# Return whole documents and further pages, not a pile of chunks

> Return whole documents and further pages, not a pile of chunks: 2 companies fully do it, 7 with a limit, 17 carry the tag. Every listing in Retrieval & vector search with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/vector-search.document-results/
- Markdown: https://www.anchorterminal.com/niches/vector-search.document-results/index.md (~1,150 tokens)
- Slim: https://www.anchorterminal.com/niches/vector-search.document-results/index.min.md (~280 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/vector-search.document-results/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

Two or three companies' docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 2 | 7 | 17 | 8 | 3 |

## What counts as done

Results are grouped by a document field so N distinct documents come back in one call, and page 2 returns the next N with no gaps or repeats. Checked with a fixture of 50 documents of 5 chunks each.

How agents ask for it: top 10 documents about X; show page 2 of the results; one result per product

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **Elasticsearch (Elasticsearch B.V.)**: Nested kNN stores one vector per chunk and returns parent documents ranked by their best chunk, with inner_hits for the passages; collapse also keeps one hit per field value.
- **Meilisearch (Meili SAS)**: A distinct attribute such as a document id returns one chunk per document, with offset and limit for the next page.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Chroma API + MCP (Chroma Inc.)**: GroupBy on a metadata key with limit and offset exists, but only in the Cloud Search API and only with a ranking expression.
- **Milvus and Zilliz Cloud API + MCP (Zilliz Inc.)**: group_by_field (for example docId) returns one or group_size hits per document, but paging to the next N groups is not documented on the grouping page.
- **OpenSearch (OpenSearch Software Foundation, a project of The Linux Foundation)**: Nested k-NN returns parent documents, with k counting documents and inner_hits naming the matching chunks, but paging to the next N documents was not shown.
- **Qdrant API + MCP (Qdrant Solutions GmbH)**: The Grouping API groups chunk hits by a document field, but page 2 needs the docs' exclusion-filter pattern rather than an offset on groups.
- **Typesense API + MCP (Typesense, Inc.)**: group_by makes per_page count groups and page walk them, but #2618 reports pages returning fewer documents than requested after mass deletions.
- **Vespa (Vespa.ai AS)**: The grouping language groups hits by any field, such as a document id, with hits per group, but paging to the next N groups was not confirmed on the page read.
- **Weaviate API + MCP (Weaviate B.V.)**: Hybrid and vector queries take a group-by on a property with group and per-group limits, but paging to the next N groups is not shown; limit and offset apply to objects.

- not this job: Cloudflare AI Search (Cloudflare, Inc.): No group-by or document-level collapse on search; results are chunks, and no pagination parameter appears in the search reference.
- not this job: Epsilla Vector Database (Epsilla, Inc.): No group-by on query; record retrieval pages with skip and limit only.
- not this job: LanceDB (LanceDB Systems, Inc.): No group-by or collapse parameter on vector or hybrid search; only limit and offset.
- not this job: Pinecone API + MCP (Pinecone Systems Inc.): No group-by and no offset or page token on Documents API search; Drupal maintainers overshoot top_k and slice to fake both.
- not this job: Upstash Vector API + MCP (Upstash, Inc.): No group-by on query; resumable queries and range page through raw results.

## What stops agents finishing

- no group-by, so the agent overshoots topK and retries
- no offset or page token on the newer search endpoints

## Around it

Before: vector-search.rag-over-docs. After: search.search-with-text.

## Where this comes from

Weight 8 of 100 in Retrieval & vector search: 3 of 37 accounts (Pinecone group-by and pagination from the Drupal maintainer, Typesense pagination after deletions)

As data: https://www.anchorterminal.com/api/v1/use-cases/vector-search.document-results.json, or the MCP tool get_use_case with id vector-search.document-results. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
