v1.2

What's New

• GPT-4o model support
• US Government cloud deployments now support Azure AI Search with semantic ranking which measurably improves search relevance by using language understanding to rank results.
• Secure mode configuration
  • Includes configuration for scenarios where infrastructure security and privacy are essential, like those in public sector and regulated industries. Key features of “secure mode” include:
    • Disabling public network access: Restrict external access to safeguard public access.
    • Virtual network protection: Deploy your Azure services within a secure virtual network.
    • Private endpoints: The deployed Azure services connect exclusively through private endpoints within a virtual network where available.
    • Data encryption at rest and in transit: Ensure encryption of data when stored and during transmission.
• Compatible with Microsoft Cloud for Sovereignty
  • Deployable within the Sovereign Landing Zone (SLZ) Online management group with secure mode
  • Compliant with the Azure Policy built-in Sovereignty Baseline Global Policy Initiative to help enforce data residency requirements.
  • Customers interested in operational transparency can also take advantage of Transparency logs (preview) by onboarding their tenant at no additional cost.
  • Learn about how Cloud for Sovereignty can help enforce sovereignty concerns within the product documentation and by exploring the getting started learning path.

What's Changed

• Merge pull request #724 from microsoft/vNext-Dev by @dayland in #728
• chore: Update tiktoken dependency to version 0.7.0 to support gpt-4o by @dayland in #729
• Update vite.config.ts by @asbanger in #736
• Update storage privacy settings by @KronemeyerJoshua in #737
• terraform support for manual app registration by @asbanger in #732
• Joshuakr/7569 backend test client by @KronemeyerJoshua in #730
• chore: Refactor file extension comparison in PartitionFile function by @wotey in #740
• Joshuakr/7031 vite fix by @KronemeyerJoshua in #752
• update infra delete by @wotey in #746
• Added data_points to return by @KronemeyerJoshua in #757
• Joshuakr/update ai search version by @KronemeyerJoshua in #756
• Sync latest hotfixes from main to vNext-Dev by @dayland in #754
• fix long running az cli command by @wotey in #759
• chore(deps): Bump azure-identity from 1.12.0 to 1.16.1 in /app/backend by @dependabot in #749
• Secure mode changes by @dayland in #764
• Add images to accepted file types and correctly navigate folders by @KronemeyerJoshua in #771
• Fix auth error on language detection by @georearl in #758
• Update node version by @KronemeyerJoshua in #768
• Remove few shot effect on ungrounded by @dayland in #774
• Documentation - fixing typographical errors by @asbanger in #779
• updating error to be removed on next question run by @wotey in #781
• Fix issue with ai service region header to be explicit by @dayland in #785
• Push images to ACR with v2 REST APIs by @dayland in #786
• Add secure mode cost estimation and Fix broken links to documentation by @dayland in #788
• Update considerations_production.md by @KronemeyerJoshua in #787
• Bump langchain-experimental from 0.0.52 to 0.0.61 in /app/backend by @dependabot in #761
• Sync main with vNext-Dev by @dayland in #789
• UpdateIA v1.1.2 Sandbox Estimation by @asbanger in #790
• Update langchain and openai dependencies and set agent_executor_kwargs by @dayland in #792
• Make AOAI model SKU values parameter based by @dayland in #794
• Make DDOS plan an optional setting via local.env by @dayland in #793
• Delete AOAI models as part of automated RG deletion by @georearl in #796
• update langchain to required versions by @wotey in #802
• update to fix csv data issue because of bad prompt by @wotey in #806
• Bump langchain-community from 0.2.5 to 0.2.9 in /app/backend by @dependabot in #801
• Remove vision SDK for REST APIs by @dayland in #811
• Update Language Detection and Translation to use Azure Cognitive Multi Service API by @ruandersMSFT in #798
• Addition of content filtering checks to prevent hanging error in chat functionality by @wotey in #809
• Merge latest main hot-fixes to vNext-Dev by @dayland in #815
• Wotey/7998 tabdata by @wotey in #813
• chore: Update deployment configuration link in features documentation by @georearl in #819
• Replace TF templates module by native function by @sinacek in #821
• Cmccleary/docs by @cmccleary8 in #823
• Upgrade hashicorp/null to 3.2.2 by @sinacek in #822
• Update code to use Entra ID Auth instead of shared keys by @dayland in #797
• Update Analysis panel to enable tooltip message when content is not available by @wotey in #825
• Update function managed identity to have right Search role by @dayland in #827
• fix the url encoding when generating sas tokens for citations by @dayland in #832
• update to handle error due to new content filtering method by api by @wotey in #837
• Bump nltk from 3.8.1 to 3.9 in /functions by @dependabot in #826
• Bump nltk from 3.8.1 to 3.9 in /functions by @dayland in #840
• New API to allow uploaded files to pass through backend service by @dayland in #839
• Route citation file view through backend service by @dayland in #838
• Add spinner overlay for file upload progress by @dayland in #841
• Updated to full connection string to support all regions by @dayland in #842
• change view back to document section on click of another citation by @wotey in #844
• Cleanup of AzDO pipeline docs to identify as examples by @dayland in #843
• Update Azure Search client configuration with audience for national c… by @dayland in #845
• File progress reversion by @wotey in #848
• Status indicator on citation load plus fixes for USGov debugging by @dayland in #849
• update to fix issue #7952 by @wotey in #851
• Enable semanitic reranking feature for USGov by @dayland in #853
• Documentation updates for secure mode and SLZ support by @dayland in #855
• Merging 1.2 features into Main by @lon-tierney in https://github.com/microsoft/PubSec-Info...

