0.3.3

Other

• adjusted format of annotations (#109)
• bump aiohttp from 3.14.3 to 3.14.4 (#114)
• bump certifi from 2025.8.3 to 2026.10.19 (#116)
• bump fsspec from 2025.7.0 to 2026.6.0 (#111)
• bump langgraph-sdk from 0.4.2 to 0.4.4 (#110)
• bump pillow from 12.3.0 to 13.0.0 (#115)

0.3.2

Features

• improve compatibility with chat 1.* (#108)

Other

• bump libpcre2-8-0 in docker to 10.46-1~deb13u3 (#107)
• bump pypdf from 6.16.2 to 6.19.0 (#106)
• bump sentence-transformers from 5.3.0 to 5.6.0 (#105)
• bump urllib3 from 2.7.0 to 2.8.0 (#104)

Deployment Changes

Added the following environment variables:

• ENABLE_ANNOTATIONS: feature flag to enable use of annotation instead of attachments to format citations; enabled by default so that this is default behavior now. If you prefer to use legacy Chat UI 0.* on your setup this should be set to false.
• ENABLE_QUOTE_THUMBNAILS: include thumbnails for image chunks in citation attachments; disabled by default since their rendering is not supported by the new chat. If you prefer to use legacy Chat UI 0.* this can be set to true.

0.3.1

Features

• added list_documents tool as a replacement for list_documents_unordered (#102)

Other

• bump pyjwt from 2.13.0 to 2.14.0 (#103)
• (deps) bump the ai-dial-ci group with 4 updates (#99)

Deployment Changes

This version does not require any deployment changes for existing applications.

0.3.0

Features

• allow overriding the current date in the generation prompt (#80)
• chunks now have their own metadata set by processing pipeline (#85)
• improve answering over charts and tables (#78)
• index page images by what they are about, not by their values (#97)
• make citations in rag_search mcp tool response self explanatory (#91)
• make the generation prompt ask for complete answers (#86)
• report chunk and character statistics of a channel (#90)
• serve document metadata as an MCP resource (#89)
• support building multiple indexes on page image descriptions (#87)
• support concatenation of text elements extracted from chunks metadata before indexing (#94)
• support sort and filtering options in GET /channel/documents (#83)

Fixes

• Jobs worker stuck in restart loop after DB connection loss (#101)
• do not set the top-level usage field (#98)
• improve stability of background jobs execution (#82)
• make unstructured don't skip text extraction for a chart-dense PDFs (#93)

Other

• bump anyio from 4.10.0 to 4.14.2 (#96)
• bump soupsieve from 2.8.4 to 2.9 (#95)
• docker image security updates (#92)
• refactored mcp module to improve code readability (#88)
• (deps) bump the ai-dial-ci group with 4 updates (#84)

Deployment Changes

This version does not require any deployment changes for existing applications.

0.3.0-rc.2

Features

• index page images by what they are about, not by their values (#97)
• support concatenation of text elements extracted from chunks metadata before indexing (#94)

Fixes

• Jobs worker stuck in restart loop after DB connection loss (#101)
• do not set the top-level usage field (#98)

Other

• bump anyio from 4.10.0 to 4.14.2 (#96)
• bump soupsieve from 2.8.4 to 2.9 (#95)

0.3.0-rc.1

Features

• make citations in rag_search mcp tool response self explanatory (#91)
• report chunk and character statistics of a channel (#90)
• serve document metadata as an MCP resource (#89)

Fixes

• make unstructured don't skip text extraction for a chart-dense PDFs (#93)

Other

• docker image security updates (#92)
• refactored mcp module to improve code readability (#88)

0.3.0-rc.0

Features

• allow overriding the current date in the generation prompt (#80)
• chunks now have their own metadata set by processing pipeline (#85)
• improve answering over charts and tables (#78)
• make the generation prompt ask for complete answers (#86)
• support building multiple indexes on page image descriptions (#87)
• support sort and filtering options in GET /channel/documents (#83)

Fixes

• improve stability of background jobs execution (#82)

CI

• (deps) bump the ai-dial-ci group with 4 updates (#84)

Deployment Changes

This version does not require any deployment changes for existing applications.

