Connect products and agents to BMJ content through one stable contract.

Discover the permitted corpus

GET /v1/customer/corpus-summary reports scoped content-type, journal, date, document, evidence and topic coverage. GET /v1/documents browses only documents admitted to the customer’s release and contract scope.

curl 'https://content-gateway.bmj-cloudflare-hackathon.com/v1/customer/corpus-summary' \
  -H "Authorization: Bearer ${BMJ_EVIDENCE_API_KEY}"

curl 'https://content-gateway.bmj-cloudflare-hackathon.com/v1/documents?journal_id=bmjonc&limit=25' \
  -H "Authorization: Bearer ${BMJ_EVIDENCE_API_KEY}"

Resolve editorial collections

GET /v1/collections discovers published static and dynamic collections. Membership is always resolved inside the authenticated customer’s release and contract scope.

curl 'https://content-gateway.bmj-cloudflare-hackathon.com/v1/collections?limit=25' \
  -H "Authorization: Bearer ${BMJ_EVIDENCE_API_KEY}"

curl 'https://content-gateway.bmj-cloudflare-hackathon.com/v1/collections/open-data-research/documents?limit=25' \
  -H "Authorization: Bearer ${BMJ_EVIDENCE_API_KEY}"

Search evidence

POST /v1/search performs semantic, keyword or hybrid retrieval against the customer policy’s active index generation.

curl -X POST 'https://content-gateway.bmj-cloudflare-hackathon.com/v1/search' \
  -H "Authorization: Bearer ${BMJ_EVIDENCE_API_KEY}" \
  -H 'Content-Type: application/json' \
  -H 'X-Request-Id: client-generated-unique-id' \
  --data '{
    "query": "What evidence discusses treatment adherence?",
    "max_results": 5,
    "retrieval": "hybrid",
    "rerank": true
  }'

Build with canonical chunks

GET /v1/chunks/:evidence_id returns a stable BMJ paragraph chunk only when the key scope, independent source-chunk entitlement, release grant and paragraph execution profile all permit it. The add-on can accompany search with or without grounded answers.

Chunk responses include offsets, content hash, citation metadata and the licensed_content_1k_tokens settlement.

Track a rolling answer

POST /v1/rolling-answers is an optional, asynchronous grounded-answer capability. It evaluates a customer-scoped rolling evidence window at monthly checkpoints, records months with no evidence, and then reports deterministic stability/change signals plus a separate trend synthesis.

curl -X POST 'https://content-gateway.bmj-cloudflare-hackathon.com/v1/rolling-answers' \
  -H "Authorization: Bearer ${BMJ_EVIDENCE_API_KEY}" \
  -H 'Content-Type: application/json' \
  -H 'X-Request-Id: client-generated-unique-id' \
  --data '{
    "query": "What does the evidence report about diabetes prevention, treatment and outcomes?",
    "window_months": 1,
    "checkpoint_from": "2025-08",
    "checkpoint_to": "2026-07",
    "max_results": 5,
    "retrieval": "hybrid",
    "rerank": true
  }'

# The POST returns HTTP 202. Poll the relative status_url with the same key.

window_months: 1 makes each checkpoint use only evidence published in that calendar month. This example uses diabetes because the current POC catalogue has diabetes-tagged articles in every month from July 2025 through July 2026. Corpus summary remains authoritative for each customer’s narrower scope.

Request

query

Required string, 2–2,000 characters

max_results

Optional, 1–50; capped by customer policy

retrieval

hybrid, vector or keyword

rerank

Optional boolean; defaults to true

match_threshold

Optional number from 0–1; defaults to 0.35

filters

Optional narrower journal/date filters; cannot broaden contract scope

window_months

Rolling answers only: evidence lookback, constrained by the named offer

checkpoint_from/to

Rolling answers only: optional monthly range in YYYY-MM; publication dates are derived per checkpoint

X-Request-Id

Recommended 8–128 character idempotency key

Response

Results include evidence_id, score, bounded text, item key and safe metadata. The execution_context reports policy, content release, logical generation, physical backend, segmentation and embedding method.

/v1/search returns retrieval directly. /v1/answers adds generation and returns the same sources, explicit component provenance, model-token meters and grounding warnings.