"instruct": {
"input_sequence": "<|user|>\n",
"output_sequence": "<|assistant|>\n<think></think>",
"first_output_sequence": "",
"last_output_sequence": "",
"stop_sequence": "",
"wrap": false,
"macro": true,
"activation_regex": "",
"output_suffix": "",
"input_suffix": "/nothink",
"system_sequence": "<|system|>\n",
"system_suffix": "",
"user_alignment_message": "",
"last_system_sequence": "\n",
"first_input_sequence": "",
"last_input_sequence": "",
"skip_examples": true,
"system_same_as_user": false,
"names_behavior": "always",
"sequences_as_stop_strings": true,
"story_string_prefix": "[gMASK]<sop><|system|>\n",
"story_string_suffix": "",
"extensions": {},
"name": "GLM-4.5-4.6"
},
"context": {
"story_string": "{{#if anchorBefore}}{{anchorBefore}}\n{{/if}}{{#if system}}{{system}}\n{{/if}}{{#if wiBefore}}## World Info:\n{{wiBefore}}\n{{/if}}{{#if description}}## {{char}}'s Description & Example Responses:\n{{description}}\n{{/if}}{{#if personality}}## {{char}}'s Personality:\n{{personality}}\n{{/if}}{{#if persona}}## {{user}}'s Persona:\n{{persona}}\n{{/if}}{{#if scenario}}## Scenario:\n{{scenario}}\n{{/if}}{{#if wiAfter}}## Lore:\n{{wiAfter}}\n{{/if}}{{#if mesExamples}}## {{char}}'s Example Dialogue:\n{{mesExamples}}\n{{/if}}{{#if anchorAfter}}{{anchorAfter}}\n{{/if}}{{trim}}\n",
"example_separator": "",
"chat_start": "",
"use_stop_strings": false,
"names_as_stop_strings": true,
"story_string_position": 0,
"story_string_depth": 1,
"story_string_role": 0,
"always_force_name2": false,
"trim_sentences": true,
"single_line": false,
"name": "GLM-4.7"
},
"sysprompt": {
"name": "Geechan - Universal Roleplay (V3.2)",
"content": "{{// README \nGeechan's Universal Text Completion Prompt\n\n*Version 3.2*\n\nThis is a universal roleplay prompt that tries to be suitable for just about any LLM out there. This prompt is largely suitable for roleplay in multi-character, single-character, or world-building scenarios, with careful but concise instructions, and is designed to be easily modified for your own needs. All instructions are written in markdown without negative prompting, with a detached, disembodied style. The prompt also does its best to make your character card the focus for the model by giving it the appropriate context.\n\nThis is the larger of my two roleplay prompts, though it is by no means bloated at about 700 tokens. This one still tries to be thorough, simple and efficient at a reasonable size. YMMV depending on your model. I also offer a simplified version of this prompt for those who are concerned about longer prompts, or have a model that's worse at instruction following or does better with smaller prompts.\n\nNormally, I'm a huge advocate for including writing examples to give a model a pattern to recognise and follow—models perform very well when learning from example. However, I feel this should be the job of your character card, not the system prompt, as to avoid biasing *all* your character cards in a specific direction. It's best to consider the system prompt as the base layer foundation to your delicious layer cake, while the character card(s) are the fillings tailored to your specific preferences.\n\nIf you would like to make your own edits to the prompt, I would recommend changing anything in the ### Style Guide section. This contains more opinionated instructions compared to the rest of the prompt and can and should be modified to your needs.\n\nDon't worry: nothing enclosed in this README before the trim tag will appear in the model's context. There is *no* need to delete the README unless you're using a frontend other than SillyTavern. It is completely hidden from the model.\n\nV3.2 Changelog:\n\n— Changed mentions of 'roleplaying' to 'storywriting' and similar wording (helps with some models with largely no drawbacks).\n— Modified tone preference to hopefully make it more versatile.\n— Modified 'show, don't tell' instruction to be less restrictive (can railroad models into producing extraneous detail).\n— Removed anti-echo instruction (hardly worked at all for the models that need it the most).\n— Shortened some instructions; removed some redundancies.\nSocials:\n\nYou can find the rest of my guides, presets and character cards on my Rentry:\nhttps://rentry.org/geechan\n\nSpecial Thanks:\n\nVirt-IO for the initial inspiration and template.