LangChain

# MISSION: Scaffold "LangChain" 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 **LangChain**.
---
## 1. PROJECT SPECIFICATIONS & BENCHMARK
- **Reference Architecture**: LangChain
- **What It Does**: LangChain is a Python framework for building LLM-powered applications and agents, organized as a monorepo of independently versioned packages: a core Runnables/Pydantic library, a LangGraph-based agent package, and dozens of separately published model-provider and tool integrations.
- **Domain & Category**: LLM Application Framework & Agent Orchestrator
- **Production Scale**: 1M+ MAU
- **Development Mode**: AI-AGENT
- **Architectural Rationale**: The current langchain package is a thin agent-building layer on top of LangGraph, with the older chain-based API kept separately as langchain-classic, and every model provider ships as its own versioned package so integrations can release independently of the core library.
- **Live Website Reference**: https://www.langchain.com
- **Source Repository**: https://github.com/langchain-ai/langchain
---
## 2. PRODUCTION TECH STACK
- **Full Stack Array**: Python, LangChain, LangGraph, Pydantic, Chroma
- **Primary Language(s)**: Python, Makefile, Shell
- **License of the reference repo**: MIT
- **Backend & APIs**: Python — Implementation language for every package in the monorepo (core, langchain, and all partner integrations).; Pydantic — Data-validation and schema layer (v2) underlying Runnables, structured output, and tool-calling schemas in langchain-core.; httpx — Async HTTP client dependency of langchain-core and the model-profiles CLI.
- **Data & persistence**: Chroma — Reference vector-store integration package (langchain-chroma) for retrieval-augmented generation examples and tests.
- **Infrastructure & deploy**: uv — Workspace and dependency manager pinning versions across all libs/* packages, kept current by a dedicated CI workflow.; Hatchling — Build backend used to produce the wheel/sdist for every package in the monorepo.
- **Tooling, testing & ops**: LangGraph — Graph-based agent orchestration engine that the current langchain package (libs/langchain_v1) is built directly on top of.; LangSmith — Tracing/observability SDK that langchain-core depends on directly for run tracing.; langchain-model-profiles — Standalone CLI package that generates and updates per-model capability metadata used by provider integrations.; Provider integration packages — Each provider (OpenAI, Anthropic, Fireworks, Groq, HuggingFace, MistralAI, Ollama, Nomic, DeepSeek, Exa) ships as its own versioned package.
---
## 3. KEY ARCHITECTURAL DECISIONS (audited from https://github.com/langchain-ai/langchain @ b3ec709)
1. **The "langchain" package is now a thin agent layer on LangGraph, not the classic chains API**: libs/langchain_v1/pyproject.toml publishes the name "langchain" (version 1.4.1) depending on langgraph>=1.2.11,<1.3.0, while libs/langchain/pyproject.toml publishes "langchain-classic" (version 1.0.8), the pre-rewrite chains/agents API kept as a separate package for backwards compatibility.
2. **Every model or tool provider ships as its own independently versioned package**: libs/partners/openai/pyproject.toml (langchain-openai 1.6.2) and libs/partners/anthropic/pyproject.toml (langchain-anthropic 1.7.2) each declare their own version and their own minimum langchain-core floor, so a provider integration can release without forcing a core or langchain release.
3. **Model capability metadata is generated by its own CLI package, refreshed in CI**: libs/model-profiles/pyproject.toml defines the langchain-model-profiles package with a langchain-profiles console script (project.scripts), and the repository runs a refresh_model_profiles.yml / _refresh_model_profiles.yml workflow pair to keep per-model data current across provider packages.
4. **Text splitting is factored out of langchain-core into its own package**: libs/text-splitters/pyproject.toml publishes langchain-text-splitters as a package depending only on langchain-core, separately versioned so splitter-only consumers do not need the rest of langchain.
5. **The repository maintains an agent-readable documentation set, kept in sync by CI**: The openwiki/ directory (agent-execution.md, agent-factory.md, mcp-integration.md, partner-pattern.md, and others) documents internal patterns specifically for AI coding agents working in the repo, and an openwiki-update.yml workflow plus check_agents_sync.yml keep it aligned with the code as it changes.
6. **uv pins dependency versions across roughly fifteen published packages in one workspace**: A bump_uv_pin.yml workflow updates the uv lock pin repo-wide, and check_versions.yml / check_extras_sync.yml verify that optional-dependency extras in libs/langchain/pyproject.toml and libs/langchain_v1/pyproject.toml stay in sync with the actual partner packages under libs/partners.
7. **Provider integrations are tested against recorded HTTP cassettes by default**: libs/partners/openai/pyproject.toml and libs/partners/anthropic/pyproject.toml both list vcrpy in their test dependency group, and a _test_vcr.yml workflow runs the suite against recorded cassettes so most CI runs do not call live provider APIs.
---
## 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**:
   - Isolate interactive UI state to leaf components. Keep core pages lightweight and performant.
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 (Chroma): 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.