Documentation
Overview
Spark is smart blob storage that replaces your dumb object store. Store objects with the Spark API, then search and query them with AI.
Make a working request
Live workspace API- 1Export your key
cURL
export SPARK_API_KEY="gsk_<accessKeyId>.<secret>" export SPARK_WORKSPACE_ID="wspQ7WsH2Np5L"Spark showed the bearer token once when you created the key.
- 2List members
cURL
curl \ --url "https://api.spark.graphon.ai/v1/workspaces/$SPARK_WORKSPACE_ID/members" \ --header "Authorization: Bearer $SPARK_API_KEY"A viewer key can read the workspace roster.
- 3Invite a member
cURL
curl --request POST \ --url "https://api.spark.graphon.ai/v1/workspaces/$SPARK_WORKSPACE_ID/invitations" \ --header "Authorization: Bearer $SPARK_API_KEY" \ --header "Content-Type: application/json" \ --data '{"emails":["grace@example.com"],"role":"editor"}'An admin key can invite up to 50 addresses at once.
What is Spark?
Spark is smart blob storage. It replaces S3, R2, and GCS in your stack. Its object API is designed to feel like S3 — buckets, objects, and keys — so it is familiar. Spark is different in one way: as you write data, it reads the data and understands it.
On top of storage, Spark gives you a fast keyword search, a vector (RAG) search, and an agentic query API that answers questions with citations. You drop your separate search stack and keep one dependency.
Why developers use Spark
- Familiar model. Buckets, objects, and keys, like S3. If you know S3, you know Spark's shape.
- Remove a dependency. Retire Elasticsearch, Pinecone, or LanceDB. Search and retrieval live in the storage layer.
- Ask questions, not just fetch bytes. Query a bucket in natural language and get an answer with sources.
- Isolate tenants. Partition data by your own customer with the
X-Spark-Tenantheader.
What you can build with Spark
Point Spark at your files and ship a product on top. A few things teams build:
Graphon company brain
Ask across Slack, docs, and meeting notes — one answer with sources.
49ers app
Fan Q&A over game footage, articles, and stats in one query.