Files
ai-gateway-go/deploy/docker-compose.yml
T
superidou b536672000 feat(m8): P2 pgvector + Ollama 向量化与语义检索
- PostgreSQL 切换 pgvector/pgvector:pg17 镜像;迁移 000024 建 vector 扩展、
  knowledge_chunks.embedding vector(1024) + HNSW 余弦索引,retrieval_mode 放宽三态
- OllamaEmbedder 本地 bge-m3 批量嵌入,404 惰性 pull 重试,维度/超时校验,可整体关闭
- SemanticRetriever/HybridRetriever + NewRetriever 按 retrieval_mode 分发,缺 embedder 回退 FTS
- 文档入库同步批量向量化;Ollama 故障降级入库 + embedding_failed 事件
- 修复 pgx CopyFrom 对 vector 列二进制编码误读:COPY 基础列后同事务 unnest 批量回填
- 修复降级路径 embeddings=nil 索引越界 panic(Add 与 Reprocess)
- 知识库列表 vectorized_chunk_count + 前端三态检索模式选择与向量化覆盖率
- 单测 embedder/retrievers + 集成 TestKnowledgeVectorLifecycle 全绿

Co-Authored-By: Claude <noreply@anthropic.com>
2026-08-12 15:16:32 +08:00

198 lines
6.1 KiB
YAML

services:
# M8 P2:pgvector 镜像与 postgres:17-alpine 数据卷兼容;CREATE EXTENSION vector
# 由迁移 000024 执行(容器内 gateway 用户即 superuser)。
postgres:
image: pgvector/pgvector:pg17
environment:
POSTGRES_DB: gateway
POSTGRES_USER: gateway
POSTGRES_PASSWORD: gateway
volumes:
- postgres-data:/var/lib/postgresql/data
healthcheck:
test: ["CMD-SHELL", "pg_isready -U gateway -d gateway"]
interval: 5s
timeout: 3s
retries: 20
redis-critical:
image: redis:8.2-alpine
command: ["redis-server", "--appendonly", "yes", "--appendfsync", "everysec", "--maxmemory-policy", "noeviction"]
volumes:
- redis-critical-data:/data
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 5s
timeout: 3s
retries: 20
redis-cache:
image: redis:8.2-alpine
command: ["redis-server", "--maxmemory", "256mb", "--maxmemory-policy", "allkeys-lfu", "--save", ""]
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 5s
timeout: 3s
retries: 20
migrator:
build:
context: ..
dockerfile: deploy/Dockerfile
args:
VERSION: ${GATEWAY_VERSION:-0.10.0}
entrypoint: ["gateway-migrator"]
environment: &gateway-environment
APP_ENV: local
HTTP_ADDR: :8080
DATABASE_URL: postgres://gateway:gateway@postgres:5432/gateway?sslmode=disable
REDIS_CRITICAL_URL: redis://redis-critical:6379/0
REDIS_CACHE_URL: redis://redis-cache:6379/0
GATEWAY_BOOTSTRAP_API_KEY: ${GATEWAY_BOOTSTRAP_API_KEY:-}
GATEWAY_BOOTSTRAP_API_KEY_ENABLED: ${GATEWAY_BOOTSTRAP_API_KEY_ENABLED:-false}
CREDENTIAL_MASTER_KEY: ${CREDENTIAL_MASTER_KEY:-AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=}
CREDENTIAL_KEK_VERSION: 1
CREDENTIAL_KEK_KEYRING: ${CREDENTIAL_KEK_KEYRING:-}
ALLOW_PRIVATE_TOOL_URLS: ${ALLOW_PRIVATE_TOOL_URLS:-false}
ALLOW_PRIVATE_WEBHOOK_URLS: ${ALLOW_PRIVATE_WEBHOOK_URLS:-false}
UPSTREAM_BASE_URL: ${UPSTREAM_BASE_URL:-https://api.openai.com}
UPSTREAM_API_KEY: ${UPSTREAM_API_KEY:-}
SHADOW_BASE_URL: ${SHADOW_BASE_URL:-}
SHADOW_API_KEY: ${SHADOW_API_KEY:-}
SHADOW_SAMPLE_RATE: ${SHADOW_SAMPLE_RATE:-0}
S3_ENDPOINT: ${S3_ENDPOINT:-http://minio:9000}
S3_ACCESS_KEY_ID: ${S3_ACCESS_KEY_ID:-gateway}
S3_SECRET_ACCESS_KEY: ${S3_SECRET_ACCESS_KEY:-gateway-secret}
S3_BUCKET: ${S3_BUCKET:-gateway-files}
