1.5.0-beta

Merge pull request #83 from cloudera/mob/main

Document Upload Error Handling and Support Jpeg and Png

1.5-beta

Markdown support + UI for non-knowledge-base-answers (#82)

* markdown file handler, small refactorings, make the file delete immediately submit a request to the reconciler to delete the file

* "delete an empty data source with an empty global storage"

* "maybe ask to delete from docstore"

* fix deleting from doc store

* "markdown rendering"

* wip on markdown tables

* "now we're thinking with styled-components"

* "css styling of the markdown tables"

* update lock file

* use markdown for docling pdf, minor ui changes to source card

* readme updates

* wip on icon for no source nodes

* "now we're thinking with colors"

* "styling/coloration of the not-a-kb-response response"

* "now we're thinking with inference models"

* Set inference model when chatting with model directly

* it ain't round

* fill a little more of the "circle"

* it's still not a circle, but it's a little bigger

* mock the model api for the test

---------

Co-authored-by: Elijah Williams <ewilliams@cloudera.com>
Co-authored-by: Michael Liu <mliu@cloudera.com>

1.4.2

Mob/main (#79)

* "fix the batch embedding response processing for CAII"

* update fe to data source ids

* "now we're thinking with panic"

* fix bug with s3 path when the prefix is not provided

* "add in exception handing for more than one kb"

* Fix docstring for qdrant.size()

* fix evals

* fix shadowing

---------

Co-authored-by: jwatson <jkwatson@gmail.com>
Co-authored-by: Michael Liu <mliu@cloudera.com>

1.4.1

Merge pull request #77 from cloudera/mob/fix-model-status

fix model status for bedrock

1.4.0

Release 1.4.0 is out! ⛷️

Major changes including:

• CAII Model Discovery. We no longer require users to specify an inference and embedding CAII model. Instead, users can now choose between any of the CAII models available in their environment.
• Switched to Bedrock Converse to support more Bedrock models
• We now support saving the inference model in a chat session
• We support saving the embedding and summarization model for a knowledge base. As part of this change, users can now disable summarization by not selecting a model.
• Added Cohere Command R+ v1 inference model
• Added Cohere Embed Multilingual v3 embedding model
• Removed Llama 3.1 405b model inference model

1.4.0-beta

Merge pull request #58 from cloudera/mob/main

Provide Sessions with Inference Model Persistence

1.3.0

Visualization, Powerpoint, Misc. Cleanup (#49)

* wip on visualizing dataset contents and vectors

* remove docling

* Make datasets represent individasiual docs

* hi lite

* make the hover point size bigger for stronger highlighting

* highlight even stronger!

* refactor and add a question entry box to show where the question lies in the vector space

* a bit of refactoring and fix unused imports

* force mypy into happiness

* Update release version to dev-testing

* fix types

* disable input while loading

* mark loading while viz is loading

* clear up some warnings

* make the active tab sticky on the ds management page

* small refactoring

* small refactor

* "wip on visualization cleanup"

* "now we're thinking with UMAP tooltips"

* "styling"

* "viz layout, continued."

* "now we're thinking with dependency hell"

* "add dependencies"

* add the last dependency needed for powerpoint parsing

* add other powerpoints

* turn on the no-kb query toggle

* tidy up the chunk metadata handling

* use models for models

* update uv.lock from main

* delete log file

---------

Co-authored-by: Michael Liu <mliu@cloudera.com>
Co-authored-by: actions-user <actions@github.com>
Co-authored-by: Elijah Williams <ewilliams@cloudera.com>

model_storage_release

This is a fake release, to provide urls to download some EasyOCR model artifacts at startup time.

DO NOT DELETE OR EDIT UNLESS YOU KNOW WHAT YOU ARE DOING

1.3.0-beta

Support parsing json and csv (#32)

* Support parsing json and csv

* parallelize requests

* try to fix deletion

* keep document id

* revert delete changes

* remove delete_document

* fix lint

1.2.0

MIsc. updates and bugfixes (#25)

* add scores to sources

* add in toggle for querying knowledge base

* wip for excluding knowledge base

* wip on llm_talk

* make direct calls to the CAII embedding model, rather than using the OpenAI library

* init the context in the super constructor call

* Fix check for CAII domain

* fix broken test compilation

* Update release version to dev-testing

* hide the knowledge base switch for now

---------

Co-authored-by: Elijah Williams <ewilliams@cloudera.com>
Co-authored-by: Michael Liu <mliu@cloudera.com>
Co-authored-by: actions-user <actions@github.com>