+ dynavec / Docs
- dynavec documentation
-A serverless hybrid vector database that runs inside your own AWS account.
+The agent stack in your AWS account
+Open-source, serverless memory, orchestration, observability and evaluation — one Python library.
-dynavec fuses Amazon DynamoDB (single-digit-millisecond -metadata & document store) with Amazon S3 Vectors -(billion-scale serverless ANN) into one Python client — a drop-in alternative to Pinecone, Qdrant, Milvus, -Weaviate, and OpenSearch that bills only when you use it.
-New here? Start with Installation then the
-Quickstart. Every feature has its own page in the sidebar.
-Explore by feature
+dynavec starts as a hybrid vector database that fuses Amazon DynamoDB +(single-digit-millisecond document store) with Amazon S3 Vectors +(billion-scale serverless ANN), and grows into agents, traces and evals. Everything runs in your AWS +account and bills only when you use it — no external control plane.
+Agent lifecycle
+
+Available
+Remember · dynavec
Store and search memory: hybrid BM25 + ANN search, GraphRAG, hot tier and caching.
Get started › +In progressBuild · dynaflow
Compose agents: chat models, ReAct and planner, tool registry with MCP, supervisor routing.
Get started › +EarlyObserve · dynalogs
Structured per-run and per-node logs with secret redaction, plus a telemetry dashboard.
Get started › +PartialMeasure · dynaevals
Score retrieval and answers with recall@k, MRR, nDCG and LLM-judged quality.
Get started › +Start here
+
+
+Installation
pip or uv, plus optional embedder and framework extras.
Install › +Quickstart
Your first upsert and semantic search in a few lines.
Start › +Configuration
Every DynavecConfig option, with sensible defaults.
Configure › +Guides
Embeddings
Pluggable, bring-your-own-key: OpenAI, Gemini, Cohere, Bedrock, local.
Search
Semantic search with metadata pre-filtering and document hydration.
@@ -103,8 +125,10 @@Explore by feature
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/installation.html b/opensource/dynavec/docs/installation.html
index 8da8239..dceb243 100644
--- a/opensource/dynavec/docs/installation.html
+++ b/opensource/dynavec/docs/installation.html
@@ -6,33 +6,44 @@
Installation · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/integrations.html b/opensource/dynavec/docs/integrations.html
index aa1b27f..aabeea5 100644
--- a/opensource/dynavec/docs/integrations.html
+++ b/opensource/dynavec/docs/integrations.html
@@ -6,33 +6,44 @@
Framework integrations · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/knowledge-graph.html b/opensource/dynavec/docs/knowledge-graph.html
index bd0c976..543db26 100644
--- a/opensource/dynavec/docs/knowledge-graph.html
+++ b/opensource/dynavec/docs/knowledge-graph.html
@@ -6,33 +6,44 @@
Knowledge graph · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/metrics-and-rerank.html b/opensource/dynavec/docs/metrics-and-rerank.html
index 8453d55..bf55294 100644
--- a/opensource/dynavec/docs/metrics-and-rerank.html
+++ b/opensource/dynavec/docs/metrics-and-rerank.html
@@ -6,33 +6,44 @@
Metrics & rerank · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/namespaces.html b/opensource/dynavec/docs/namespaces.html
index d77cf1c..35e9390 100644
--- a/opensource/dynavec/docs/namespaces.html
+++ b/opensource/dynavec/docs/namespaces.html
@@ -6,33 +6,44 @@
Namespaces · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/quantization.html b/opensource/dynavec/docs/quantization.html
index 850cbbc..8160bc2 100644
--- a/opensource/dynavec/docs/quantization.html
+++ b/opensource/dynavec/docs/quantization.html
@@ -6,33 +6,44 @@
Product quantization · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/quickstart.html b/opensource/dynavec/docs/quickstart.html
index cd12af8..1f5a52f 100644
--- a/opensource/dynavec/docs/quickstart.html
+++ b/opensource/dynavec/docs/quickstart.html
@@ -6,33 +6,44 @@
Quickstart · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/release-notes.html b/opensource/dynavec/docs/release-notes.html