v1.1.1

What's New

• Support for migrating or upgrading from v1.0 to v1.1.x. More details can be found in the documentation at Moving from v1.0 to v1.1
• API Version updates
  • openai=1.17.0
  • langchain=0.1.16
  • langchain-openai=0.1.3
  • tiktoken=0.5.2
  • Azure OpenAI API='2024-02-01'
• Streaming of responses in the UX. Answers will begin streaming in chats without having to wait on full answer before returning.
• Support for adding Service Management Reference and multiple owners on Azure Entry object created for Info Assistant. Following recommended security baselines, these new variables can be used to satisfy policy requirements when deploying.

What's Changed

• Update README.md (Typo "Azure Open AI"→"Azure OpenAI") by @hyoshioka0128 in #680
• Add MAX_CSV_FILE_SIZE parameter to limit the size of CSV files in Tab… by @dayland in #687
• Aparmar/upgrade azure open ai by @ArpitaisAn0maly in #682
• Update BICEP references to Terraform in README and documentation files by @dayland in #689
• Don't reprocess soft deletes by @georearl in #686
• Update Bicep references to Terraform in deployment documentation and … by @georearl in #688
• Bump langchain-experimental from 0.0.49 to 0.0.52 in /app/backend by @dependabot in #676
• Geearl/7540 storage retention policy by @georearl in #691
• Add tags to azurerm_linux_function_app and azurerm_search_service res… by @georearl in #692
• Update Azure web app logging configuration in enrichmentapp.tf by @georearl in #697
• Update diagnostic setting name in enrichmentapp.tf by @georearl in #696
• Geearl/7547 credential lifetime hotfix 3 by @georearl in #698
• Update architectural_decisions.md (Typo "Azure Open AI"→"Azure OpenAI") by @hyoshioka0128 in #700
• Joshuakr/7432 Chat Streaming by @KronemeyerJoshua in #701
• Geearl/upgrade and migrate by @georearl in #679
• Fix line spacing issue by @KronemeyerJoshua in #709
• Add application insights connection string and key to function app by @dayland in #706
• Apply ServiceTreeEntryID via local.env by @georearl in #703
• Call is now cancelled when chat mode is switched midstream by @KronemeyerJoshua in #711
• updte azurerm_monitor_diagnostic_setting by @georearl in #712
• Geearl/7566 clean up by @georearl in #708
• cleanup script - added TF state removal by @georearl in #716
• Geearl/7564 additional owners by @georearl in #718
• chore: Update role assignments for various apps by @georearl in #719
• chore: Add check for empty object_id before adding to array by @dayland in #720
• update demo question in chat by @wotey in #722
• Fixed Generating Answer being displayed on load by @KronemeyerJoshua in #725
• Joshuakr/7653 comparison fix by @KronemeyerJoshua in #726
• v1.1.1 - Hotfixes for v1.1 by @dayland in #724

