Skip to Content
Overview

OpsDNA Engineering Docs

OpsDNA is an agentic CFO platform for venture capital firms. It exists because fund back offices run on spreadsheets, email, and trust in whichever admin last touched the numbers — and none of that is replayable or auditable. So the platform ingests the raw facts of a fund’s financial life — bank transactions, capital calls, LPAs and side letters, valuation marks, portfolio KPIs — and turns them into an auditable, deterministic general ledger with per-LP capital accounting (ASC 946 / ASC 820), plus a permissioned runtime where LLM-assisted answers and workflows stay constrained behind typed schemas, registries, and evidence.

These docs are for engineers working in the OpsDNA monorepo: the product repo at opsdna/opsdna, a Bun + Turborepo workspace of ~120 packages and 7 apps. They are internal — assume the reader has repo access, and treat CLAUDE.md, DATA-MODEL.md, DESIGN.md, and docs/adrs/ in the repo as the normative sources these pages summarize and link together.

If you just cloned the repo, the whole loop is four commands:

cd ~/development/opsdna-workspace/opsdna
bun install
bun run dev:stack     # local portal + API + Docker Postgres, zero credentials
bun run verify        # the mandatory pre-push gate (~3-8 min)

Sections

Getting StartedArchitectureEntity ModelPersistenceFund AccountingSource PipelineRuntimeAPI (apps/api)Portal (apps/portal)InfrastructureGlossary

Where to start

A reading path that matches how the system is layered:

  1. Getting Started — clone, bun install, bun run dev:stack, and the one non-negotiable habit: bun run verify before every push (and why verify is not the whole story).
  2. Architecture — the whole-system map (Cloudflare portal → Lambda API → RDS, the Plaid/SQS ingestion plane, the query/policy runtime) and the house design principles you will see repeated in every section.
  3. Entity Model — the layer cake (tenant → organization → fund / legal entity → extensions), the identifier policy, and tenancy/RLS. Everything else keys off this.
  4. Persistence and Fund Accounting — how data lands (migrations, repositories, roles/RLS) and what the money core does with it (ledger kernel, Application/Trace/Row records, policy IR, Acceptance Lab).
  5. Then branch by what you’re touching: Source Pipeline for bank/document ingestion, API and Portal for the serving planes, Runtime for query/policy/AI surfaces, Infra for deploys and operations.

When a term stops you cold, the Glossary links every recurring concept to its owning page.

Last updated on
Overview