index ac8aaa9..bc8e08e 100644
--- a/opensource/dynavec/docs/release-notes.html
+++ b/opensource/dynavec/docs/release-notes.html
@@ -6,33 +6,44 @@
Release notes · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/search-index.js b/opensource/dynavec/docs/search-index.js
new file mode 100644
index 0000000..9a5a435
--- /dev/null
+++ b/opensource/dynavec/docs/search-index.js
@@ -0,0 +1,2 @@
+/* generated by tools/build_docs.py */
+window.DV_SEARCH = [{"u": "index.html", "t": "Overview", "g": "Getting started", "d": "Open-source, serverless memory, orchestration, observability and evaluation — one Python library.", "x": "dynavec starts as a hybrid vector database that fuses Amazon DynamoDB (single-digit-millisecond document store) with Amazon S3 Vectors (billion-scale serverless ANN), and grows into agents, traces and evals. Everything runs in your AWS account and bills only when you use it — no external control plane. Agent lifecycle Available Remember · dynavec Store and search memory: hybrid BM25 + ANN search, GraphRAG, hot tier and caching. Get started › In progress Build · dynaflow Compose agents: chat models, ReAct and planner, tool registry with MCP, supervisor routing. Get started › Early Observe · dyn"}, {"u": "index.html#agent-lifecycle", "t": "Agent lifecycle", "g": "Overview", "d": "", "x": ""}, {"u": "index.html#start-here", "t": "Start here", "g": "Overview", "d": "", "x": ""}, {"u": "index.html#guides", "t": "Guides", "g": "Overview", "d": "", "x": ""}, {"u": "installation.html", "t": "Installation", "g": "Getting started", "d": "Base install is boto3 + numpy. Everything else is an optional extra.", "x": "Install with pip or uv — both pull from PyPI. # base (boto3 + numpy only) pip install dynavec uv add dynavec # with an embedder pip install \"dynavec[openai]\" pip install \"dynavec[sentence-transformers]\" # everything (all embedders + framework adapters) pip install \"dynavec[all]\" Optional extras Extra Adds openai , gemini , cohere hosted embedders (bring your own key) sentence-transformers local / offline embedder, also used for cross-encoder rerank langchain , llamaindex , crewai framework vector stores / tools redis RedisCache (AWS ElastiCache) mcp MCP client ingestion benchmark pandas / matp"}, {"u": "installation.html#optional-extras", "t": "Optional extras", "g": "Installation", "d": "", "x": ""}, {"u": "quickstart.html", "t": "Quickstart", "g": "Getting started", "d": "Provision, upsert, and search in a dozen lines.", "x": "With auto_provision=True , dynavec creates the S3 vector bucket, the vector index, and the DynamoDB table on first use. from dynavec import Dynavec, DynavecConfig, Document from dynavec.embeddings import OpenAIEmbedder cfg = DynavecConfig( vector_bucket=\"my-vectors\", index=\"docs\", table=\"dynavec_docs\", dimension=1536, region=\"us-east-1\", auto_provision=True, ) db = Dynavec(cfg, embedder=OpenAIEmbedder(model=\"text-embedding-3-small\")) db.upsert([ Document(id=\"a\", text=\"Mitochondria power the cell.\", metadata={\"topic\": \"bio\"}), Document(id=\"b\", text=\"Rockets reach orbit near 28,000 km/h.\", metad"}, {"u": "configuration.html", "t": "Configuration", "g": "Getting started", "d": "Everything the client needs, in one frozen dataclass.", "x": "DynavecConfig is an immutable description of your resources and tuning. It never holds secrets — credentials are passed separately (see Credentials & IAM ). from dynavec import DynavecConfig cfg = DynavecConfig( vector_bucket=\"my-vectors\", # S3 vector bucket name index=\"docs\", # vector index name table=\"dynavec_docs\", # DynamoDB table name dimension=1536, # must match your embedder distance_metric=\"cosine\", # \"cosine\" or \"euclidean\" (S3 Vectors native) region=\"us-east-1\", filterable_keys=[\"topic\"], # metadata keys pushed to S3 Vectors for filtering store_text_in_s3vectors=True, # mirror trunca"}, {"u": "configuration.html#key-parameters", "t": "Key parameters", "g": "Configuration", "d": "", "x": ""}, {"u": "faq.html", "t": "FAQ", "g": "Getting started", "d": "Common questions about regions, limits, consistency, costs, and architecture.", "x": "Regions & Availability Which AWS regions are supported? dynavec runs in any AWS region where both Amazon S3 Vectors and Amazon DynamoDB are supported (for example, us-east-1 , us-west-2 , ap-south-1 , eu-west-1 , and others). Specify your target region in DynavecConfig(region=\"...