0.14.1

Fixes

• Gemini: fixed bug in a finish reason handler (#161) (#162)

0.14.0

Features

• support streaming with multiple candidates (#158)

Other

• add integration tests for multi-input embeddings (#159)
• pin Docker image to alpine-3.20 (#160)
• bump aidial-sdk from 0.15.0 to 0.16.0 (#156)

0.13.0

Features

• Add Google Search Grounding static tool (#147)
• Calculate Google Search Grounding price in tokens, allow configuration fo Search usage (#153)

Fixes

• Gemini: reporting safety error properly (#149)
• fixed bug in truncate prompt (#143)
• utilize reference url and markdown in attachments for Grounding (#151)

Other

• bump aidial-sdk from 0.14.0 to 0.15.0 (#154)
• bump aiohttp from 3.10.2 to 3.10.11 (#152)
• bump epam/ai-dial-ci from 1.9.3 to 1.9.4 (#145)
• bump epam/ai-dial-ci from 1.9.4 to 1.10.0 (#155)
• cleanup untagged images (#146)
• improved integration tests (#124)
• migrated from Ubuntu (ubuntu:23.10) to the Alpine image (python:3.11-alpine) (#113)

0.12.1

Fixes

• fixed bug in truncate prompt (#143) (#144)

0.12.0

Features

• Gemini: supported prompt truncation and request.max_prompt_tokens parameter (#141, #139)
• Gemini: supported system_instruction and tools in token counting (#137)
• supported content parts in all models (#133)
• added Docker healthcheck (#120)

Other

• bump epam/ai-dial-ci from 1.9.1 to 1.9.2 (#138)
• bump epam/ai-dial-ci from 1.9.2 to 1.9.3 (#142)
• bump starlette from 0.36.3 to 0.40.0; fastapi from 0.109.2 to 0.115.2 (#140)
• fixed README table formatting (#135)

0.11.0

Features

• Supported Gemini 1.5 Flash/Pro 002 (#128)

Fixes

• fixed incorrect handling of content filter errors for Gemini models (#126)
• fixed incorrect handling of multi-input embedding requests (#130)

Other

• bumped github workflow versions (#134)
• removed dead code (#132)

0.10.0

Features

• added Multimodal embeddings (#106, #114)
• added code=truncate_prompt_error to errors raised by prompt truncation algorithm (#111)
• improved error message for unsupported request.max_prompt_tokens (#122)

Fixes

• fixed incorrect handling of system message for Gemini models (#121)
• fixed potential server deadlock on start-up (#117)
• supported zero-argument tools/functions (#111)

Other

• bump aidial-sdk from 0.10.0 to 0.13.0 (#111)
• bump aiohttp from 3.9.4 to 3.10.2 (#112)
• bump cryptography from 42.0.8 to 43.0.1 (#118)

0.9.0

Features

• supported plain text file attachments in Gemini (#102)

Fixes

• fixed max number of PDF pages for Gemini 1.5 (#99)
• supported encrypted PDFs with empty password (#98)
• fixed incorrect usage of log.exception (#97)

Other

• bump aidial-sdk from 0.8.0 to 0.10.0 (#105, #108)
• bump certifi from 2023.7.22 to 2024.7.4 (#94)
• bump epam/ai-dial-ci from 1.8.1 to 1.8.3 (#93)
• bump epam/ai-dial-ci from 1.8.3 to 1.9.0 (#107)
• bump setuptools from 68.2.2 to 70.0.0 (#103)
• bump urllib3 from 1.26.18 to 1.26.19 (#91)
• bump zipp from 3.17.0 to 3.19.1 (#95)
• bumped to aidial-sdk=0.10.0 (#109)

0.8.1

Other

• bump zipp from 3.17.0 to 3.19.1 (#95)
• bump certifi from 2023.7.22 to 2024.7.4 (#94)
• bump epam/ai-dial-ci from 1.8.1 to 1.8.3 (#93)
• bump urllib3 from 1.26.18 to 1.26.19 (#91)

0.8.0

Features

• added Gemini 1.5 Flash as gemini-1.5-flash-001 (#88)
• added citation sources to attachments (#89)

Other

• bump epam/ai-dial-ci from 1.8.0 to 1.8.1 (#90)
• bump requests from 2.31.0 to 2.32.0