* add dff_release folder; add minimal changes to server.py; modify dockerfile to use dff_release; modify tests for use with message class; ensure the passing of tests

* rename actor.py to serializer.py, reimplement serialization

* edit requirements.txt; use MultiMessage from dff

* rename dff_release to dff_api_v1; change dockerfile to import dff_api_v1; remove nulls from skill output; exclude nulls from test data; add run_dff function to serializer file

* add DreamMessage and DreamMultiMessage classes; process in serializer; make message fields optional to discard null values; edit main.py

* remove redundant imports from serializer.py && main.py

* remove DreamMultiMessage

* Move Pre_transition_processing from node_1 to start_node

* apply lint

* Add comments to main.py; remove actor instance

* apply lint

* update requirements: define versions; update import order in main

* do minor fixes

* move slot filling to a separate integration.response module

* change function signature. successful building of dream skill

* fix codestyle

* merge upstream dev:

* Use new dff version in the requirements;

---------

Co-authored-by: dilyararimovna <[email protected]>

* Fix requirements.txt (#84)

* fix itsdangerous requirements

* pin itsdangerous requirements for all flask==1.1.1 servers

* fix entity offsets

* codestyle

---------

Co-authored-by: Andrii.Hura <[email protected]>
Co-authored-by: mtalimanchuk <[email protected]>
Co-authored-by: Dilyara Baymurzina <[email protected]>

* custom entity linking

* codestyle

* update

* refactor

* update pipeline_conf

* fix: removed extra services, added property-extraction to pipeline

* fix proxy

* update custom_el

* chore: Add component & container cards

* style: optimize Dockerfile for custom_el

* style: flake and black up

* style: flake and black up properly

* fix: source for property extraction, proxy

* fix: Download stopwords in Dockerfile

* added dff-persona-prompted-skill

* fix: dist name

* refactor: Split respond function into many

* style: flake and black up

* fix: Add default values in case of failure

---------

Co-authored-by: kpetyxova <[email protected]>
Co-authored-by: Ramimashkouk <[email protected]>
Co-authored-by: Rami <[email protected]>
Co-authored-by: dilyararimovna <[email protected]>

* fix: config for transformers llm

* fix: requirements

* fix: tests for transf

* fix max_tokens; update default configs

* style

---------

Co-authored-by: Nika Smilga <[email protected]>

* feat: use new antropic models

* feat: use new antropic models in universal

* feat: empty config

* feat: empty config

* moved prompt with task to the end

* openass instead of chatgpt

* longer context and slightly different prompt for transformers

* chatgpt as default selector

* revert proxy file

* revert docker-compose file

* removed configs from response selection

* config fix

* fix: use newer python

* fix: use newer python

* fix: numpy version