\") . Can I query across multiple AWS regions? Each DynavecConfig connects to a single AWS region. Keeping your S3 vector bucket and DynamoDB table in the same region as your application workloads (e.g. Lambda, ECS, EC2) ensures single-digit millisecond latency and eliminates cross-region data transfer fees. If you nee"}, {"u": "faq.html#regions-availability", "t": "Regions & Availability", "g": "FAQ", "d": "", "x": ""}, {"u": "faq.html#limits-constraints", "t": "Limits & Constraints", "g": "FAQ", "d": "", "x": ""}, {"u": "faq.html#consistency-latency", "t": "Consistency & Latency", "g": "FAQ", "d": "", "x": ""}, {"u": "faq.html#costs-billing", "t": "Costs & Billing", "g": "FAQ", "d": "", "x": ""}, {"u": "faq.html#security-privacy-architecture", "t": "Security, Privacy & Architecture", "g": "FAQ", "d": "", "x": ""}, {"u": "embeddings.html", "t": "Embeddings", "g": "Writing data", "d": "Pluggable, bring-your-own-key — or bring your own vectors.", "x": "Choose an embedder and supply your own API key, or skip the embedder entirely and pass pre-computed vectors. Embedder backends are imported lazily, so the base install stays light. from dynavec.embeddings import ( OpenAIEmbedder, GeminiEmbedder, MistralEmbedder, BedrockEmbedder, SentenceTransformerEmbedder, ) # hosted (BYO key via env var or argument) emb = OpenAIEmbedder(model=\"text-embedding-3-small\") # 1536-d emb = GeminiEmbedder(model=\"text-embedding-004\") # 768-d emb = MistralEmbedder(model=\"mistral-embed\") # 1024-d # in-account (no third party) or fully local / offline emb = BedrockEmbed"}, {"u": "embeddings.html#bring-your-own-vectors", "t": "Bring your own vectors", "g": "Embeddings", "d": "", "x": ""}, {"u": "upsert.html", "t": "Upsert", "g": "Writing data", "d": "Write documents to both stores in one call.", "x": "Each document carries an id , either text (which gets embedded) or a vector , and optional metadata . dynavec splits metadata: a small filterable subset goes to S3 Vectors, the full copy plus text goes to DynamoDB. from dynavec import Document db.upsert([ Document(id=\"1\", text=\"apple pie recipe\", metadata={\"cat\": \"food\", \"rating\": 5}), Document(id=\"2\", text=\"rocket launch schedule\", metadata={\"cat\": \"space\"}), ], namespace=\"kb\", auto_metadata=True) The metadata switch You provide metadata — stored verbatim. auto_metadata=True — dynavec also derives created_at , content_hash , word_count , char"}, {"u": "upsert.html#the-metadata-switch", "t": "The metadata switch", "g": "Upsert", "d": "", "x": ""}, {"u": "upsert.html#document-size-limit", "t": "Document size limit", "g": "Upsert", "d": "", "x": ""}, {"u": "update-and-lambda.html", "t": "Update & Lambda", "g": "Writing data", "d": "Change text, vector, or metadata — and transform data in-account.", "x": "update() is a read-modify-write: metadata merges by default, and the vector is only re-derived when the text changes or you pass a new vector. # merge new metadata, keep existing text + vector db.update(\"1\", namespace=\"kb\", metadata={\"rating\": 4}) # change text -> re-embbeds and overwrites the vector db.update(\"1\", namespace=\"kb\", text=\"new content\") Transform pipeline Transforms are plain callables run on each document before it is written — for enrichment, redaction, or deriving vectors elsewhere. from dynavec.transforms import TransformPipeline def redact(ctx): ctx.metadata[\"pii\"] = False