Full Changelog: v1.1...v1.1.1

v1.1

What's New

• New Chat Modes that include Work + Web, that includes search the Web (via Bing APIs) and compare and contrast answers between my Work documents and Web search results. Also included is Ungrounded which allows direct interaction with an LLM without grounding to allow for creative and fully generative responses.
• Enhanced File Management for files loaded into IA to including filtering, resubmit and delete file(s). Plus detailed views on processing status and/or error messages without needing IT to look in a database.
• SharePoint (document libraries, not lists or websites) as a source configurable to point at multiple sites and/or folders. Documents are copied into IA for processing, therefore RBAC is not maintained. The process supports all CRUD operations for files in configured SharePoint document libaries.
• Compatible with US Gov sovereign clouds including Azure OpenAI hosted in US Gov regions
• Documentation for use of Bing Safe Search and Azure OpenAI Content Filtering controls for Azure customers to tailor the content filtering behavior to their needs while aiming to prevent potentially harmful generated content and any copyright violations from public content

Preview Features

• Assistants/ Autonomous Agents are stateful evolution of traditional Generative AI application where you have to manage conversation state, tool integrations, and execute them manually. Assistants/Autonomous Agents are stateful and they automatically manage tool/ function integrations and conversation state. We are introducing two assistants with this release:
  • Math Assistant. How do you get a Large "Language" Model to reason over order of operations in Math? We have included a tool based on Assistants patterns.
  • Tabular Data Assistant. How do you reason over a set of tabular data that you do not know the schema for to answer complex questions in natural language? We show how when a user asks question in natural language and Tabular Data Assistant generates python code to understand tabular data, reason over it, and generate a response.

Coming Soon

• Secure Deployment: Updates to ensure all traffic and data is private and encrypted. We will also be compatible with Microsoft's Cloud for Sovereignty Security Baselines, Sovereign Landing Zone Baseline for Online and Corp landing zones.
• Migrate or Upgrade from v1.0 to v1.1 without the need to reload and reprocess all your files.

Issues Fixes

Fixes #295
Fixes #345
Fixes #467
Fixes #481
Fixes #487
Fixes #493
Fixes #496
Fixes #597

