Release v1.7.0

• fixed source of combined classifier in dream_light pipeline #490
• fix: add chatgpt-16k to universal #492
• Upd/dff/template skill #303
• Filter entities by offsets #497
• custom entity linking #335
• Fix/config for transformers llm #500
• Feat/anthropic to prod #493
• fix: claude display name
• Fix/better llm response selector #498
• Feat/newer python dockerfile #502
• refactor: openai limit increased from 100 to 500M #503

Release v1.6.0

• Add/light dist #369
• Dockerfiles refactoring #375
• Feat/use configs and function from common #478
• google-api-skill utilizing LangChain's google api wrapper #377
• refactor: tests project name #488
• feat: tests of new prompted demo dists #480
• Feat/antropic models #479
• combined mtl lightweight #484
• Feat/utilize llm resp selector #483
• feat: add chatgpt 16k #491

Release v1.5.1

• Fix/update dff tests #473
• Feat/add gptjt model #464
• Fix/update ports #472
• Fix/configs and llm resp selection #476

Release v1.5.0

• fix: codestyle
• fix: remove space replacement #465
• fix: increase timeout to 120s #466
• Feat/description based skill selector #463
• fix: add endpoints for llms #468
• fix: prompt selector ru #467
• refactor: removed command for universal prompted skill #469
• fix: prompt to list #470
• Fix/prompt formatting #471

Release v1.4.1

• fix: use correct oasst name #461

Release v1.4.0

• fix: getting url not name #449
• feat: gpt-4 and gpt-4 32k services #456
• feat: update oasst model #411
• feat: update openai for gpt-4 #458

Release v1.3.0

• Fix/resp selector attributes #453
• Fix/prompt deeppavlov #452
• fix: upd models descriptions #454

Release v1.2.0

• feat: llms cards #448
• fix: check for env vars values #450
• Feat/better deepy #441

Release v1.1.0

• feat: thresholds for toxic_cls #437
• fix: update pip #439
• feat: smaller context for prompt selector #438
• Add compose definitions to component cards #384
• Fix/remove duplicates #443
• Multiskill update #440
• Fix/increase all timeouts #444
• fix: 5sec to 20 also for dream persona
  1eb3bac
• refactor: promp selector and prompted skill dockerfiles #446

Release v1.0.0

Most important changes:

• fixed prompt selector
• changed names, descriptions, timeouts of public prompt-based distributions and their components
• changed prompts of public prompt-based distributions
• removed spelling and blacklisted words from public prompt-based distributions
• Ranking-based Dialog Management (also utilized by public prompt-based distributions)
• LLM-based Dialog Management
• template distribution for public prompt-based distributions
• switch to a new PROXY address
• sending API KEYS via requests to agents (not only from env)
• New Multi-skill Public Assistant combining 5 prompts (inferring at most 2 in parallel)

In Details (all pull requests):

• fix: prompt selector - empty prompt! #407
• feat: remove spelling from prompted dists (#406)
• Fix/black style (#409)
• Fix/namings (#408)
• fix: model descriptions (#410)
• fix: sync dream_persona with demos (#413)
• Feat/llm-based and ranking-based dm (#405)
• fix: add proxy usage (#416)
• feat: new proxy address (#417)
• prompt asks chatbot to be concise (#418)
• added gptj to proxy file (#424)
• fix: dummy score decrease (#423)
• Feat/api keys from attributes (#415)
• prompt to ask fewer questions (#429)
• feat: universal dist requires api keys in request (#425)
• feat: template_template (#414)
• fixed cutoff for AI utterance (#426)
• feat: list of public dists (#433)
• fix: remove badlists from prompted dists (#431)
• added info about payments (#432)
• timeout 30 for fairytales (#427)
• longer generation and timeout to reduce cutoffs (#420)
• change prompts for ai faq and nutrition (#430)
• fix: template dist and description (#435)
• Feat/multiskill assistant (#434)