r"}, {"u": "update-and-lambda.html#transform-pipeline", "t": "Transform pipeline", "g": "Update & Lambda", "d": "", "x": ""}, {"u": "update-and-lambda.html#run-the-transform-in-your-own-aws-lambda", "t": "Run the transform in your own AWS Lambda", "g": "Update & Lambda", "d": "", "x": ""}, {"u": "ingestion.html", "t": "Ingestion & MCP", "g": "Writing data", "d": "Suck in content from anywhere — including any MCP server.", "x": "A source is any iterable of records. ingest() chunks, embeds, and upserts them. from dynavec.ingest import ingest, IterableSource src = IterableSource([ {\"id\": \"doc1\", \"text\": long_text, \"metadata\": {\"src\": \"wiki\"}}, ]) ingest(db, src, namespace=\"kb\", chunk_size=1000, overlap=150) From a Markdown or text directory MarkdownSource reads UTF-8 .md and .txt files recursively. Install dynavec[ingest] for YAML front matter support. from dynavec.ingest import MarkdownSource, ingest source = MarkdownSource(\"./notes\") # Or select files with a root-relative glob: source = MarkdownSource(\"./notes\", glob="}, {"u": "ingestion.html#from-a-markdown-or-text-directory", "t": "From a Markdown or text directory", "g": "Ingestion & MCP", "d": "", "x": ""}, {"u": "ingestion.html#from-any-mcp-server", "t": "From any MCP server", "g": "Ingestion & MCP", "d": "", "x": ""}, {"u": "search.html", "t": "Search", "g": "Searching", "d": "ANN in S3 Vectors, document hydration from DynamoDB.", "x": "Provide a query string (embedded for you) or a raw vector . S3 Vectors returns the nearest keys; dynavec hydrates the full documents from DynamoDB via BatchGetItem . hits = db.search( \"how do cells make energy?\", top_k=5, namespace=\"kb\", filter={\"topic\": \"biology\"}, # metadata pre-filter (S3 Vectors) ) for h in hits: print(h.score, h.id, h.text, h.metadata) Metadata filtering Filters use the S3 Vectors dialect — bare {\"k\": v} is equality; operators like $gte , $in , $and , $or are supported. Only keys in filterable_keys can be filtered. db.search(\"q\", filter={\"$and\": [{\"topic\": \"bio\"}, {\"year\""}, {"u": "search.html#metadata-filtering", "t": "Metadata filtering", "g": "Search", "d": "", "x": ""}, {"u": "metrics-and-rerank.html", "t": "Metrics & rerank", "g": "Searching", "d": "Cosine, dot, euclidean, manhattan, weighted combinations, and MMR.", "x": "The S3 Vectors index metric is cosine or euclidean. On top of the returned candidates, dynavec can rescore with any metric — or a weighted combination — client-side. # single metric db.search(\"q\", top_k=5, rescore=\"manhattan\") # weighted combination (normalized per candidate set) db.search(\"q\", top_k=5, rescore={\"cosine\": 0.7, \"dot\": 0.3}) MMR diversity rerank Maximal Marginal Relevance balances relevance against diversity so results are not near-duplicates. db.search(\"q\", top_k=5, rerank=\"mmr\", mmr_lambda=0.5) # 1=relevance, 0=diversity Both over-fetch top_k * over_fetch candidates first. You"}, {"u": "metrics-and-rerank.html#mmr-diversity-rerank", "t": "MMR diversity rerank", "g": "Metrics & rerank", "d": "", "x": ""}, {"u": "namespaces.html", "t": "Namespaces", "g": "Searching", "d": "Multi-tenant / multi-collection isolation on a single index.", "x": "A namespace gives each tenant an isolated view of the same dynavec index. Every vector is tagged with its namespace, and searches are automatically scoped to that namespace. This lets multiple customers share the same infrastructure while keeping their retrieval results separated. Namespace-per-tenant Use db.namespace(\"tenant-id\") to create a lightweight handle bound to one tenant. The same document ID can exist in multiple namespaces because the namespace is part of the storage key. acme = db.namespace(\"acme\") globex = db.namespace(\"globex\") acme.upsert([Document(id=\"refund-policy\", text=\"Acm"}, {"u": "namespaces.html#namespace-per-tenant", "t": "Namespace-per-tenant", "g": "Namespaces", "d": "", "x": ""}, {"u": "namespaces.html#end-to-end-multi-tenant-rag", "t": "End-to-end