MCP Event Driven Ledger

分类账——第9-10周入门代码

快速开始

# 1. Install dependencies
pip install -r requirements.txt

# 2. Start PostgreSQL
docker run -d -e POSTGRES_PASSWORD=apex -e POSTGRES_DB=apex_ledger -p 5432:5432 postgres:16

# 3. Set environment
cp .env.example .env
# Edit .env — add your ANTHROPIC_API_KEY

# 4. Generate all data (companies + documents + seed events → DB)
python datagen/generate_all.py --db-url postgresql://postgres:apex@localhost/apex_ledger

# 5. Validate schema (no DB needed)
python datagen/generate_all.py --skip-db --skip-docs --validate-only

# 6. Run Phase 0 tests (must pass before starting Phase 1)
pytest tests/test_schema_and_generator.py -v

# 7. Begin Phase 1: implement EventStore
# Edit: ledger/event_store.py
# Test: pytest tests/test_event_store.py -v

什么开箱即用

• 完整事件模式（45种事件类型）-- ledger/schema/events.py
• 完整的数据生成器（GAAP PDF、Excel、CSV、1200多个种子事件）
• 事件模拟器（所有5个代理管道，确定性）
• 模式验证器（根据EVENT_REGISTRY验证所有事件）
• 第0阶段测试：10/10通过

你实施什么

按阶段划分的闸门测试

pytest tests/test_schema_and_generator.py -v  # Phase 0: all must pass before Phase 1
pytest tests/test_event_store.py -v           # Phase 1
pytest tests/test_domain.py -v               # Phase 2
pytest tests/test_narratives.py -v           # Phase 3: all 5 must pass
pytest tests/test_projections.py -v          # Phase 4
pytest tests/test_mcp.py -v                  # Phase 5

账本——代理事件存储和审计基础架构

Apex金融服务贷款决策平台。五个LangGraph代理。完整的事件源审计跟踪。PostgreSQL支持。

______________________________________________________________________

先决条件

• Python 3.11+
• PostgreSQL 14+
• 无烟煤API键

______________________________________________________________________

安装

pip install -r requirements.txt

______________________________________________________________________

环境

复制 .env.example 到 .env 并填写您的值：

cp .env.example .env

所需变量：

ANTHROPIC_API_KEY=sk-ant-...
DATABASE_URL=postgresql://ledger:ledger@localhost:5432/apex_ledger
TEST_DB_URL=postgresql://ledger:ledger@localhost:5432/apex_ledger_test
DOCUMENTS_DIR=./documents
REGULATION_VERSION=2026-Q1
LOG_LEVEL=INFO

______________________________________________________________________

数据库设置

创建数据库：

psql -U postgres -c "CREATE DATABASE apex_ledger;"
psql -U postgres -c "CREATE DATABASE apex_ledger_test;"
psql -U postgres -c "CREATE USER ledger WITH PASSWORD 'ledger';"
psql -U postgres -c "GRANT ALL PRIVILEGES ON DATABASE apex_ledger TO ledger;"
psql -U postgres -c "GRANT ALL PRIVILEGES ON DATABASE apex_ledger_test TO ledger;"

运行迁移：

psql -U ledger -d apex_ledger -f src/schema.sql
psql -U ledger -d apex_ledger_test -f src/schema.sql

______________________________________________________________________

种子数据

运行数据生成器（一次--幂等）：

python datagen/generate_all.py \
    --applicants 80 \
    --db-url postgresql://ledger:ledger@localhost:5432/apex_ledger \
    --docs-dir ./documents \
    --output-dir ./data \
    --random-seed 42

预计产量：80家公司，320份文件，1847场种子活动。

______________________________________________________________________

运行测试

# Phase 1 — InMemoryEventStore (no DB required)
pytest tests/phase1/ -v

# Concurrency test (mandated)
pytest tests/test_concurrency.py -v

# All tests
pytest tests/ -v

______________________________________________________________________

运行管道

端到端处理一个应用程序：

python scripts/run_pipeline.py --app APEX-0007 --phase document
python scripts/run_pipeline.py --app APEX-0012 --phase credit

______________________________________________________________________

MCP 服务器

python -m ledger.mcp_server

默认情况下，服务器在端口8765上运行。

______________________________________________________________________

项目结构

apex-ledger/
├── schema/events.py          # Canonical event schema — never modify
├── agents/                   # LangGraph agents
├── registry/client.py        # Applicant Registry — read-only
├── src/                      # Core implementation
│   ├── schema.sql            # PostgreSQL schema
│   ├── event_store.py        # EventStore + InMemoryEventStore
│   ├── models/events.py      # Re-exports from schema/events.py
│   ├── aggregates/           # Domain aggregates
│   ├── commands/handlers.py  # Command handlers
│   ├── projections/          # Read model projections
│   ├── upcasting/            # Schema evolution
│   ├── integrity/            # Audit chain + Gas Town
│   └── mcp/                  # MCP server
├── tests/                    # Test suite
├── datagen/                  # Data generator
├── scripts/                  # Pipeline + demo scripts
└── artifacts/                # Generated outputs