Actions · Trivy + CodeQL + gitleaks · Caddy 2.

The process is the portfolio as much as the code. Every build stream cleared three gates before merge:

1. Codex challenge — a second-brain CLI attacks the design, plan, and code; findings triaged and resolved.
2. Independent Claude review — a gating review subagent re-checks against source until clean.
3. Live evidence — drive the app as a real user in a browser, locally and on prod, on top of unit/E2E/a11y suites. Running-the-app evidence is required, not optional.

At the 2.0.0 release: backend 1,421 tests / frontend 468 tests, all green; WCAG 2.2 AA axe-core gate (11 surfaces, 0 violations); en + ar i18n parity; visual-regression baselines; Playwright E2E on Chromium and WebKit. The UI itself went through a 20-loop redesign (30+ Radix-backed primitives, ⌘K command palette, dark/light themes) with five in-loop Codex rescue passes plus a final Codex review. A green main auto-deploys to production.

_{⌘K — navigate, search courses, switch theme, sign out.}

Prereqs: Docker Desktop 4.30+ (or Engine 27 + Compose v2).

git clone https://github.com/ahmedEid1/lumen.git
cd lumen
cp .env.example .env
make up && make migrate && make seed

Open http://localhost:3000 and sign in:

Without an LLM key the AI features fall back to a deterministic noop provider — the rest of the app still works. For the real thing (define/build, tutor, evals), a free Groq key is enough:

LLM_PROVIDER=openai
OPENAI_API_BASE=https://api.groq.com/openai/v1
OPENAI_API_KEY=<your-groq-key>
LLM_MODEL=llama-3.3-70b-versatile

The same LLMProvider abstraction takes native Anthropic or OpenAI by env var — no code changes. Feature flags (FEATURE_BYOK_ENABLED, FEATURE_PRIVATE_PUBLISH_ENABLED, CLONE_ENABLED, FEATURE_TUTOR_STREAMING) default off; set them in .env once their prerequisites (e.g. a real BYOK master key) are in place. make demo-seed adds the richer agentic-demo bundle.

More screenshots — dashboard, catalog, the agent-replay home page, the public eval page, a freshly built course, the brief review

• Live at lumen.ahmedhobeishy.tech — 2.0.0-two-role, shipped 2026-06-06. Free tier end-to-end: Groq + Cloudflare Workers AI + one AWS t4g.small (runbook: docs/deployment/aws-vps.md). Budget guards and request quotas cap spend; expect free-tier latency under load.
• Email verification is off in prod (EMAIL_ENABLED=false — no SMTP configured). A flag, not a code limitation.
• Eval scores include the weak ones, with causes documented — see the table above and docs/eval/ for raw JSONL.
• Test counts are release-time snapshots (the suites keep moving with main); CI is the live source of truth.

Ahmed Hobeishy — AI / Agent Engineer in Essen, Germany. Lumen started as a 2020 Django side-project; five years and one model revolution later it's the centrepiece of my agentic-AI work: agents that are measured (golden evals, LLM-as-judge), auditable (per-call cost/latency traces, citation checks), and shipped (live, CI-gated, self-hostable).

Open to AI / Agent Engineer roles in Germany where evaluation and observability are first-class.

LinkedIn · GitHub · or open an issue here.