multi-tenant RAG", "g": "Namespaces", "d": "", "x": ""}, {"u": "namespaces.html#runnable-example", "t": "Runnable example", "g": "Namespaces", "d": "", "x": ""}, {"u": "streaming.html", "t": "Streaming", "g": "Searching", "d": "Deliver results to agents page-by-page as they arrive.", "x": "search_stream() is a generator: it yields one hit at a time as S3 Vectors paginates, so an agent can start consuming the first results before the full set returns. Amazon S3 Vectors returns at most 100 vectors per response page; dynavec follows nextToken up to the service limit of 10,000 results. for hit in db.search_stream(\"large query\", top_k=250, namespace=\"kb\", page_size=50): handle(hit) # one hit at a time; page_size is the DynamoDB hydration batch page_size (or DynavecConfig(top_k_page_size=50) ) only changes how many hits are hydrated from DynamoDB per batch. It does not change the S3 V"}, {"u": "caching.html", "t": "Caching", "g": "Searching", "d": "Skip the vector DB for repeated or similar queries.", "x": "Attach a cache and repeated queries are served without hitting S3 Vectors. Three backends: from dynavec import Dynavec, SemanticCache, DynamoDBCache, RedisCache # 1) in-process semantic cache — also serves near-duplicate queries db = Dynavec(cfg, embedder=emb, cache=SemanticCache(threshold=0.97)) # 2) durable, shared cache with TTL jitter to spread simultaneous expiry db = Dynavec( cfg, embedder=emb, cache=DynamoDBCache(cfg, ttl_seconds=3600, ttl_jitter_seconds=300), ) # 3) sub-millisecond shared cache on Redis / AWS ElastiCache db = Dynavec(cfg, embedder=emb, cache=RedisCache(\"redis://my-elas"}, {"u": "knowledge-graph.html", "t": "Knowledge graph", "g": "Advanced", "d": "Attach meaning to embeddings and traverse it to guide search.", "x": "Alongside the vector index, dynavec keeps a lightweight entity-relationship graph in DynamoDB. Entities link to documents; you can traverse the graph first (cheap key lookups) to gather a candidate set, then rank only those against the query embedding. That is the DynamoDB → S3 Vectors reference join. # build the graph db.graph_add_edge(\"acme\", \"competes_with\", \"globex\", namespace=\"kb\") db.graph_link(\"acme\", [\"doc-1\", \"doc-2\"], namespace=\"kb\") # GraphRAG: traverse from seeds, then rank related docs by the query hits = db.graph_search( \"recent product launches\", seed_entities=[\"acme\"], hops=2, "}, {"u": "knowledge-graph.html#removing-nodes-and-edges", "t": "Removing nodes and edges", "g": "Knowledge graph", "d": "", "x": ""}, {"u": "hot-tier.html", "t": "In-memory hot tier", "g": "Advanced", "d": "Pinecone-class latency for the hot working set — without a paid cluster.", "x": "Amazon S3 Vectors is cheap and serverless, but it is an object-backed ANN: its per-query server time is hundreds of milliseconds. For the hot working set, dynavec can keep vectors in RAM (via the built-in SPFreshHotIndex ) so a warmed namespace is served entirely from memory — no S3 Vectors query and no DynamoDB hydration, since the hot index already holds text and metadata. That collapses p50 from hundreds of ms to sub-millisecond, and it costs nothing extra: the index lives in the compute you already run. Correctness first. A namespace is served from RAM only when it is authoritative — every"}, {"u": "hot-tier.html#enable-it", "t": "Enable it", "g": "In-memory hot tier", "d": "", "x": ""}, {"u": "hot-tier.html#how-it-works", "t": "How it works", "g": "In-memory hot tier", "d": "", "x": ""}, {"u": "quantization.html", "t": "Product quantization", "g": "Advanced", "d": "Compress cached vectors up to 32× with asymmetric distance.", "x": "S3 Vectors stores float32 and manages its own layout, so PQ does not change what it stores. PQ compresses the vectors dynavec caches — the in-memory hot tier and local candidate caches — turning a dim × 4 byte vector into m bytes. from dynavec import ProductQuantizer pq = ProductQuantizer(m=96, nbits=8).fit(training_vectors) # 