What's Changed

• brrohrer/modelcheckerror by @rohrerb in #400
• Fix fileupload status window bug and logger bug by @ryonsteele in #402
• Geeal/6317 embeddings retry by @georearl in #412
• Ryonsteele/6275 backend fastapi by @ryonsteele in #416
• Ryonsteele/manual merge by @ryonsteele in #406
• brrohrer/RetryErrored by @rohrerb in #407
• Enable file deletion by @nhwkuhns in #419
• Update Document Intelligence to latest GA to resolve unicode issues by @ryonsteele in #432
• Remove functions_flow.md and update related documentation by @georearl in #454
• Ryonsteele/manual merge v1.0 by @ryonsteele in #458
• Added azureRoles JSON by @paullizer in #468
• Resolve function debug issue and add logic for multiple table spans by @ryonsteele in #462
• Add support to document viewer for more document types by @ryonsteele in #473
• Geearl/6323 large tables by @georearl in #429
• Brohrer/6188 - by pass rag part 1 by @rohrerb in #448
• Citation De-duplication from @TaylorN15 by @dayland in #476
• Multilingual support by @ArpitaisAn0maly in #470
• Syncing 3 PRs related to large table updates to 1.0 HF (main) by @dayland in #478
• Remove extra tag quoting when passing to search function by @ryonsteele in #484
• Update deployment.md by @avidunixuser in #486
• Bump fastapi from 0.103.2 to 0.109.1 in /app/enrichment by @dependabot in #489
• Fast-api version increase by @ryonsteele in #490
• Fix multitag index array values to allow for proper search by @ryonsteele in #491
• Commented out code for Media Service and AVAM Service by @georearl in #502
• Remove media_service and avam modules from BICEP by @dayland in #505
• Geearl/6515 delete UI by @georearl in #507
• Geearl/7023 delete files from UI by @georearl in #510
• Add "All" option to dropdownTimespanOptions by @georearl in #511
• Dabiscup/aoaimagchanges - AOAI in Gov Changes by @rohrerb in #494
• Geearl/7021 tag filter by @georearl in #509
• Geearl/7024 retry errors by @georearl in #514
• Conversion from BICEP to Terraform - Part 1 by @dayland in #508
• SharePoint Ingestion Logic App by @KronemeyerJoshua in #513
• Remove base_url breaking param by @ryonsteele in #518
• Resolve issue with new gov logic on aoai endpoint resolution by @ryonsteele in #522
• Jeffmausolf/6187 add md4c info to prod considerations by @mausolfj in #524
• MathTutor Code by @ArpitaisAn0maly in #512
• deletion of chunks on upload by @georearl in #515
• Geearl/7045 show processing log by @georearl in #520
• Geearl/7045 show processing log by @georearl in #528
• Fix statusLog document upsert in FileUploadedFunc and add flag for pr… by @georearl in #523
• Geearl/7083 consolidate cosmos by @georearl in #529
• Conversion from BICEP to Terraform - Part 2 by @dayland in #516
• addition off Status component and new Panel for Status log by @wotey in #535
• Geearl/7162 search key error by @georearl in #536
• Hot-fix for type on app.py for Terraform conversion by @dayland in #539
• Geearl/6403 unstructured updates by @georearl in #530
• Fix Azure SDK issue and update translation API endpoint by @dayland in #540
• Ryonsteele/bing search poc by @ryonsteele in #534
• joshuakr/7160 - ARM and Sharepoint Site Iteration by @KronemeyerJoshua in #547
• change buttons to icons by @georearl in #552
• Fix condition for checking tags in get_tags_and_upload_to_cosmos func… by @georearl in #553
• BICEP to Terraform deployment change by @dayland in #537
• Merge pull request #523 and #530 from vNext-Dev for large table fixes by @dayland in #558
• Geearl/7188 tags to start 2 by @georearl in #560
• Aparmar/6512 structured data analysis csv by @ArpitaisAn0maly in #559
• Jeffmausolf/6968 update python version and tool extensions by @mausolfj in #525
• Update environment variables and pipeline templates by @dayland in #563
• Added documentation on the upload status page by @georearl in #561
• brrohrer/6188 - Bypass RAG - Part 2 by @rohrerb in #308
• Add Azure Blob Storage and URL parsing*** by @georearl in #564
• F...

1.0

What's New

Lots of new documentation

• Troubleshooting guides for common issues
• Swagger API documentation
• New click thru guides to walkthrough deployment and using the system for the first time
• Considerations for production implementation added
• Sample data links provided
• Updated to Cost Estimation Guide for v1.0
• Advanced Architecture documents now available
• Sharing architectural decisions for the Information Assistant
• Functional Test guidance
• How to use your own UX with Info Assistant APIs using client credentials flow

Performance

• Enrichment App metric based autoscaling
• Function Apps metric based autoscaling

More secure deployment

• Use of KeyVault in BICEP deployments to avoid MS Defender alerts
• KeyVault implementation for App Services

Dependency Updates

• Document Intelligence API Updated to 2023-07-31
• Updates to NPM packages addressing CVEs

Hotfixes to resolve bugs

• #443
• #405
• #397
• #427
• #355
• #352
• #285
• #344
• #343
• Allow spaces in folder name and tags
• Improvements in prompt engineering for citations, ungrounded responses, and follow up questions

