LibreChat

# MISSION: Scaffold "LibreChat" Production Architecture
You are an expert Senior Staff Software Architect and Full-Stack Engineer. Your mission is to scaffold and implement a production-grade, highly reliable, and modular codebase following the proven architecture of **LibreChat**.
---
## 1. PROJECT SPECIFICATIONS & BENCHMARK
- **Reference Architecture**: LibreChat
- **What It Does**: LibreChat is a self-hostable, open-source chat interface that unifies many AI providers (OpenAI, Anthropic, Azure, Bedrock) behind one UI, adding agents, MCP tool support, code interpreters, and multi-user auth.
- **Domain & Category**: Multi-Model AI Agent & Chat Workspace
- **Production Scale**: 6-20 People
- **Development Mode**: HYBRID
- **Architectural Rationale**: A Node/Express API with a Mongoose-modeled MongoDB store handles chat state, while Vite builds the React client separately; optional Docker sidecars for pgvector, Meilisearch, and a Go OpenTelemetry relay keep RAG, search, and observability decoupled from the core API.
- **Live Website Reference**: https://librechat.ai
- **Source Repository**: https://github.com/danny-avila/LibreChat
---
## 2. PRODUCTION TECH STACK
- **Full Stack Array**: TypeScript, JavaScript, React, Vite, Express, MongoDB, Redis, pgvector, Model Context Protocol, Go, Docker, Tailwind CSS
- **Primary Language(s)**: TypeScript, JavaScript
- **License of the reference repo**: MIT
- **Frontend**: React — Renders the chat UI in client/src as a single-page app.; Vite — Bundles and dev-serves the React client.; Tailwind CSS — Styles the client via the shared tailwind config.
- **Backend & APIs**: Express — Serves the REST API and static client from the api/server entrypoint.; Model Context Protocol — Agents and the Assistants API connect to external MCP servers and tools.
- **Data & persistence**: MongoDB — Primary datastore for conversations, users, and agents.; Redis — Backs caching, rate limiting, and resumable streams across scaled instances.; pgvector — Stores document embeddings for the optional RAG API vector search.
- **Infrastructure & deploy**: Docker — docker-compose.yml wires api, mongodb, meilisearch, vectordb, and rag_api together.; Go — otel/langfuse-fanout is a small Go service relaying OpenTelemetry traces to Langfuse.; GitHub Actions — Builds, tests, and publishes multi-arch Docker images across dozens of workflows.
- **Tooling, testing & ops**: Model Context Protocol — The Agents and Assistants endpoints call MCP servers for tool use via the SDK dependency.
---
## 3. KEY ARCHITECTURAL DECISIONS (audited from https://github.com/danny-avila/LibreChat @ 12d7890)
1. **MongoDB, not PostgreSQL, is the single source of truth**: Despite bundling a pgvector-backed vectordb container for RAG embeddings, core conversation, user, and agent data lives in MongoDB via Mongoose models in packages/data-schemas, per the mongodb service defined in docker-compose.yml.
2. **Client, backend, and shared logic are npm workspaces, not separate repos**: Root package.json declares workspaces for api, client, and packages/*, letting packages/data-provider, packages/data-schemas, and packages/api ship as versioned internal packages consumed by both api and client.
3. **RAG and search run as optional sidecar services**: docker-compose.yml adds vectordb (pgvector) and rag_api as separate containers, and search/compose.yml configures Meilisearch independently, so a minimal deployment can skip both without touching api/server code.
4. **A dedicated Go service fans out OpenTelemetry traces to Langfuse**: otel/langfuse-fanout/go.mod defines a small Go binary, wired up in docker-compose.langfuse-fanout.yml, that receives OTLP traces from the Node API and forwards them to a Langfuse instance, isolating that integration from the main backend process.
5. **Authentication supports many strategies through one Passport-based layer**: api/package.json lists passport-google-oauth20, passport-ldapauth, passport-jwt, passport-apple, passport-discord, passport-facebook, passport-github2, and @node-saml/passport-saml, all wired through api/strategies so OAuth, LDAP, SAML, and email/password login share one session model.
6. **MCP support lets agents call external tools instead of hardcoded plugins**: api/package.json depends on @modelcontextprotocol/sdk, and docs/skills-management-api.md plus the skill/ directory show LibreChat exposing SKILL.md-based agent capabilities layered on top of the MCP protocol.
---
## 4. NON-NEGOTIABLE ARCHITECTURAL GUARDRAILS
1. **Monorepo & Modular Separation**:
   - Structure as a Turborepo monorepo with strict package boundaries:
     - `apps/web`: Application UI, routing, layouts, and server endpoints.
     - `packages/ui`: Shared design tokens, CSS variables, and Radix UI primitive components.
     - `packages/db`: Database schemas, client singleton, declarative migrations, and seed scripts.
     - `packages/config`: Shared TypeScript, ESLint, and build configurations.
2. **Strict Type Safety & Zero `any` Policy**:
   - Enable `strict: true`, `noImplicitAny: true`, and `strictNullChecks: true`.
   - Validate ALL external inputs, API request bodies, and query parameters with **Zod** schemas before execution.
3. **Frontend & Rendering Guidelines**:
   - Utilize Vite 6 with TanStack Router for fully type-safe routing. Manage server state and caching via TanStack Query v5 with optimistic UI updates.
4. **Design System & Aesthetics**:
   - Keep every color, radius, shadow and font in a single token file (CSS variables) and consume tokens everywhere; never hardcode hex values in components.
   - Prefer crisp 1px borders and one subtle shadow scale over blurry default shadows. Pair one sans-serif for body/headings with one monospace for tags, badges, metrics, and code.
5. **Data Layer & Reliability**:
   - Write declarative schema definitions with foreign keys, composite indexes on queried filters, and automated timestamp triggers.
   - Use connection pooling and prepared statements for serverless database execution.
---
## 5. STEP-BY-STEP SCAFFOLDING ROADMAP
- **Phase 1: Workspace & Root Config**: Initialize package manager, monorepo configuration (`turbo.json`, `tsconfig.base.json`, `package.json`).
- **Phase 2: Database Schema & Client**: Set up the data layer (MongoDB, Redis, pgvector): client, connection pool, models, and migration scripts.
- **Phase 3: Design Tokens & UI Primitives**: Build accessible `Button`, `Input`, `Card`, `Badge`, and layout wrappers inside `packages/ui`.
- **Phase 4: Core Application Routes & Handlers**: Implement primary authentication, user session handling, and application routes.
- **Phase 5: Quality Assurance & Build Verification**: Run `tsc --noEmit`, ESLint, Prettier, and smoke test suites to ensure zero compilation or runtime errors.
---
## 6. EXECUTION INSTRUCTIONS
1. Review all specifications, architectural guardrails, and stack choices above.
2. Present the full monorepo directory tree structure.
3. Systematically generate the complete, production-ready codebase according to the 5-phase roadmap above — starting with the root workspace setup, followed by the database schema, UI design system package, and full-stack application routes until the repository is fully scaffolded and ready to run.