768-d -> 96 bytes (32x) codes = pq.encode(vectors) # uint8 codes dists = pq.asymmetric_distances(query, codes) # ADC, fast at scale print(pq.reconstruction_error(vectors)) # Persist and reload codebooks across restarts pq.save(\"pq_model.npz\") loaded_pq = ProductQuantizer"}, {"u": "concurrency.html", "t": "Concurrency", "g": "Advanced", "d": "GIL-aware thread pool for I/O-bound AWS calls.", "x": "dynavec's workload is I/O-bound (network calls to AWS), and Python releases the GIL during those calls — so a thread pool gives real parallelism without an async rewrite. Batched writes fan out across threads, and search_many runs several queries concurrently. # many queries at once results = db.search_many([\"q1\", \"q2\", \"q3\"], top_k=5, namespace=\"kb\") # tune the pool cfg = DynavecConfig(..., max_workers=16, parallel_writes=True) # clean up the pool (or use the client as a context manager) with Dynavec(cfg, embedder=emb) as db: ... A native asyncio client ( aioboto3 ) is on the roadmap for very"}, {"u": "credentials.html", "t": "Credentials & IAM", "g": "Advanced", "d": "Connect to your account, with least-privilege permissions.", "x": "dynavec uses the standard boto3 credential chain, so exported env vars just work. You can also pass keys explicitly or assume a cross-account role. from dynavec import Dynavec, AWSCredentials # explicit keys / profile / cross-account role creds = AWSCredentials( access_key_id=\"AKIA...\", secret_access_key=\"...\", region=\"us-east-1\", # profile_name=\"prod\", # assume_role_arn=\"arn:aws:iam::OTHER_ACCOUNT:role/dynavec\", ) db = Dynavec(cfg, credentials=creds) Step-by-step: create the IAM user & keys AWS Console → IAM → Users → Create user . Name it dynavec (programmatic access only — no console sign-i"}, {"u": "credentials.html#step-by-step-create-the-iam-user-keys", "t": "Step-by-step: create the IAM user & keys", "g": "Credentials & IAM", "d": "", "x": ""}, {"u": "credentials.html#least-privilege-iam-policy", "t": "Least-privilege IAM policy", "g": "Credentials & IAM", "d": "", "x": ""}, {"u": "integrations.html", "t": "Framework integrations", "g": "Ecosystem", "d": "LangChain, LlamaIndex, and a tool for any agent framework.", "x": "LangChain from dynavec.integrations.langchain import DynavecVectorStore store = DynavecVectorStore(db, namespace=\"kb\") retriever = store.as_retriever(search_kwargs={\"k\": 4}) LlamaIndex from dynavec.integrations.llamaindex import DynavecLlamaStore from llama_index.core import VectorStoreIndex, StorageContext store = DynavecLlamaStore(db, namespace=\"kb\") ctx = StorageContext.from_defaults(vector_store=store) index = VectorStoreIndex.from_documents(docs, storage_context=ctx) LangGraph / CrewAI / Strands A framework-agnostic retriever tool — just a callable that takes a query and returns text. fro"}, {"u": "integrations.html#langchain", "t": "LangChain", "g": "Framework integrations", "d": "", "x": ""}, {"u": "integrations.html#llamaindex", "t": "LlamaIndex", "g": "Framework integrations", "d": "", "x": ""}, {"u": "integrations.html#langgraph-crewai-strands", "t": "LangGraph / CrewAI / Strands", "g": "Framework integrations", "d": "", "x": ""}, {"u": "integrations.html#fastmcp-server-claude-desktop-cursor-ai-agents", "t": "FastMCP server (Claude Desktop, Cursor, AI agents)", "g": "Framework integrations", "d": "", "x": ""}, {"u": "benchmarking.html", "t": "Benchmarking", "g": "Ecosystem", "d": "Recall, latency, and cost — with tables and charts.", "x": "The suite measures recall@k and latency against a labeled dataset and models cost versus Pinecone, OpenSearch, Qdrant, Weaviate, and Milvus across dimensions and scale. pip install \"dynavec[benchmark]\" # recall + latency (local control, or real AWS) python -m benchmarks.run_benchmark --backend local --n 50000 --dim 384 python -m benchmarks.run_benchmark --backend dynavec --bucket my-vectors \\ --index bench --table dynavec_bench --n 100000 --dim 1536 # cost comparison + charts (dimensions 384-3072, 100K -> 1B vectors) python -m benchmarks.report --qpm 1_000_000 Cost figures are cost-model estim"}, {"u": "release-notes.html", "t": "Release notes", "g": "About", "d": "What changed in each version of dynavec, newest first.", "x": "dynavec follows semantic versioning . Upgrade with: pip install --upgrade dynavec uv pip install --upgrade dynavec The full machine-readable history lives in CHANGELOG.md , and every version is a GitHub Release . For what's coming next, see the Ecosystem overview and its phased roadmap. 