Updated bug report templates

• Improved bug reports

What's Changed

• multilanguage support for response by @ArpitaisAn0maly in #346
• status message displayed in UI by @georearl in #332
• Remove input restriction add support folder and tag spaces by @ryonsteele in #342
• Ryonsteele/6239 tokenauth doc by @ryonsteele in #340
• Resolve random issue of no citation based on file moniker by @ryonsteele in #348
• fixes to queue names by @georearl in #350
• Update default bicep params to use default flavor model for deployment name by @ryonsteele in #351
• Ryonsteele/6264 util blob sas fix by @ryonsteele in #356
• Update README.md by @asbanger in #339
• known issues by @georearl in #361
• Merge 0.4-Delta HFs from main to vNext-Dev by @dayland in #359
• Encode tags on frontend and decode in backend by @ryonsteele in #363
• Azure cost estimation for delta release by @asbanger in #369
• Ryonsteele/6267 keyvault deployment by @ryonsteele in #370
• Add documentation for workbook and func test usage by @ryonsteele in #371
• Ryonsteele/6270 correct appid by @ryonsteele in #377
• Aparmar/6280 architectural decisions.md by @ArpitaisAn0maly in #379
• Add logic to check for nltk unzip issue by @ryonsteele in #384
• Remove unused cogservices account by @ryonsteele in #392
• Sync vNext-Dev with Main by @ryonsteele in #394
• Update the process flow architecture with latest product naming by @asbanger in #398
• assign SP ID for Ci/CD shared by @dayland in #403
• Ryonsteele/function autoscale by @ryonsteele in #390
• Add check for existing CUA deployment object and remove by @ryonsteele in #415
• Update costestimator.md to reflect current release by @mausolfj in #418
• prompt engineering to fix citation with no answer questions and follo… by @ArpitaisAn0maly in #421
• Resolve npm dependencie vulerabilities by @ryonsteele in #427
• Link broken-configure your own language by @asbanger in #423
• Mausolfj/6301 user experience by @mausolfj in #422
• Implement runtime keyvault secrets for app services by @ryonsteele in #424
• Document Intelligence version bump hotfix by @ryonsteele in #433
• Updates to docs based on community feedback by @dayland in #434
• Update example list with new questions by @dayland in #436
• Update configure_local_dev_environment.md by @ArpitaisAn0maly in #435
• Resolve issue with aoai key reference when not using existing deployment by @ryonsteele in #438
• Documenation on considerations for production adoption by @ryonsteele in #426
• Correct typo in sentence of new documentation by @ryonsteele in #440
• 1.0 - update for high level architecture by @asbanger in #444
• Resolve issue with chunks created statuslog by @ryonsteele in #447
• Updating documentation by @asbanger in #445
• Add Architecture Document in /docs by @dayland in #446
• Ltierney/deploy-video by @lon-tierney in #441
• Azure estimation 1.0 release by @asbanger in #449
• Update deployment troubleshooting and improve UX analysis panel docs by @dayland in #450
• Updating hard link to redirect link for YouTube by @lon-tierney in #451
• Fixed typo and broken image links by @KronemeyerJoshua in #452
• Update sample data links in user_experience.md by @dayland in #453
• Merge pull request #454 from microsoft/geearl/function-flow-doc by @dayland in #456
• Update bug_report.md template with additional instructions and details by @dayland in #455

New Contributors

• @mausolfj made their first contribution in #418
• @KronemeyerJoshua made their first contribution in #452

Full Changelog: v0.4-Delta...v1.0

0.4 Delta

What's New:

• Vector Hybrid Search which combines vector similarity with keyword matching to enhance search accuracy.
  • Added document processing pipeline steps to generate embeddings for text-based files. Bring your embedding (Azure Open AI or open-source embedding model).
  • Extended document processing pipeline with richer language detection and translation to avoid common error with OOTB Azure Cognitive Search skillsets
  • Switched to direct search index inserts instead of Azure Cognitive Search Indexer and skillsets
  • Restructured and added vector columns to Azure Cognitive Search Index (expanded JSON into index fields)
  • Update UX to embed users query and execute Vector Hybrid Search with Semantic
  • Added pipeline to process images and store enrichments as keywords in Azure Cognitive Search index which would allow user to do text to image search.
• Added iFrame document and image rendering of source material under citation panel of UX.
• Added support for several new file types using Unstructured.io
  • Text-based: pdf, docx, html, htm, csv, md, pptx, txt, json, xlsx, xml, eml, msg
  • Images: jpg, jpeg, png, gif, bmp, tif, tiff
• Added support for US Government deployments
• Added filtered query support for Azure Cognitive Search index fields
  • Enabled upload to a folder and adding tags to uploaded file in UX
  • Enabled filtering search by "folder" and/or "tags" fields in Adjust panel in UX
• Added function testing of document pre-processing pipelines and embeddings REST APIs
• Added branding updates that allow a warning banner and UX title updates
• Enhanced infrastructure and application logging
  • Detailed chunk-based logging for embeddings and indexing
  • New Azure Workbook to help investigate infrastructure level errors (i.e. App Service not starting up correctly)

