v5.29.1

What's Changed

• Fixing gen_answer failover leaving raw_answer blank by @jamesbraza in #1077

Note @jamesbraza did a clean rebase of a few commits to beyond this release, so this CHANGELOG entry was manually constructed.

Full Changelog: v5.29.0...v5.29.1

v5.29.0

What's Changed

• Preventing Greek name from crashing by @jamesbraza in #1048
• Better invalid name logs by @jamesbraza in #1049
• Fix paperqa/configs link in README.md by @chrisranderson in #1051
• Fixing all broken links, pymarkdown by @jamesbraza in #1052
• chore(deps): lock file maintenance by @renovate[bot] in #1053
• chore(deps): chore(deps): update actions/checkout action to v5 @renovate[bot] in #1058
• chore(deps): chore(deps): update actions/download-artifact action to v5 by @renovate[bot] in #1059
• Moved mypy to use local hook to unsilence it on missing dependencies by @jamesbraza in #1055
• Removed dead patch from test_add_clinical_trials_to_docs by @jamesbraza in #1056
• Restored UnpaywallProvider by updating expected response by @jamesbraza in #1057
• Refreshed test_crossref_journalquality_fields_filtering cassette by @jamesbraza in #1060
• Updating journal_quality.csv from script by @jamesbraza in #1061
• Better lookup failure message in Settings.from_name by @jamesbraza in #1064
• Lower bibtex logging to debug by @mskarlin in #1067
• Fixed SearchIndex default index_directory being a default_factory by @jamesbraza in #1070
• skip bibtex call to crossref by @mskarlin in #1072
• Adding EmptyDocsError failover to gen_answer by @jamesbraza in #1073
• Removed a duplicate computation of body's filehash by @jamesbraza in #1074

Note @jamesbraza did a clean rebase of a few commits to beyond this release, so this CHANGELOG entry was manually constructed.

New Contributors

• @chrisranderson made their first contribution in #1051

Full Changelog: v5.28.0...v5.29.0

v5.28.0

What's Changed

• Added PyPDF package to mypy_path by @jamesbraza in #1036
• Removed trailing whitespace from summary_json_system_prompt by @jamesbraza in #1037
• Allow users to skip citation stripping from evidence by @mskarlin in #1035
• Re-raising original exception in clinical_trials tools by @jamesbraza in #1038
• Fixed test_agent_types after Gemini 1.5 deprecation by @jamesbraza in #1043
• Renamed a few more answer to session by @jamesbraza in #1042
• Move context str generation into configurable function by @mskarlin in #1039
• chore(deps): lock file maintenance by @renovate[bot] in #1045

Full Changelog: v5.27.0...v5.28.0

v5.27.0

What's Changed

• Expose answer evidence score cutoff as config by @mskarlin in #1031
• Support grouping contexts by question in the final context by @mskarlin in #1032
• Journal quality fixes by @mskarlin in #1034
• Avoiding citation prompt hallucinating doi.org URLs by @jamesbraza in #1033

Full Changelog: v5.26.0...v5.27.0

v5.26.0

What's Changed

• Fixed aiohttp.ClientOSError crashing us by @jamesbraza in #1016
• S2 also retrying 403's by @jamesbraza in #1017
• Fixed _s2_get_with_retrying multiple values issue by @jamesbraza in #1018
• chore(deps): lock file maintenance by @renovate[bot] in #1020
• Removed JSONType for Pydantic's equivalent JsonValue by @jamesbraza in #1022
• Removed extra test test_can_read_normal_pdf_reader by @jamesbraza in #1019
• uv workspace layout, broke out PyMuPDF by @jamesbraza in #1021
• Created PyPDF reader package by @jamesbraza in #1023
• Fixed Docs.aadd giving bad docname after upgrade by @jamesbraza in #1025
• Removed dead code repopulate_doc_id_from_doi from DocDetails by @jamesbraza in #1024
• Made DeprecationWarning on moved Settings more precise by @jamesbraza in #1027
• Open sourced test of encode_id by @jamesbraza in #1030

Full Changelog: v5.25.0...v5.26.0

v5.25.0

What's Changed

• Sorting Doc.fields_to_overwrite_from_metadata serialization by @jamesbraza in #1010
• Ensuring manifest deserialization works for non-primitive fields by @jamesbraza in #1011
• Removing leading/trailing newlines from Context creation prompt by @jamesbraza in #1013
• Adding variable timeouts for clients by @mskarlin in #1015

Full Changelog: v5.24.0...v5.25.0

v5.24.0

What's Changed

• Exposing environment on-reset hook for Docs preparation by @jamesbraza in #1003
• Removed reset_log_levels fixture by @jamesbraza in #1004
• Lowering OpenAI base client log verbosity during CI by @jamesbraza in #1007
• Added convenience method DocDetails.make_filename by @jamesbraza in #1006
• Fixed Context not always requiring a call-arg type ignore by @jamesbraza in #1009
• Fixed DocDetails.pages having leading/trailing whitespace by @jamesbraza in #1008