0.7.0 — 2026-10-01 Concurrency and scale: a native async client, a lexical/hybrid retrieval path, per-document TTL, a new framework connector, and a graph data-loss fix. Added Native async client — AsyncDynavec with async / await upsert / search / get / delete for concurrent agent and web workloads. BM25 lexica"}, {"u": "release-notes.html#v0-7-0", "t": "0.7.0 — 2026-10-01", "g": "Release notes", "d": "", "x": ""}, {"u": "release-notes.html#v0-6-0", "t": "0.6.0 — 2026-09-25", "g": "Release notes", "d": "", "x": ""}, {"u": "release-notes.html#v0-5-0", "t": "0.5.0 — 2026-09-16", "g": "Release notes", "d": "", "x": ""}, {"u": "release-notes.html#v0-4-0", "t": "0.4.0 — 2026-09-12", "g": "Release notes", "d": "", "x": ""}, {"u": "release-notes.html#v0-3-0", "t": "0.3.0 — 2026-09-10", "g": "Release notes", "d": "", "x": ""}, {"u": "release-notes.html#v0-2-0", "t": "0.2.0 — 2026-08", "g": "Release notes", "d": "", "x": ""}, {"u": "dashboard.html", "t": "Telemetry dashboard", "g": "Ecosystem", "d": "A native, in-your-brand observability dashboard — a Langfuse-style view of real query telemetry.", "x": "Attach a recorder to your client and every search is captured with latency, cache outcome, result count, and score stats. No simulated data. Live interactive preview: codeforstartups.github.io/dynavec/dashboard (landing-page theme). 1. Expose real telemetry Attach a recorder to your client and serve the API: from dynavec import Dynavec, DynavecConfig, SemanticCache from dynavec.telemetry import TelemetryRecorder from dynavec.dashboard import serve rec = TelemetryRecorder() db = Dynavec(cfg, embedder=emb, cache=SemanticCache(), telemetry=rec) # ... your app runs searches; the recorder fills aut"}, {"u": "dashboard.html#1-expose-real-telemetry", "t": "1. Expose real telemetry", "g": "Telemetry dashboard", "d": "", "x": ""}, {"u": "dashboard.html#2-run-the-dashboard", "t": "2. Run the dashboard", "g": "Telemetry dashboard", "d": "", "x": ""}, {"u": "dashboard.html#tracing-view", "t": "Tracing view", "g": "Telemetry dashboard", "d": "", "x": ""}, {"u": "ecosystem.html", "t": "Ecosystem overview", "g": "Ecosystem", "d": "Four projects that together cover remembering, running, observing and measuring an AI application.", "x": "dynavec started as a vector database, and it's growing into a full agent stack that runs inside your own cloud account. The plan has four parts. Each one answers a question about an AI application: what does it know, what does it do, what happened when it ran, and how good were the results? This page follows the ecosystem epics ( #273 – #276 ). It's a living plan: the order is a proposal rather than a fixed schedule, and it will change as work lands. The four projects Project Tagline What it covers Status dynavec Remember The vector database: S3 Vectors for search, DynamoDB for documents, plus"}, {"u": "ecosystem.html#the-four-projects", "t": "The four projects", "g": "Ecosystem overview", "d": "", "x": ""}, {"u": "ecosystem.html#how-they-fit-together", "t": "How they fit together", "g": "Ecosystem overview", "d": "", "x": ""}, {"u": "ecosystem.html#what-you-can-use-today", "t": "What you can use today", "g": "Ecosystem overview", "d": "", "x": ""}, {"u": "ecosystem.html#roadmap", "t": "Roadmap", "g": "Ecosystem overview", "d": "", "x": ""}];
diff --git a/opensource/dynavec/docs/search.html b/opensource/dynavec/docs/search.html