What's Changed

• Merge pull request #204 from microsoft/vNext-Dev by @dayland in #208
• hot fix for missing subtitle in non-pdf by @georearl in #210
• Forward integrate HF for non-PDF missing subtitle attribute by @dayland in #214
• Update README.md for model version support by @ArpitaisAn0maly in #216
• Geearl/5974 acr bicep by @georearl in #199
• Gov init by @rohrerb in #212
• Model check Fix by @rohrerb in #223
• Bash variable if check repair by @rohrerb in #224
• Sovereign Setup Instructions by @rohrerb in #222
• Geearl/5977 container app service by @georearl in #205
• Temperature setting by @asbanger in #215
• udpated default model to gpt-35-turbo-16k by @dayland in #226
• Geearl/5982 appservice debug 2 by @georearl in #217
• Fix for Deployment Name by @rohrerb in #229
• Parameter for environment so we can share the vNext.yml by @rohrerb in #228
• Gov Automation Fix. by @rohrerb in #232
• Some more az login fixes for automation. by @rohrerb in #233
• Temp Build Fix for Gov by @rohrerb in #234
• Geearl/5972 generate embeddings by @georearl in #230
• 6012 Add utils method for extracting html charset else default to utf8 by @ryonsteele in #239
• brrohrer/6010 enrichment func gov by @rohrerb in #237
• AAD Regression Fix by @rohrerb in #240
• Lwilk/5892 image func by @lmwilki in #227
• Geearl/6053 large files missing chunks by @georearl in #242
• 5993 Change how page numbers are collected and stored for chunks by @ryonsteele in #243
• requeue logic with max retries by @georearl in #241
• Geearl/6055 missing index key values by @georearl in #244
• Fix passing of api variable to function by @dayland in #245
• brrohrer/6050-GovIngestPipeline by @rohrerb in #247
• Move installation into startup entrypoint script to reduce build by @ryonsteele in #246
• Ryonsteele/6063 default responselength change by @ryonsteele in #249
• Add build-containers to pipelines by @dayland in #250
• updates to process flow diagram for embeddings by @dayland in #248
• Aparmar/5661 hybrid search by @ArpitaisAn0maly in #255
• 6105/brrohrer stack trace logging by @rohrerb in #253
• Geearl/6070 non pdf by @georearl in #256
• Ryonsteele/int test by @ryonsteele in #254
• Geearl/6112 missing chunk index by @georearl in #257
• Fix embeddings mismatch between pipeline and REST APi by @dayland in #258
• Improvements to functional tests by @dayland in #260
• Branding and UX updates by @dayland in #261
• Merging 0.3-Gamma hotfixes into vNext-Dev by @dayland in #264
• Fix to have non-pdf files honor the DEV_CODE flags by @dayland in #265
• Add image enrichment to index for image function by @ryonsteele in #259
• Add more test case files and types by @ryonsteele in #266
• Extended logging to UX for uploaded files by @dayland in #267
• Enable citations to use new vector index fields by @dayland in #268
• Ryonsteele/6108 cicd functional by @ryonsteele in #272
• Aparmar/6151 enable hybrid semantic search by @ArpitaisAn0maly in #275
• Updates for Sprint 12 arch changes by @dayland in #273
• Ryonsteele/6115 api tests by @ryonsteele in #277
• Update to use model name over deployment name by @dayland in #279
• Fix to put calculated chunk name in index for images by @dayland in #280
• Create seperate pipeline ymls for gov and com by @ryonsteele in #283
• Updating the GPT-35-Turbo-16K capacity to default by @asbanger in #284
• Fix for AOAI BICEP deployment by @dayland in #287
• Fixing the GitHub broken Links by @asbanger in #281
• Ryonsteele/indexing statuslog by @ryonsteele in #289
• Change to create seperate func ASP and resolve enrichment concurrency by @ryonsteele in #293
• Geearl/6184 function timeout by @georearl in #294
• ReadMe update for Image Search for Delta release. by @ArpitaisAn0maly in #296
• Prompt to delete index on deploy if vector size changes by @ryonsteele in #297
• Vector Search Doc Update to ReadMe by @ArpitaisAn0maly in #300
• Update preprocessing feature document for unstructured and image by @ryonsteele in #298
• Remove stack from upload statuslog by @ryonsteele in #301
• Geearl/6130 azure caps by @georearl in #303
• Fix env variable mispelling by @ryonsteele in #304
• Dabiscup/applicationtitle2 by @danimal521 in #302
• Resolve startup failure with aoai model usage by @ryonsteele in #307
• brrohrer/6199 - Unstructured J...