Full Changelog: v5.23.0...v5.24.0

v5.23.0

What's Changed

• Rolling out markdownlint-cli to lint Markdown by @jamesbraza in #979
• Modernized CI by @jamesbraza in #982
• Allow extras to be kept in user output by @mskarlin in #983
• Assertions that chunked Text.name have correct page ranges by @jamesbraza in #985
• Refactor to expose PDF parsing settings by @jamesbraza in #984
• Removed dead Docs.index_path and generalized aadd types by @jamesbraza in #987
• Fixing missing bibtex_source, and created BibTeXSource enum by @jamesbraza in #989
• Implemented Text.__eq__ by @jamesbraza in #988
• Change contexts to use ids rather than text by @mskarlin in #990
• Fixing crash on possessive query by @jamesbraza in #992
• Moved debug config to use Claude 3 Haiku by @jamesbraza in #991
• Centralized codespell config and refreshed tooling/lock file versions by @jamesbraza in #996
• Documenting why we drop embeddings by @jamesbraza in #995
• Began deprecation cycle of exclude_text_filter by @jamesbraza in #994
• Supporting bytes in stored documents by @jamesbraza in #997
• Specifying default= in AnswerSettings.answer_length by @jamesbraza in #1000
• Denying empty text going to citation_prompt by @jamesbraza in #998
• Supporting more kwargs in ParsingSettings.parse_pdf by @jamesbraza in #999
• chore(deps): lock file maintenance by @renovate in #1001
• Removed extraneous mention of 'relevance scores' from QA prompt template by @jamesbraza in #1002

Full Changelog: v5.22.0...v5.23.0

v5.22.0

What's Changed

• Exposing IndexSettings.files_filter for programmatic use by @jamesbraza in #960
• Fixing flaky test_ensure_robust_to_timeouts by lowering timeout by 10X by @jamesbraza in #963
• Fixing code chunking working with binary files by @jamesbraza in #961
• Restoring typeguard functionality by @jamesbraza in #962
• Removing dead print_callback and empty_callback by @jamesbraza in #964
• Fixed chunk_code_text's UnboundLocalError crash on empty code file by @jamesbraza in #965
• Added missing PyPI classifiers by @jamesbraza in #967
• chore(deps): update astral-sh/setup-uv action to v6 by @renovate in #968
• chore(deps): update all non-major dependencies by @renovate in #969
• chore(deps): lock file maintenance by @renovate in #970
• add client response errors to our catching stack by @mskarlin in #972
• Making use of new Environment.get_id by @jamesbraza in #973
• Fix pdf parsing bug by @markenki in #938
• Fixed README typo from answer.k to answer.evidence_k by @jamesbraza in #974
• Fixed flaky test test_chain_completion via resolving Cannot add callback by @jamesbraza in #975
• Allow for non-mutated user specified doc-ids by @mskarlin in #977
• Remove threaded pdf parsing by @mskarlin in #978
• chore(deps): lock file maintenance by @renovate in #980

New Contributors

• @markenki made their first contribution in #938

Full Changelog: v5.21.0...v5.22.0

v5.21.0

What's Changed

• Pulling in Pydantic 2.11, latest tooling, doctests by @jamesbraza in #920
• chore(deps): lock file maintenance by @renovate in #914
• Removed LFRQA prompts by @jamesbraza in #923
• Integrating typos tool by @jamesbraza in #922
• Resolved a few deprecation warnings by @jamesbraza in #926
• Fixed README typo on basename by @jamesbraza in #927
• Began deprecation of non-async methods in Docs by @jamesbraza in #925
• chore(deps): lock file maintenance by @renovate in #930
• Gave instructions to correct citation on answer by @maykcaldas in #929
• Pulled in latest fhaviary for tool schemae changes by @jamesbraza in #933
• chore(deps): lock file maintenance by @renovate in #934
• Updated repo badges by @maykcaldas in #935
• Removed Gemini special case in Settings.adjust_tools_for_agent_llm by @jamesbraza in #936
• Expanded test_chain_completion to check IDs and text by @jamesbraza in #944
• Pulled in latest litellm and fhlmi by @jamesbraza in #945
• chore(deps): lock file maintenance by @renovate in #943
• Moved Renovate config validation to CI by @jamesbraza in #931
• Supporting Python 3.13 by @jamesbraza in #946
• Add ipykernel as dev dependency by @tobiasdiez in #947
• Usable "getting started" docs in README, both interpreter and Jupyter by @jamesbraza in #953
• Add error handling for index file loading in SearchIndex by @tobiasdiez in #948

New Contributors

• @tobiasdiez made their first contribution in #947

Full Changelog: v5.20.0...v5.21.0