> Source: https://vitna.ai/docs/search/live-engine
> Raw Markdown: https://vitna.ai/docs/search/live-engine.md
> Corpus index: https://vitna.ai/llms.txt
# Migrate Bright Data and Live/Hybrid clients
Vitna retrieves actual Google and Bing organic SERPs through its own
transport and parser at `POST /v1/serp`, with a URL-based compatibility path
at `POST /request`. The former Bright Data search adapter has been removed.
Native indexed search and cited research remain separate capabilities.
This page keeps its existing URL to help older clients migrate.
## Actual Google/Bing results
Change the retrieval URL and bearer key to the Vitna search service. For a
new client, use the explicit SERP contract:
```bash
curl http://127.0.0.1:8789/v1/serp \
-H "Authorization: Bearer $VITNA_SEARCH_KEY" \
-H 'Content-Type: application/json' \
-d '{"engine":"google","query":"lunar science","country":"us","language":"en","maxResults":5,"format":"json"}'
```
For a supported URL-based client, use the compatibility alias:
```bash
curl http://127.0.0.1:8789/request \
-H "Authorization: Bearer $VITNA_SEARCH_KEY" \
-H 'Content-Type: application/json' \
-d '{"url":"https://www.google.com/search?q=lunar+science&hl=en&gl=us","format":"json"}'
```
Both return actual search-engine organic listings in `organic`, with
`rank`, `title`, and `link`, plus optional snippets and display links.
The independent SERP route needs
no Bright Data key, zone, or API call. It does not require a local index or
native crawl, and it does not use indexed pages as an empty-result fallback.
This alpha supports a bounded organic text-search contract. Ads, shopping,
maps, images, news, knowledge panels, asynchronous vendor jobs, and the full
vendor output are not implemented. Country and language parameters are
localization hints; they do not prove the geographic location of egress.
Challenges and unrecognized pages return explicit failures. Review the
[SERP API contract](serp.md) for supported URL parameters, formats, and
operator transport setup.
Former `POST /v1/search` requests with `mode: "live"` or
`mode: "hybrid"` still return HTTP 400 with
`error.code: "invalid_request"`. Move those actual-SERP requests to
`POST /v1/serp` rather than changing them to an indexed query.
## Vitna's indexed corpus
Send `mode: "index"` for native retrieval and omit legacy `provider` and
`fallback`. Search returns ranked indexed pages with source identity,
observed crawl timestamps, and index provenance. It does not return a Google
result list or fetch pages inline.
For bounded coverage expansion, add `expand: true`. Add `discover: true`
to examine selected sites' homepages and sitemaps; domains are optional
guidance. Use `discover: false` to refresh known or supplied targets only.
Name and topic searches can also admit automatic discovery when configured.
```bash
curl https://search.vitna.ai/v1/search \
-H "Authorization: Bearer $VITNA_SEARCH_KEY" \
-H 'Content-Type: application/json' \
-d '{"query":"lunar science","mode":"index","expand":true,"discover":true,"domains":["nasa.gov"]}'
```
Expansion returns a separate private job. Poll
`GET /v1/search/jobs/<id>` with the same client's key, then repeat the query
after completion to use newly indexed content. Completion means the bounded
pass finished; inspect indexed, unchanged, failed, and unmet target counts
for actual coverage.
`POST /v1/research` also accepts `expand` and `discover`.
It assembles cited verbatim passages from matched indexed pages. It accepts
no mode or paid provider fields. See the [API contract](overview.md) for
search filters, passage budgets, quotas, and response fields.
Research now selects from bounded candidate metadata before reading and
rechecks each extracted page's URL and title. A small PostgreSQL, Python,
and Node.js registry identifies known primary domains and canonical version
paths. When the query identifies one requested version, explicit
incompatible versions are excluded; unknown or mixed versions remain
eligible. Version evidence comes from recognized paths or
subject-specific titles, not snippet numbers. Documentation queries prefer
documentation metadata, known primary domains, and compatible versions,
while equal-priority candidates retain their input order. This selection
policy does not verify content facts or establish complete coverage.
## Update the deployment
Remove former Bright Data search credentials and search adapter settings from
the service's private environment. Enable independent SERP retrieval with
`VITNA_SEARCH_SERP_ENABLED=true` and `VITNA_SERP_RESALE_APPROVED=true`.
The second setting now records the operator's decision that collecting Google
and Bing results and redistributing them through Vitna's API is permitted. A
value kept from the Bright Data adapter confirmed a reseller/OEM agreement for
that vendor's service, not this decision, so decide again before keeping it.
Startup refuses SERP retrieval enabled without this confirmation. `auto` is
the default transport. It can
try direct HTTPS followed by browser retrieval, while configured proxies
restrict attempts to pinned browser routes. The browser transport uses
Vitna Unblocker's separately installed runtime. Distinct attempts, route
concurrency, cooldowns, and a short JSON cache bound retrieval work. Send
`fresh: true` for a new capture. It has no CAPTCHA solver. The same
search client registry and request quotas apply to SERP and native requests.
See [SERP setup](serp.md#run-without-an-index).
Native retrieval needs an initialized
`VITNA_SEARCH_INDEX` and `VITNA_SEARCH_PROVIDERS=vitna`, with dedicated
client hashes. Query-guided expansion also needs the separate frontier and
demand worker. Site discovery needs a shared private discovery key and the
relevant local catalogs. Starting the API does not start those workers.
Follow [focused index setup](self-hosted.md) or
[general web coverage expansion](general-web.md#query-guided-coverage-expansion)
for crawler setup, budgets, and timers. The optional Brave, Exa, and Tavily
gateway remains available only through separately configured provider keys
and routing; it is separate from native crawling and can incur provider charges.