Can you share the horde template in previous version? The new beta seems to have deleted it? Or it's a bug because I saw the horde.ts file is still there. Anyway, thanks.

The link is broken. I found the.md file in the /lib folder.

Got weird black screen when importing the template file. No errors found in log

Is there a public api I can use to query the model sending queries via Tasker to model (gemma-3-4b-it-q4_k_m.gguf offline) through ChatterUI. My goal is to interact with model via voice commands, hands free using Tasker as sort of a middleman. when the model replies, Tasker will intercept reply, parce relevant data and trigger any Tasker tasks, etc.which are setup to be activated based on model's replies.
Example:
I activate voice recognition and give query to Tasker via get voice or AutoVoice, etc.
"JARVIS . (Trigger) Tell me a funny joke to start the day and then I need the weather Forcast for the day."
Tasker sends the query to the model who processes the request, sends reply back to Tasker which will read joke Aloud via "say" action, and fire up the weather forecast task. I could of course obviously just load the app up on screen and type away to the model and wait for my reply and then just fire up the weather but that's no fun and not very cool when you can converse with the "assistant" while you're scrolling through Instagram or something and Tasker is firing up weather to be spoken to me Aloud while I'm in another app as well. Front end is great for sitting down and reading responses comfortably but if you're trying to multitask on the go, a basically handsfree conversation with the model and the model's replies helping trigger useful tasker tasks would be a Game changer.

Outstanding app, btw.
It IS my daily driver! 💯

I'm having a lot of trouble figuring out how to make this custom template work.

For reference, I'm running Bitnet on ik_llama.cpp server in termux.

input:

{
	"prompt": "<|im_start|>system\nYou are a helpful assistant<|im_end|>\n<|im_start|>user\nHello, who are you?<|im_end|>\n<|im_start|>assistant\n",
	"temperature": 0.7,
	"n_predict": 128,
	"stop": [
		"<|im_end|>"
	]
}

output:

{
 "content": "I am an AI assistant designed to help you find information and answer your questions. How can I assist you today? ",
 "id_slot": 0,
 "stop": true,
 "model": "./models/bitnet.gguf",
 "tokens_predicted": 28,
 "tokens_evaluated": 47,
 "generation_settings": {
 "n_ctx": 4096,
 "n_predict": -1,
 "model": "./models/bitnet.gguf",
 "seed": 4294967295,
 "temperature": 0.699999988079071,
 "dynatemp_range": 0,
 "dynatemp_exponent": 1,
 "top_k": 40,
 "top_p": 0.949999988079071,
 "min_p": 0.05000000074505806,
 "tfs_z": 1,
 "typical_p": 1,
 "repeat_last_n": 64,
 "repeat_penalty": 1,
 "presence_penalty": 0,
 "frequency_penalty": 0,
 "penalty_prompt_tokens": [],
 "use_penalty_prompt_tokens": false,
 "mirostat": 0,
 "mirostat_tau": 5,
 "mirostat_eta": 0.10000000149011612,
 "penalize_nl": false,
 "stop": [
 "<|im_end|>"
 ],
 "n_keep": 0,
 "n_discard": 0,
 "ignore_eos": false,
 "stream": false,
 "logit_bias": [],
 "n_probs": 0,
 "min_keep": 0,
 "grammar": "",
 "samplers": [
 "top_k",
 "tfs_z",
 "typical_p",
 "top_p",
 "min_p",
 "temperature"
 ]
 },
 "prompt": "<|im_start|>system\nYou are a helpful assistant<|im_end|>\n<|im_start|>user\nHello, who are you?<|im_end|>\n<|im_start|>assistant\n",
 "truncated": false,
 "stopped_eos": false,
 "stopped_word": true,
 "stopped_limit": false,
 "stopping_word": "<|im_end|>",
 "tokens_cached": 74,
 "timings": {
 "prompt_n": 47,
 "prompt_ms": 4828.648,
 "prompt_per_token_ms": 102.7371914893617,
 "prompt_per_second": 9.733573455758215,
 "predicted_n": 28,
 "predicted_ms": 4568.995,
 "predicted_per_token_ms": 163.17839285714285,
 "predicted_per_second": 6.128262342156208
 }
}

I have managed to get a correct payload to send and response from my server, but chatter seems unable to parse it.

The template guide it seems is focused on request formatting, but not response parsing. Not sure if I'm overlooking something?

Does anyone have a template for Google's API that includes the safetySettings parameters?

Could you please add Deepseek and Grok to the list of API providers?

Chutes anyone?
I can't connect to chutes API directly.

I use qwen 3.5 9b model on llmacpp server. It works perfectly fine without thinking.

How do I make the template work with thinking tough?

Any ideas?