0.2.0

Features

• accept text documents in addition to PDF (#63)
• added get_pages tool as a replacement for get_page (#59)
• channel export api (#62)
• channel import api (#71)
• document indexes and document search endpoint (#67)
• document update endpoint (#54)
• endpoint to check document existence by name (#57)
• endpoint to set document's metadata (#41)
• explicit option to allow document overwrite on upload and import (#55)
• get_citation_url mcp tool (#73)
• limited support for indexing in background (#44)
• store descriptions produced by page_description indexer in the index (#79)

Fixes

• bugfix generation prompt (#48)
• correct off-by-one in get_pages page limit check (#61)
• fixed incorrect order of chunks returned by retrievers (#65)
• fixed numba errors when running on read-only filesystem (#33)
• max_image_size of page_description indexer has no effect (#34)
• order chunk queries by chunk_id (#64)
• search on vector indexes can produce result with duplicates (#30)
• state cited page explicitly in citation attachment title (#75)
• use structured output in page_description indexer (#70)

CI

• (deps) bump the ai-dial-ci group across 1 directory with 4 updates (#15)
• (deps) bump the ai-dial-ci group with 4 updates (#68)
• (deps) bump the ai-dial-ci group with 4 updates (#77)
• re-use venv (#31)

Other

• added gpt-5* models to local config
• bump pillow from 12.2.0 to 12.3.0 (#22)
• bump pyasn1 from 0.6.3 to 0.6.4 (#27)
• bump pypdf from 6.14.2 to 6.15.0 (#53)
• bump pypdf from 6.15.0 to 6.16.2 (#76)
• bump setuptools from 82.0.1 to 83.0.0 (#28)
• bump sqlparse from 0.5.5 to 0.6.0 (#60)
• migrated to uv in Dockerfile (to address security issues in pip) (#56)
• refactoring of retrievers and answering stages reporting (#58)
• update ai-dial-sdk and unstructured to latest versions (#81)

Deployment Changes

This version does not require any deployment changes for existing application, but depending on you needs it's recommended to set the following environment variables:

• DIAL_API_KEY: should be set in order to enable background job processing
• ENABLE_DEBUG_STAGES: it's recommended to set it on dev/test environments in order to enable reporting of stages with debug information in the chat (starting from this version they are disabled by default)

0.2.0-rc.0

Features

• accept text documents in addition to PDF (#63)
• added get_pages tool as a replacement for get_page (#59)
• channel export api (#62)
• channel import api (#71)
• document indexes and document search endpoint (#67)
• document update endpoint (#54)
• endpoint to check document existence by name (#57)
• endpoint to set document's metadata (#41)
• explicit option to allow document overwrite on upload and import (#55)
• get_citation_url mcp tool (#73)
• limited support for indexing in background (#44)
• store descriptions produced by page_description indexer in the index (#79)

Fixes

• bugfix generation prompt (#48)
• correct off-by-one in get_pages page limit check (#61)
• fixed incorrect order of chunks returned by retrievers (#65)
• fixed numba errors when running on read-only filesystem (#33)
• max_image_size of page_description indexer has no effect (#34)
• order chunk queries by chunk_id (#64)
• search on vector indexes can produce result with duplicates (#30)
• state cited page explicitly in citation attachment title (#75)
• use structured output in page_description indexer (#70)

CI

• (deps) bump the ai-dial-ci group across 1 directory with 4 updates (#15)
• (deps) bump the ai-dial-ci group with 4 updates (#68)
• (deps) bump the ai-dial-ci group with 4 updates (#77)
• re-use venv (#31)

Other

• added gpt-5* models to local config
• bump pillow from 12.2.0 to 12.3.0 (#22)
• bump pyasn1 from 0.6.3 to 0.6.4 (#27)
• bump pypdf from 6.14.2 to 6.15.0 (#53)
• bump pypdf from 6.15.0 to 6.16.2 (#76)
• bump setuptools from 82.0.1 to 83.0.0 (#28)
• bump sqlparse from 0.5.5 to 0.6.0 (#60)
• migrated to uv in Dockerfile (to address security issues in pip) (#56)
• refactoring of retrievers and answering stages reporting (#58)
• update ai-dial-sdk and unstructured to latest versions (#81)

0.1.4

Features

• page description indexer (#21)