rust/lancedb
1
0
Fork 0
mirror of https://github.com/lancedb/lancedb.git synced 2026-08-18 20:18:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(python): added support for WatsonxReranker component (#3642)

Browse Source 
## Summary

Adds `WatsonxReranker` to the Python bindings, integrating the [IBM
watsonx.ai text rerank
API](https://cloud.ibm.com/docs/apis/watsonx-ai#text-rerank) via the
`ibm_watsonx_ai` SDK (`pip install ibm-watsonx-ai`).

## Parameters

| Parameter | Default | Description |
|---|---|---|
| `model_name` | `"cross-encoder/ms-marco-minilm-l-12-v2"` | Rerank
model ID |
| `column` | `"text"` | Table column used as document input |
| `top_n` | `None` | Return only the top-n results |
| `return_score` | `"relevance"` | `"relevance"` or `"all"` |
| `api_key` | `None` | Falls back to `WATSONX_API_KEY` env var |
| `project_id` | `None` | Falls back to `WATSONX_PROJECT_ID` env var —
mutually exclusive with `space_id` |
| `space_id` | `None` | Falls back to `WATSONX_SPACE_ID` env var —
mutually exclusive with `project_id` |
| `url` | `None` | Defaults to `https://us-south.ml.cloud.ibm.com` |
| `truncate_input_tokens` | `None` | Token truncation limit |

## Usage

```python
from lancedb.rerankers import WatsonxReranker

# credentials from environment variables
reranker = WatsonxReranker()

# or passed explicitly
reranker = WatsonxReranker(
    api_key="<key>",
    project_id="<project-id>",   # or space_id="<space-id>"
    top_n=5,
)
```

## Testing

Integration test added in `test_rerankers.py`, skipped unless
`WATSONX_API_KEY` and one of `WATSONX_PROJECT_ID` / `WATSONX_SPACE_ID`
are set.
This commit is contained in:
Mateusz Szewczyk
2026-07-14 00:58:32 +02:00
committed by GitHub
parent cde48fad95
commit 5b982f2f05
3 changed files with 199 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/python/tests/test_rerankers.py
+17
View File
@@ -23,6 +23,7 @@ from lancedb.rerankers import (
AnswerdotaiRerankers,
VoyageAIReranker,
MRRReranker,
WatsonxReranker,
)
from lancedb.table import LanceTable
@@ -727,3 +728,19 @@ def test_linear_combination_missing_fts_is_penalised():
f"Document with FTS score (rowid 0, {scores[0]:.4f}) should beat "
f"document with no FTS match (rowid 1, {scores[1]:.4f})"
)
@pytest.mark.skipif(
os.environ.get("WATSONX_API_KEY") is None
or (
os.environ.get("WATSONX_PROJECT_ID") is None
and os.environ.get("WATSONX_SPACE_ID") is None
),
reason="WATSONX_API_KEY and one of WATSONX_PROJECT_ID / "
"WATSONX_SPACE_ID must be set",
)
def test_watsonx_reranker(tmp_path):
pytest.importorskip("ibm_watsonx_ai")
table, schema = get_test_table(tmp_path)
reranker = WatsonxReranker()
_run_test_reranker(reranker, table, "single player experience", None, schema)