index 3d7c0ba..82a1e6e 100644
--- a/opensource/dynavec/docs/search.html
+++ b/opensource/dynavec/docs/search.html
@@ -6,33 +6,44 @@
Search · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/streaming.html b/opensource/dynavec/docs/streaming.html
index e3f5d69..1c72f20 100644
--- a/opensource/dynavec/docs/streaming.html
+++ b/opensource/dynavec/docs/streaming.html
@@ -6,33 +6,44 @@
Streaming · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/update-and-lambda.html b/opensource/dynavec/docs/update-and-lambda.html
index 52c2427..ea12c6d 100644
--- a/opensource/dynavec/docs/update-and-lambda.html
+++ b/opensource/dynavec/docs/update-and-lambda.html
@@ -6,33 +6,44 @@
Update & Lambda transforms · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/docs/upsert.html b/opensource/dynavec/docs/upsert.html
index 3970ff0..eb74780 100644
--- a/opensource/dynavec/docs/upsert.html
+++ b/opensource/dynavec/docs/upsert.html
@@ -6,33 +6,44 @@
Upsert · dynavec docs
-
+
+
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star —
+
+
+
+
+
diff --git a/opensource/dynavec/favicon.svg b/opensource/dynavec/favicon.svg
new file mode 100644
index 0000000..1bf64da
--- /dev/null
+++ b/opensource/dynavec/favicon.svg
@@ -0,0 +1,16 @@
+
diff --git a/opensource/dynavec/images/brand/dynavec-logo-dark.svg b/opensource/dynavec/images/brand/dynavec-logo-dark.svg
new file mode 100644
index 0000000..9c11e26
--- /dev/null
+++ b/opensource/dynavec/images/brand/dynavec-logo-dark.svg
@@ -0,0 +1,6 @@
+
diff --git a/opensource/dynavec/images/brand/dynavec-logo-light.svg b/opensource/dynavec/images/brand/dynavec-logo-light.svg
new file mode 100644
index 0000000..4febd61
--- /dev/null
+++ b/opensource/dynavec/images/brand/dynavec-logo-light.svg
@@ -0,0 +1,6 @@
+
diff --git a/opensource/dynavec/images/brand/dynavec-mark.svg b/opensource/dynavec/images/brand/dynavec-mark.svg
new file mode 100644
index 0000000..1bf64da
--- /dev/null
+++ b/opensource/dynavec/images/brand/dynavec-mark.svg
@@ -0,0 +1,16 @@
+
diff --git a/opensource/dynavec/index.html b/opensource/dynavec/index.html
index cce5601..b708f72 100644
--- a/opensource/dynavec/index.html
+++ b/opensource/dynavec/index.html
@@ -3,169 +3,462 @@
- dynavec — serverless hybrid vector database on AWS
-
+ dynavec — the serverless agent stack in your AWS account
+
+
+
+
+
+
-
+
+
-
-
-
-
Skip to content
-
-
-
+
+
+
+
+
+
+ esc
+
+
-
-
- dynavec
-
-
-
- Star
- 27
+
+
+
-
-
-
+
+ 
+ dynavec
+
+
+
+
+ dynavec
+
+
+
+
+
+
+ Star —
+
+ Get started
+
+
-
-
- The serverless vector database
-
Open source · Apache-2.0 · runs in your AWS account
-The serverless vector database
that lives in your AWS account.
- - dynavec fuses Amazon DynamoDB (single-digit-millisecond metadata & document store) - with Amazon S3 Vectors (billion-scale, serverless ANN) into one Python client — - a drop-in alternative to Pinecone, Qdrant, Milvus, Weaviate, and OpenSearch that - bills only when you use it and keeps every byte in your own region. -
- -
-
+
+
+
-
-
-
+
+
+ New dynaflow agents, tool registry & supervisor routing →
+
-
-
- The agent stack that lives in your AWS account
++ Start with dynavec — a serverless hybrid vector database on Amazon DynamoDB + Amazon S3 Vectors — and grow into + agents, traces and evals. One open-source Python library. Pay only when you use it; every byte stays in your region. +
+
+ Get started
+ Read the docs
+
+ pip install dynavec
-
+
-
- uv add dynavec
-
+
+
+
-
+
+
+ 
+
+
+
+
+
+
- Read the docs
- View on GitHub
+
- -
+
- —GitHub stars +
- $469/mo at 1B vectors* +
- 100%in your account +
- 0servers to run +
-
-
-
-
+
+
+
-
-
+
+
-
- Built on AWS · works with the tools you already use
+
+
-
+ Amazon DynamoDBAmazon S3 VectorsAmazon BedrockAWS Lambda
+ LangChainLlamaIndexCrewAIHaystackDSPySemantic KernelOpenAIAnthropicMCP
+
+
+
-
-
- —GitHub stars -
- $469/mo at 1B vectors* -
- 100%in your account -
- 0servers to run -
-
Built on two AWS primitives you already trust
-
-
-
-