danvics/haiku.rag
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
haiku.rag/tests
History
Yiorgis Gozadinos d63199d96d
Keep cross-encoder rerank scores apart when they saturate 
mxbai-rerank-base-v2 ships a Sigmoid activation and evaluates it in bf16, so
every strongly-relevant candidate rounds to exactly 1.0. Ties then leave the
order to the stable sort, which preserves the incoming hybrid ranking: on 100
t2_finqa retrieval cases the reranker scored MAP 0.661 against 0.659 with no
reranker at all, and 0.742 once the scores separate.

Ask the model for logits and apply the sigmoid here, where it runs in float64.
Scores stay 0-1, matching the cohere, vllm and zeroentropy rerankers.

Also drop the remaining pyright references; the project type-checks with ty.
2026-08-07 14:07:53 +03:00
..
capabilities Scope the limit notice and spend the cite window on own turns only 2026-07-30 19:14:14 +03:00
cassettes Recover from worker death and deny metadata writes 2026-07-28 19:11:19 +03:00
chat Constrain the chat prompt container height 2026-07-25 14:40:40 +03:00
data Remove repliqa from test setup 2026-06-01 10:40:52 +03:00
docker Add a docker-compose for integration test services 2026-06-04 11:13:16 +03:00
ingester Classify obstore config errors as permanent 2026-08-06 12:53:45 +03:00
sandbox Point the limit comments at the right limits 2026-07-28 19:11:37 +03:00
store Load document blobs only when asked 2026-08-06 13:04:51 +03:00
tools Report tool failures with ToolFailed 2026-07-27 18:26:42 +03:00
__init__.py Use LLM-as-a-judge to test QA 2025-06-25 18:18:28 +03:00
conftest.py Consolidate duplicated client-side tests 2026-07-26 13:30:59 +03:00
json_body_serializer.py Keep cross-encoder rerank scores apart when they saturate 2026-08-07 14:07:53 +03:00
services.py Add a docker-compose for integration test services 2026-06-04 11:13:16 +03:00
test_ask_images.py Accept images on ask/analyze 2026-07-25 10:23:55 +03:00
test_cascade_delete.py Skip embedding validation on document deletion 2026-07-22 10:48:00 +03:00
test_chunk.py Fix defects found reviewing the coverage work 2026-07-27 10:44:32 +03:00
test_chunker.py Fix defects found reviewing the coverage work 2026-07-27 10:44:32 +03:00
test_cli.py Accept images on CLI ask/analyze and MCP tools 2026-07-25 10:35:25 +03:00
test_client.py Load document blobs only when asked 2026-08-06 13:04:51 +03:00
test_client_analyze.py replace haiku.skills with native Pydantic AI capabilities 2026-07-24 15:26:17 +03:00
test_config.py Cover the lazy config init and restore the uv cache in CI 2026-07-27 13:35:50 +03:00
test_context.py Fix defects found reviewing the coverage work 2026-07-27 10:44:32 +03:00
test_context_enhancement.py Shrink the docling-serve context-expansion cassette 2026-06-29 15:15:26 +03:00
test_converters.py Use a neutral dotfile in the docling name examples 2026-07-30 14:13:23 +03:00
test_database_autocreate.py Test vacuum against real table state 2026-07-27 12:11:01 +03:00
test_docling_serve_client.py Cover MCP, ingester and converter error paths 2026-07-26 19:14:26 +03:00
test_doctor.py Reach and enforce 100% coverage 2026-07-26 20:11:02 +03:00
test_document.py Load document blobs only when asked 2026-08-06 13:04:51 +03:00
test_download_models.py Reach and enforce 100% coverage 2026-07-26 20:11:02 +03:00
test_duplicate_ingestion.py Re-check URI under the write lock to prevent duplicate ingestion 2026-07-08 11:26:45 +03:00
test_embedder.py Pool HTTP clients for vLLM embedding and vLLM/Jina reranking 2026-07-09 10:32:44 +03:00
test_embedder_config.py Fix defects found reviewing the coverage work 2026-07-27 10:44:32 +03:00
test_filter.py Rename document_meta identity column document_id to id 2026-07-08 16:32:55 +03:00
test_info.py Harden history against tag annotation failures; document tag restore 2026-07-16 13:11:34 +03:00
test_inspector.py Surface database-open errors in the TUIs 2026-07-25 11:27:29 +03:00
test_lancedb_connection.py Fix defects found reviewing the coverage work 2026-07-27 10:44:32 +03:00
test_logging.py Additional tests 2026-02-13 18:08:46 +02:00
test_mcp.py Cover MCP, ingester and converter error paths 2026-07-26 19:14:26 +03:00
test_pdf_attachments.py compare off-loop work against actual event-loop thread, fix ty 2026-06-22 10:17:02 +03:00
test_pdf_concurrency.py Serialize all in-process pdfium access under a shared lock 2026-06-14 10:50:42 +03:00
test_pdf_split.py Cover MCP, ingester and converter error paths 2026-07-26 19:14:26 +03:00
test_picture_description.py cover picture-description provider and rebuild edge cases 2026-05-05 14:20:37 +03:00
test_picture_in_context.py Reach and enforce 100% coverage 2026-07-26 20:11:02 +03:00
test_processing.py Normalize pragma comments to the codebase's single-line form 2026-07-26 20:18:19 +03:00
test_reader.py async versions for convert_file() convert_text() 2025-11-26 17:46:25 +02:00
test_rebuild.py Stop loading page rasters on the title and update paths 2026-08-06 13:04:51 +03:00
test_reranker.py Keep cross-encoder rerank scores apart when they saturate 2026-08-07 14:07:53 +03:00
test_s3_integration.py Clean up S3 test data per-test to keep SeaweedFS writable 2026-06-26 16:41:20 +03:00
test_s3_source.py canonical metadata keys: source_revision + content_type, bump to 0.50.0 2026-05-26 11:43:30 +03:00
test_search.py Cover get_model provider branches and visualize_chunk edge paths 2026-07-26 19:23:54 +03:00
test_settings.py Cover store engine and repository paths 2026-07-26 19:36:37 +03:00
test_telemetry.py Route eval scripts through the shared telemetry configure 2026-07-10 11:38:33 +03:00
test_title_generation.py Reach and enforce 100% coverage 2026-07-26 20:11:02 +03:00
test_utils.py Map thinking onto Pydantic AI's unified setting 2026-07-27 18:26:42 +03:00
test_vacuum_debounce.py Serialize tag operations, vacuum, and metadata refresh with writes 2026-07-16 12:32:57 +03:00
test_versioning.py Throttle background auto-vacuum to at most once per 5 minutes 2026-06-12 10:17:18 +03:00