v0.1.0-beta

Introduction

Welcome to this is the first official release of the humanitarian AI assistant!

This version contains 14 demo recipes and supports the creation and review of recipes and memories by LLM-powered recipe manager, as well as serving these recipes to end users upon request via a chat interface. See the main README for all the goodies!

Features - High level

• Ask questions about HDX data: End users can ask questions about a connected data source and receive pre-approved, or LLM-generated responses as text- or images.
• Transparent model output: Every answer contains a flag indicating whether it has been approved by a human fact-checker or not
• Transparent data source: Human-approved answers contain a link to the source data that was used to create it
• Recipe management: Technical users/ admins can generate new recipes or edit existing ones, using their preferred IDE.

Components

• Data - Database: Holds the source data used to generate answers and recipes
• Recipe - Database: Holds recipes and memories - i.e. the artifacts that are presented to the end user during their chat session
• Chat Interface: A Chainlit user frontend for end users to interact with recipes and memories
• Recipes Server: The “connective tissue” between front- and backend, querying the recipe database for suitable answers to a user request
• Recipe Manager: A command line interface for technical users to create and edit recipes
• Data Ingestion: A data pipeline to save data from a specified data source such as HAPI into the "data database"

The above components are containerized into separate docker containers within the same docker network to allow for platform-independent deployment

Contributors

• @dividor
• @JanPeterDatakind
• @rachelauryn

Full Changelog: https://github.com/datakind/data-recipes-ai/commits/v0.1.0-beta

Additional Information

For more details, please check out the ReadMe, CONTRIBUTION.md, or contact the developer team.

What's Changed

• Feat/integration scaffolding by @dividor in #1
• Feat/integration scaffolding by @dividor in #3
• Feat/recipe server by @dividor in #2
• Feat/recipe server by @dividor in #4
• Assistants creation scripts by @dividor in #7
• Renaming UI by @dividor in #8
• Putting plugin assistants under source control by @dividor in #9
• Feature/assistants scripts by @dividor in #11
• Added check out script by @JanPeterDatakind in #10
• adjusted directory + added checksum to SQL scripts by @JanPeterDatakind in #13
• Documentation improvements for first ship by @dividor in #12
• Maint/documentation updates by @dividor in #14
• Tweaks to README by @dividor in #16
• Minor fixes by @dividor in #17
• Added secrets check GH action by @dividor in #19
• Maint/open source standards by @dividor in #20
• Adding autogen studio and agentic workflow for creating new recipes by @dividor in #18
• Draft vesion of autogen team for creating recipes, with autogen studio in build. Will refine over next month by @dividor in #21
• Secrets scan in progress by @dividor in #22
• Feature/autogen coding team by @dividor in #23
• Moved post processing functions out of core code and added flag in HAPI for latest ref period by HDX ID by @dividor in #24
• Renaming GH actions for clarity by @dividor in #25
• Renaming by @dividor in #26
• Bringing feature branch up to date by @JanPeterDatakind in #31
• Adding Recipe Management by @JanPeterDatakind in #32
• Update README.md by @rachelauryn in #29
• Update README.md by @rachelauryn in #28
• Recipe for each HAPI endpoint + refactor of DB schema by @dividor in #34
• Refactored recipes search into one module, we had two instances by @dividor in #35
• Refactored recipes search into one module, we had two instances by @dividor in #36
• Feat/create recipe for every hapi endpoint by @dividor in #37
• Feat/create recipe for every hapi endpoint by @dividor in #38
• Feat/create recipe for every hapi endpoint by @dividor in #39
• Cosmetic fixes for demo video by @dividor in #40
• PR for enabling dynamic execution of recipes by @dividor in #42
• Feat/recipe runner by @dividor in #44
• Feat/recipe runner by @dividor in #46
• Rapid prototype of chaiinlit + Fast API , lighter weight than robocorp actions and Librechat. by @dividor in #47
• Maint/upgrade chainlit 1.1.301 by @dividor in #48
• Feat/ingestion free standing by @dividor in #49
• Feat/ingestion free standing by @dividor in #50
• Added parquet support by @dividor in #51
• Better error handling for unset env variables, eg the HAPI key by @dividor in #52
• Added parquet support by @dividor in #53
• Feat/intent prompt engineering by @dividor in #54
• Feat/intent prompt engineering by @dividor in #55
• Feat/audio input by @dividor in #56
• Feat/audio input by @dividor in #57
• Minor bug fixes by @dividor in #58
• Simplified install instructions a bit by @dividor in #59
• Feat/pf tests v2 by @dividor in #60
• Feat/pf - End-to-End Github actions tests by @dividor in #61
• Adding self test for the get_memory function by @JanPeterDatakind in #64
• Chore/create assistant - enhancements to create assistant for latest architecture and features by @dividor in #63
• Documentation updates by @dividor in #66
• Feat/selenium e2e tests by @dividor in #65
• Chore/doc updates by @dividor in #67
• Chore/doc updates by @dividor in #68
• Fix to column width by @dividor in #69
• Feat/selenium e2e 3 by @dividor in #71
• Migration of tests to use Azure instead of OpenAI by @dividor in #72
• Chore/doc updates2 by @dividor in #73
• First v0.1.0-beta release by @JanPeterDatakind in #74
• Release/v0.1.0 beta by @JanPeterDatakind in #76