danvics/haiku.rag
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Linting, doc edits

Browse source
This commit is contained in:
Lawrence Akka 2026-08-18 12:04:25 +02:00
parent 3a9ded2218
commit 47feeb32ee
4 changed files with 5 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
CHANGELOG.md
View file

@ -5,11 +5,12 @@
- `evaluations run --filter/-f CLAUSE`: SQL `WHERE` clause over document columns, applied to the retrieval benchmark's searches and to every capability search during QA. Recorded as `document_filter` in experiment metadata. - `evaluations run --filter/-f CLAUSE`: SQL `WHERE` clause over document columns, applied to the retrieval benchmark's searches and to every capability search during QA. Recorded as `document_filter` in experiment metadata.
- `mtrag_clapnq` / `mtrag_clapnq_rewrite` / `mtrag_clapnq_live` / `mtrag_clapnq_live_uncompacted` evaluation datasets: multi-turn QA with gold-prefix and live-session conversation replay, live arms with and without `EvidenceCompactionCapability`, Recall@k/nDCG@k retrieval metrics, eligibility-aware citation scoring, refusal precision/recall, per-turn tool-traffic attributes, and `citation_status` / `turn_citation_status` eval attributes. - `mtrag_clapnq` / `mtrag_clapnq_rewrite` / `mtrag_clapnq_live` / `mtrag_clapnq_live_uncompacted` evaluation datasets: multi-turn QA with gold-prefix and live-session conversation replay, live arms with and without `EvidenceCompactionCapability`, Recall@k/nDCG@k retrieval metrics, eligibility-aware citation scoring, refusal precision/recall, per-turn tool-traffic attributes, and `citation_status` / `turn_citation_status` eval attributes.
- Raw chunk metadata is now exposed to search and citation results, through `SearchResult.chunk_meta` and `Citation.chunk_meta`. For context-expanded results, the metadata is that of the anchor chunk.
### Changed ### Changed
- `import_documents` embeds chunks across the whole batch in one pass instead of per document. - `import_documents` embeds chunks across the whole batch in one pass instead of per document.
- Raw chunk metadata is now exposed to search and citation results, through `SearchResult.chunk_meta` and `Citation.chunk_meta`. For context-expanded results, the metadata is that of the anchor chunk.
### Removed ### Removed

4
docs/custom-pipelines.md
View file

@ -117,9 +117,7 @@ Chunks are returned with:
- `embedding` - `None` (not yet embedded) - `embedding` - `None` (not yet embedded)
- `document_id` - `None` (not yet stored) - `document_id` - `None` (not yet stored)
A custom `DocumentChunker` can provide other keys and values in `metadata`. They will A custom `DocumentChunker` can provide other keys and values in `metadata`. They will be stored with the chunk and are accessible when it is returned in a search result or citation, within `SearchResult.chunk_meta` / `Citation.chunk_meta`.
be stored with the chunk and are accessible when it is returned in a search result or citation, within
`SearchResult.chunk_meta` / `Citation.chunk_meta`.
## Embed ## Embed

2
docs/python.md
View file

@ -196,7 +196,7 @@ for result in results:
print(f"Document ID: {result.document_id}") print(f"Document ID: {result.document_id}")
``` ```
Each result carries the parent document's metadata in `result.document_meta`. It is not shown to the model during QA. Each result carries the parent document's metadata in `result.document_meta` and the relevant chunk's verbatim metadata in `result.chunk_meta`. Neither is shown to the model during QA.
Search with different search types: Search with different search types:
```python ```python

2
haiku_rag_slim/haiku/rag/store/models/chunk.py
View file

@ -133,7 +133,7 @@ class SearchResult(BaseModel):
``document_meta`` carries the parent document's metadata for citation ``document_meta`` carries the parent document's metadata for citation
consumers (UIs). Never part of ``format_for_agent`` output. consumers (UIs). Never part of ``format_for_agent`` output.
``chunk_meta``is the anchor chunk's unparsed ``Chunk.metadata`` and does not ``chunk_meta`` is the anchor chunk's unparsed ``Chunk.metadata`` and does not
include the metadata of any other chunks merged with it. Never part of include the metadata of any other chunks merged with it. Never part of
``format_for_agent`` output. ``format_for_agent`` output.
""" """