S3_REGION: ${S3_REGION:-us-east-1}
S3_USE_SSL: ${S3_USE_SSL:-false}
S3_MAX_FILE_BYTES: ${S3_MAX_FILE_BYTES:-134217728}
# M8 P2:本地 Ollama 向量化(bge-m3)。EMBEDDINGS_ENABLED=false 可整体关闭。
EMBEDDINGS_ENABLED: ${EMBEDDINGS_ENABLED:-true}
OLLAMA_BASE_URL: ${OLLAMA_BASE_URL:-http://ollama:11434}
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-bge-m3}
EMBEDDING_DIM: ${EMBEDDING_DIM:-1024}
EMBEDDING_BATCH_SIZE: ${EMBEDDING_BATCH_SIZE:-64}
EMBEDDING_TIMEOUT: ${EMBEDDING_TIMEOUT:-120s}
depends_on:
postgres:
condition: service_healthy
gateway-api:
build:
context: ..
dockerfile: deploy/Dockerfile
args:
VERSION: ${GATEWAY_VERSION:-0.10.0}
environment: *gateway-environment
ports:
- "${GATEWAY_PORT:-8080}:8080"
depends_on:
migrator:
condition: service_completed_successfully
redis-critical:
condition: service_healthy
redis-cache:
condition: service_healthy
minio:
condition: service_started
# 向量化不阻断 API 启动:ollama 未就绪时 knowledge 入库降级(embedding 置 NULL)。
ollama:
condition: service_started
restart: unless-stopped
# M8: 对象存储。MinIO 不暴露主机端口,凭据只留在 API 容器内;
# 上传/下载全部经网关代理。桶由 gateway-api 启动时的 EnsureBucket 兜底创建。
minio:
image: minio/minio:latest
command: ["server", "/data", "--console-address", ":9001"]
environment:
MINIO_ROOT_USER: ${S3_ACCESS_KEY_ID:-gateway}
MINIO_ROOT_PASSWORD: ${S3_SECRET_ACCESS_KEY:-gateway-secret}
volumes:
- minio-data:/data
restart: unless-stopped
# M8 P2:本地向量化。模型默认 bge-m3(~1.2GB),首次 embed 惰性 pull;
# 无需对外暴露端口,仅网关容器访问。restart=unless-stopped 以便重启后自愈。
ollama:
image: ollama/ollama:latest
volumes:
- ollama-models:/root/.ollama
restart: unless-stopped
admin-web:
build:
context: ..
dockerfile: deploy/Dockerfile.web
args:
APP: admin
ports:
- "${ADMIN_PORT:-8081}:80"
depends_on:
gateway-api:
condition: service_started
restart: unless-stopped
portal-web:
build:
context: ..
dockerfile: deploy/Dockerfile.web
args:
APP: portal
ports:
- "${PORTAL_PORT:-8082}:80"
depends_on:
gateway-api:
condition: service_started
restart: unless-stopped
outbox-worker:
build:
context: ..
dockerfile: deploy/Dockerfile
args:
VERSION: ${GATEWAY_VERSION:-0.10.0}
entrypoint: ["gateway-outbox-worker"]
environment: *gateway-environment
depends_on:
migrator:
condition: service_completed_successfully
redis-critical:
condition: service_healthy
restart: unless-stopped
maintenance-worker:
build:
context: ..
dockerfile: deploy/Dockerfile
args:
VERSION: ${GATEWAY_VERSION:-0.10.0}
entrypoint: ["gateway-maintenance"]
environment: *gateway-environment
depends_on:
migrator:
condition: service_completed_successfully
restart: unless-stopped
notification-worker:
build:
context: ..
dockerfile: deploy/Dockerfile
args:
VERSION: ${GATEWAY_VERSION:-0.10.0}
entrypoint: ["gateway-notification-worker"]
environment: *gateway-environment
depends_on:
migrator:
condition: service_completed_successfully
redis-critical:
condition: service_healthy
restart: unless-stopped
volumes:
postgres-data:
redis-critical-data:
minio-data:
ollama-models: