Capability · ETL & search

Ingest, index and search, without a vector DB.

Turn URLs, HTML and text into a searchable vector store, and run live web and deep-research search — a stateless white-label of Twin Browser, billed straight through with no markup.

Get started Read the docs

POST /v1/etl · /v1/etl/query · /v1/search

What you get

Built for regulated workloads

ETL and search are a curated, stateless proxy over Twin Browser’s browser-execution API. POST sources to /v1/etl and the upstream extracts, transforms, chunks, embeds and stores them; query them back with /v1/etl/query — no vector database to run. /v1/search adds live web search, and depth:"deep" runs an async research pass you poll at /v1/jobs/{token}. phi-cloud holds no state: the vector store, collections and jobs all live upstream, isolated per customer and billed exactly at the upstream’s rate.

Pipeline

Extract → embed → store

One POST to /v1/etl runs a browser-grade extract, transform, chunk, embed and store over your urls, html or text. Nothing to host — the pipeline and vector store live upstream; you get back a collection you can query.

Retrieval

Semantic query

/v1/etl/query returns the best-matching chunks from your collection with scores — a RAG index without standing up or scaling a vector DB yourself.

Web + deep research

/v1/search returns live web results synchronously; depth:"deep" kicks off an asynchronous research run that you poll at /v1/jobs/{token} until it settles.

Isolation

Per-customer namespace

Every customer’s content is namespaced to its own collection, derived server-side from your API key (never the request body) — a query can only ever reach your own rows. Async jobs return a customer-bound token, not a raw upstream id.

Billing

Pass-through pricing

You pay exactly what the upstream charges, converted credits→USD (1 USD = 1000 credits) with no phi-cloud markup. Async jobs settle once, on the first terminal poll.

PHI

General traffic only

Twin Browser is not a PHI-eligible host, so an X-PHI:true request returns phi_blocked. Paid keys only — ETL and search are off the free test tier so evaluators can’t spend the shared balance.

Availability & pricing

Where it runs, what it costs

Every route is region-resident and the PHI gate is enforced per call. Prices include the flat +10% gateway margin and mirror the live /v1/pricing rate card.

Region	Provider	Model	Tier	Price
All	Twin Browser	ETL ingest · /v1/etl	General	Pass-through
All	Twin Browser	ETL query · /v1/etl/query	General	Pass-through
All	Twin Browser	Web search · /v1/search	General	Pass-through
All	Twin Browser	Deep research · /v1/search (async)	General	Pass-through
Any	—	ETL / search (PHI)	No route	phi_blocked

Billed pass-through — you pay exactly the upstream’s reported credits, converted at 1 USD = 1000 credits with no phi-cloud markup. General traffic only, and paid keys only (not the free test tier).

Try it

A real call, end to end

Ingest a few sources, then query them semantically — no vector DB to run. Everything is scoped to your own collection.

curl

# 1 — ingest sources into a collection
curl https://phi-cloud.com/api/v1/etl \
  -H "Authorization: Bearer $PHI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "sources": ["https://example.com/handbook"], "collection": "handbook" }'

# 2 — semantic query over that collection
curl https://phi-cloud.com/api/v1/etl/query \
  -H "Authorization: Bearer $PHI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "query": "what is the refund window?", "collection": "handbook" }'

response

x-phi-routed: etl-query/EU/non_phi
x-phi-upstream: twin-browser
x-phi-cost-micro: 4000
{ "matches": [{ "text": "Refunds within 30 days...", "score": 0.82 }],
  "collection": "handbook" }

Statelessness & isolation

phi-cloud is a stateless proxy — the vector store, collections and job state all live in Twin Browser, not here. Nothing about your content is persisted by the gateway.
Every customer shares one upstream tenant, isolated by a per-customer collection namespace derived from your API key (never the request body), so a query can only reach your own content.
Async deep-search jobs return an HMAC token bound to your customer id; /v1/jobs/{token} only un-wraps it for the same customer, so job ids can’t be guessed across tenants.
General traffic only: Twin Browser is not PHI-eligible, so X-PHI:true returns phi_blocked. Paid keys only — the free test tier is off.

Good to know

· Deep search is asynchronous — /v1/search with depth:"deep" returns a job token to poll at /v1/jobs/{token}; shallow search is synchronous.
· The captcha-solving and agentic-automation surface (run, live, skills, monitors) is intentionally not exposed — off-posture for a compliance product.

FAQ

Common questions

In Twin Browser, the upstream that runs the pipeline and hosts the vector store. phi-cloud is a stateless proxy — it authenticates your key, namespaces your collection, forwards with our server key and bills pass-through. It holds no vector store, no collection list and no job table.

No. Every customer’s collection is namespaced server-side from your API key identity, never from the request body, so a query is scoped to your namespace and can’t reach another tenant’s rows. Async jobs return a customer-bound token, not the raw upstream job id.

No. Twin Browser is not a PHI-eligible host, so any X-PHI:true request returns phi_blocked — use ETL and search for general content only. It is also paid-keys only; the free test tier is disabled so evaluators can’t spend the shared balance.

Pass-through. You pay exactly the credits Twin Browser reports for the call, converted to USD at 1 USD = 1000 credits with no phi-cloud markup. Async jobs settle once, on the first terminal poll.

Embeddings & search API reference All capabilities

Ready when you are

Put etl & search in production — without giving up your data.

Spin up a key in minutes. The residency and PHI posture above arrives unchanged.

Get started See pricing →

Free to test · Prepaid credits, no subscription · No data retained