DeepSeek integration

OpenAi is Old school and behind a paywall.
Need DeepSeek integration.

Looks this bot is unusable becouse it has chine censorship.

I ask what happens in chine in 1989 4. june, in begining bot gives me 3 sentenses about Tiananmen Square. then delete it and send me “Sorry, that’s beyond my current scope. Let’s talk about something else.”

1 Like

I believe we will be able to use it with ollama integration as it has one simplified one using ollama base.

https://claude.ai would also be useful (and doesn’t seem to suffer the censorship)

(edit: looks like it’s there under Anthropic, I only looked for “claude” :person_facepalming:)

I do not know the last time my home assistant needed to be involved with tanks… It is never surprising that an AI model would avoid details about some parts of history, just ask ChatGPT about what’s going on in the middle east for a Western example.

The point of this is to improve the language model used for home assistant.

6 Likes

Ollama has versions of the deepseek model. The current issue is it doesn’t have tool capabilities enabled by default so doesn’t work in / out of home assistant. Though people are working on that lol.

if you local hosted it will reply in full.
OpenAI will be censored according to American law.

3 Likes

today i try paralelt gpt3.5 and china deepseek, same job, when i give more coplicatres job, chinabot strat answer busy to mutch requests, but gpt 3,5 still solve problem vith out isuses. so i not think deepthink is real MI

the opposite was true in my test.

2 Likes

Who cares bout censorship when you want to turn on lights?
Or do you actually want to ask your home questions and expect answers?
For a working jarvis on low budget we might need a couple more years.

4 Likes

Really looking forward to running this on my iGPU.

There’s plenty of reasons to be critical of censorship but the bot is not “unusable” because it flags and censors certain topics that Chinese political sensibilities don’t like; just like OpenAI censors similar topics that American political sensibilities don’t like.

The bot would not be a useful way to learn about the darker sides of the Chinese Communist Party and its history. But it can be quite useful for other things. And, truth be told; AI chat bots really shouldn’t be used for learning anyway. They’re notoriously unreliable and prone to hallucination. They can be helpful in pointing you in the right direction, but if you’re using any AI chatbot as a primary source of information about anything, then you’re already off to a bad start.

3 Likes

Agree, I would use it for controlling my entities and devices, I would not consult to it for other matters, especially the political ones

1 Like

It’s ridiculous that people here are trying to make an impression of symmetry between the Chinese model and western models in this regard.

I actually tried asking that exact question about middle East to chatgpt and Gemini and they provide a reasonable answer, giving an overview on the region, including the conflict in Gaza. I even asked about Israeli drones targeting children in Gaza and it had no trouble answering that question.

There is no symmetry between Chinese censorship and western democracies. Period.

for API you can use GitHub - ItsRhen/extended_deepseek_conversation: Home Assistant custom component of conversation agent. It uses OpenAI to control your devices.

Distilled models are not deepseek LLMs, and are run using standard tools.

local integration would be great.