0.3 Gamma

What's New

• Support for GPT-4 using Chat/Completion APIs (backwards compatible with GPT-3.5-Turbo)
• New UX "Info" panel to show common configuration values (Azure Open AI, Azure Search, and Language settings)
• Improved the UX for citation chunks to show as HTML rather than raw JSON
• Ability to resubmit chat questions using the "Regenerate" button. Regenerating an answer will use the past question with the current "Adjust" settings.
• Removal of the "Ask a Question" section in the UX
• CUA (Customer Usage Attribution) enablement
• Adoption Workshop self-paced learning available
• Debugging support for the app/frontend Typescript code

Known Issues

See our updated Known Issues list. Please check the Issues board for any issues you encounter.

What's Changed

• Azure cost estimation - Creation of costestimator.MD by @asbanger in #136
• Sync Cost Calculator HF from main to vNext-Dev by @dayland in #137
• Fix - updating Azure OpenAI retired model by @asbanger in #138
• Hot fix for Response length. by @ArpitaisAn0maly in #139
• Set EasyAuth settings for MSIT compliance by @dayland in #142
• Geearl/5872 fr internal error by @georearl in #140
• update branch name in dev pipeline by @dayland in #143
• 5858- Conversion from Completion API end point to Chat Completion API… by @ArpitaisAn0maly in #146
• Geearl/5884 pdf enrichment queue by @georearl in #147
• Geearl/5885 parsing other to enrichment by @georearl in #148
• Documentation - Updated Readme.MD by @asbanger in #152
• Geearl/5896 vi infra by @georearl in #145
• Update retired models by @asbanger in #151
• media upload by @georearl in #144
• Geearl/5891 translation function by @georearl in #149
• Geearl/5913 enrichment flag by @georearl in #155
• update function app configuration by @asbanger in #164
• fix for function deployment issue by @asbanger in #165
• Telemetry Tracking Using Customer Usage Attribution by @asbanger in #159
• Removed unused "Adjust" options from UX by @dayland in #166
• Lon tierney patch 2 by @lon-tierney in #169
• Update known issues : Azure OpenAI service region availability. by @asbanger in #167
• Geearl/5903 model to UI by @georearl in #170
• Updates to prompts to remove duplication and simplify directions to LLM by @dayland in #173
• Julian/5623 allow a user to resubmit the last question by @georearl in #172
• Ltierney adoption workshop docs by @lon-tierney in #175
• Updates to improve citation responses on gpt-35-turbo by @dayland in #177
• Ltierney/DevelopmentDocsUpdate by @lon-tierney in #180
• webapp debug and regenerate a question, not just last question by @georearl in #178
• Ltierney/TransparencyUpdates by @lon-tierney in #179
• Update cost estimation for 0.3 gama by @asbanger in #183
• Updates for model specific prompting to better show citations by @dayland in #184
• Ltierney/depolyment docs updates2 by @lon-tierney in #185
• Add Navigating the Source Code section by @dayland in #186
• document html pdf docx function by @asbanger in #182
• Convert chunk JSON to HTML display by @dayland in #187
• GPT3.5 token limit error fix. Conditional max token count based on GP… by @ArpitaisAn0maly in #188
• Update default sort to Last Modified DESC by @dayland in #191
• fix to process sections and titles by @georearl in #192
• Ltierney/CUA-telemetry-updates by @lon-tierney in #189
• fix for FR nuance by @georearl in #193
• Geeartl/5973 title issue by @georearl in #194
• Fix remaining Gamma testing bugs by @dayland in #195
• Updates to the Introduction section of README.md by @dayland in #198
• Fix token size default and model variant support by @dayland in #203
• 0.3-Gamma Release Candidate by @dayland in #204

