How codeforstartups/dynavec works
dynavec is a Python hybrid vector database for AWS: its client embeds or accepts vectors, stores vectors and full documents in separate AWS services, and searches by retrieving vector candidates then hydrating and ranking results. The repository also includes ingestion, retrieval extensions, GraphRAG, agent and framework integrations, evaluation, MCP, and a separate dashboard. Relationships are limited to README-documented flows and source-index references; unsampled subsystem internals and dashboard-to-backend wiring are not asserted.
Main entry points: client.py Sync client · s3vectors.py S3 Vectors index · dynamodb.py DynamoDB documents · async_client.py Async client
Components 29
Client and ingestion
- Sync client
src/dynavec/client.py - Async client
src/dynavec/async_client.py - Configuration
src/dynavec/config.py - Documents and results
src/dynavec/models.py - Content ingestion
src/dynavec/ingest.py - Embedding backends
src/dynavec/embeddings/base.py - Transform pipeline
src/dynavec/transforms.py - Shared client helpers
src/dynavec/client_common.py
Search and ranking
- Vector retrieval
src/dynavec/retrieval.py - Query expansion and BM25
src/dynavec/retrievers.py - Namespace routing
src/dynavec/namespace.py - Hybrid fusion
src/dynavec/fusion.py - Query cache
src/dynavec/cache.py - In-memory hot tier
src/dynavec/hot.py
AWS data plane
- S3 Vectors index
src/dynavec/stores/s3vectors.py - DynamoDB documents
src/dynavec/stores/dynamodb.py - AWS provisioning
src/dynavec/provisioning.py
Product extensions
- GraphRAG
src/dynavec/graph.py - Agent workflows
src/dynavec/agents - Framework adapters
src/dynavec/integrations - MCP server
src/dynavec/mcp/server.py - Retrieval evaluation
src/dynavec/eval
Observability and UI
- Telemetry
src/dynavec/telemetry.py - Metadata handling
src/dynavec/metadata.py - Structured logging
src/dynavec/logging.py - Dashboard UI
dashboard/app/page.tsx - Dashboard data API
dashboard/lib/api.ts
Other components
- Application user
- AWS services
Connections 22
- Application user → Sync client: upserts and searches
- Sync client → Embedding backends: embeds text
src/dynavec/client.py - Sync client → S3 Vectors index: writes vectors; queries candidates
- Sync client → DynamoDB documents: writes and hydrates documents
- S3 Vectors index → Sync client: returns vector keys
- DynamoDB documents → Sync client: returns hydrated documents
- Sync client → Shared client helpers: uses shared helpers
src/dynavec/client.py - Shared client helpers → Metadata handling: splits metadata
src/dynavec/client_common.py - Sync client → Transform pipeline: applies transforms
src/dynavec/client.py - Sync client → Vector retrieval: uses retrieval helpers
src/dynavec/client.py - Sync client → Query expansion and BM25: exposes retriever support
src/dynavec/client.py - Namespace routing → Sync client: delegates operations
src/dynavec/namespace.py - Query expansion and BM25 → Vector retrieval: uses retrieval helpers
src/dynavec/retrievers.py - Sync client → Query cache: checks query cache
src/dynavec/client.py - Sync client → In-memory hot tier: uses hot tier
src/dynavec/client.py - Content ingestion → Sync client: upserts ingested content
src/dynavec/ingest.py - Configuration → Sync client: configures client
src/dynavec/client.py - Async client → DynamoDB documents: hydrates documents
src/dynavec/async_client.py - Async client → S3 Vectors index: queries vector index
src/dynavec/async_client.py - MCP server → Sync client: exposes search and GraphRAG
src/dynavec/mcp/server.py - Dashboard UI → Dashboard data API: loads dashboard data
dashboard/app/page.tsx - AWS provisioning → AWS services: creates AWS resources