Mission: Understand risk-graph-indexer deeply, end to end

Why

The user wants a deep, correct, mechanism-level understanding of the entire

risk-graph-indexer system (Forta Network) — how every layer actually works and why it's

built that way — before they start contributing. Contributing is still the eventual goal;

it's now sequenced after full comprehension, not abandoned. (Re-sequenced 2026-06-10 from

"onboarding to contribute [now]": understand everything deeply first → then contribute on a

solid foundation. See LR-0007.)

Two phases: (1) NOW — deep end-to-end understanding (mechanisms, trade-offs, failure modes).

(2) LATER — guided real contributions, once phase 1 is complete. Don't push phase 2 yet; the

already-covered architecture (Lessons 1–6) + the connective-tissue/failure lessons are the foundation it will build on.

Success looks like

• Can explain every layer end to end (ingest → index → enrich → risk) at the level of why,

not just what — including the failure modes and the design trade-offs behind each choice.

• Can trace a real on-chain event (e.g. an ERC-20 Transfer) end-to-end into a specific graph edge,

AND explain what happens when things go wrong (crash, reorg, consumer lag, stale write).

• Understands the connective tissue, not just the binaries: Redis Streams, consumer groups,

backpressure, the single-writer model, cursor/atomicity/idempotency as one coherent story.

• Has a solid-enough grip on the Go / Memgraph(Cypher) / Redis-streams stack to read any hot path

and reason about its correctness.

• Can read the package layout and predict, for any subsystem, how it plugs into the whole.

Constraints

• Strong on blockchain/EVM (blocks, logs, traces, ERC-20, DeFi protocols).
• Newer to Go, graph databases / Cypher, and streaming / event pipelines — lean on the EVM anchor and scaffold these.
• Learning for deep comprehension: bias lessons toward how it really works and why it's built this way

(mechanisms, trade-offs, failure modes) over "where would I change this." Still cite real code/files.

• Keep the format that's working: one tightly-scoped lesson, real code citations, EVM anchors, end-of-lesson quiz with instant feedback. (User has affirmed the lessons are excellent.)

Out of scope (for now)

• Deep risk-math formulas (DebtRank/centrality/HHI weighting, exposure BFS propagation math).

Risk-engine concepts are now in-scope at orientation level (covered in Lesson 6: AT_RISK cells,

exposure BFS, incremental delta-driven recompute) — but the heavy math stays deferred until the

user has shipped simpler changes. The risk engine is explicitly NOT a first-PR area (parity constraints).

• Ops/infra (k8s deploy, Grafana, Prometheus tuning) beyond knowing they exist.
• The frozen legacy Python batch pipeline — treat as nonexistent per repo CLAUDE.md.