Full Changelog: v0.2-Beta...v0.3-Gamma

0.2 Beta

What's New

• Improved prompt engineering focused on reducing hallucinations and ensuring citation generation.
  • Using Chain-of-Thought prompting to help reduce hallucinations
• Language specific deployment options. You can now configure the target language of the search index, search skillsets, and prompt engineering.
• New Content Management view in the web site. This now provides the ability to view the status up uploaded files.
• Improved upload processing of PDF files. We have been able to increase the "per load" limit of PDFs to ~200 documents or ~4500 pages that can be processed at once.

Known Issues

• Uploaded PDF files in large batches may get stuck in "Queued" status. If some of your PDF files are stuck in the "Queued" status for more than 30 minutes, simply upload the files again to restart the processing.

What's Changed

• Geearl/5760 file form rec submission pdf by @georearl in #79
• Build pipeline for Red/Blue deployment. by @asbanger in #72
• Geearl/5762 file form rec polling pdf by @georearl in #83
• Geearl/5797 parser error 2 by @georearl in #87
• Geearl/5793 document map by @georearl in #89
• fixed blank folder by @georearl in #91
• Geearl/5803 pdf chunks by @georearl in #90
• Add NONE indicator to source list when none are available by @dayland in #92
• fix bug where check for first large para was blocking smaller chunk ouputs by @dayland in #93
• add import of nltk and punkt by @dayland in #95
• fix large paragraph chunking logic by @dayland in #97
• Geearl/5763 non pdf document map by @georearl in #96
• extend the pipeline to deploy azure functions by @asbanger in #82
• add "allowSkillsetToReadFileData" property to search indexer by @dayland in #101
• Update AnalysisPanel to render DOCX in Office viewer by @dayland in #100
• Geearl/5814 status complete by @georearl in #99
• Doccumentation update-developing in a codespace using vscode by @asbanger in #102
• various changes by @georearl in #103
• Hallucination Resistance prompt with Chain of Thoughts by @ArpitaisAn0maly in #107
• adding new pipeline for vNext dev branch by @dayland in #108
• Dayland/5753 add support for webapp ad security in automation by @dayland in #104
• fixes to auto deployment for limited permissions by @dayland in #109
• Adding missed parameter to CI/CD pipeline env file by @dayland in #110
• support for current file debug by @georearl in #112
• Adding Content page with File Upload and File Status as sub-pages by @dayland in #114
• Expand ASP to B3 with 3 nodes for function performance by @dayland in #116
• Fix sorting on file list and add loading... component by @dayland in #117
• Fix syntax error on new AOAI deployment by @dayland in #118
• update build to make "shared_code" dir before copy by @dayland in #119
• Adding configurable language support by @dayland in #122
• fix for docker daemon socket permission. by @asbanger in #121
• Geearl/5857 code error by @georearl in #126
• aparmar/5751-citation-bug. Code changes to enforce citation lookup di… by @ArpitaisAn0maly in #127
• Changed to GA FR API Vers and fixed error log by @lmwilki in #128
• MD documentation - update prerequisites by @asbanger in #129
• MD documentation - update known issues by @asbanger in #130
• MD documentation for configuration of local dev environment by @asbanger in #132
• changed to to dictionary to map value of response length and passed r… by @ArpitaisAn0maly in #131
• 0.2 Beta Release Candidate by @dayland in #133

Full Changelog: v0.1-Alpha...v0.2-Beta

v0.1-Alpha

This is the first Alpha release of the Information Assistant Accelerator. We believe we have achieved enough basic functionality for the accelerator to be deployed and provide these basic features:

• Chat and Q&A interfaces
• File Upload and automated chunking and indexing for PDF, HTML, and DOCX
• Explores various options to help users evaluate the trustworthiness of responses with citations, tracking of source content, etc.
• Shows possible approaches for data preparation, prompt construction, and orchestration of interaction between model (ChatGPT) and retriever (Cognitive Search)
• Settings directly in the UX to tweak the behavior and experiment with options