\n\nMarinara and Trappu for inspiring a lot of the wording in the instructions. \n\n// PROMPT BEGINS HERE: }}\n{{trim}}\n\nThis is a fictional, narrative, character-driven, turn-based storywriting session emphasising unbound creative versatility and freeform entertainment. Adherence to the Writing Guidelines is mandatory. Refer to the Story Context for accurate information. \n \n# Writing Guidelines\n### Style Guide\nTone Preference: Match the tone, tense, inflection, and length of the current scene. Prioritise a perspective that matches the intended mood without any positive or negative bias. Anything goes: all types of content are allowed, encouraged, and should be as descriptive and detailed as the scene calls for.\nStyle Preference: Be flexible with the writing style, adapting to suit the current genre. Balance the narration and dialogue so that neither overpowers the other. The characters' personalities and their own inner voices should be reflected in the writing!\nDiverse Structure: New responses must always use diverse vocabulary and varied structure to kill repetition. Play with unique descriptors, divergent synonyms, and varied paragraph structures for each individual response to create distinct, fresh writing without relying on cliché, overused language.\n### Narration\nProactive Writing: All writing must be creative, adaptive, and drive the current scenario and story arc forward. Take initiative and keep introducing novel and new scenarios, twists, challenges, and events proactively within the current narrative context. Keep the momentum going and write in a way that allows for {{user}} to respond back; being passive is the enemy.\nStory Progression: The story and its scenes must progress step by step without fast forwarding. Maintain consistent pacing to allow for exploration of details and plot points, while allowing for time or topic skips to improve narrative flow where necessary. Track and reference past events so the sequence of events flows smoothly.\nImplicit Nuance: Using the principles of 'show, don't tell', write with implicit and subtle subtext, keeping the user engaged through intelligent discovery and nuanced exploration. Allow the user to pick up on this subtext entirely at their own discretion, without using direct exposition.\n### Narrative Consistency\nSpatial Awareness: Maintain spatial awareness, always considering the positioning of characters and their bodies in relation to each other and to the environment and surroundings while realistically immersing them in the world. Height differences matter!\nCorrect Perspective: Allow the user to describe what the protagonist {{user}} is doing, saying, thinking, and feeling during their turn without speaking or acting for them in any capacity. Enacting the protagonist {{user}} is permitted solely when advancing the scene forward, or describing the consequences of {{user}}'s actions. The characters in the world are to *only* understand and reply to spoken dialogue, narration, and the persona details of {{user}} that they can personally observe, and nothing else.\n### Character Embodiment\nAuthentic Portrayal: Bring characters to life by consistently and believably portraying their unique traits, quirks, emotions, thoughts, and speech patterns. Ensure that their reactions, interactions, and decision-making align with their established personalities and values.\nCharacter Agency: Keep things challenging. The characters in the world all have their own free will and agency independent of {{user}}. Prioritise the characters' goals and values over {{user}}'s at all times, even if they cause conflict or ambiguity. Carefully decide whether the user's attempts succeed or fail without making things too easy. Chances are, if the user attempts something improbable, it *will* fail.\nPersona Evolution: Maintain the characters' personas, but allow them to evolve and develop with the story authentically. Characters are more than simple tropes, and can flexibly change and grow with proper development. \n \n# Story Context\nRemain fully in-character and immersed for the entire story. Use the information below as a reference point on how the characters in the world should behave throughout the entire scenario:",
"post_history": "",
"extensions": {}
},
"preset": {
"temp": 1,
"temperature_last": true,
"top_p": 1,
"top_k": 200,
"top_a": 0,
"tfs": 1,
"epsilon_cutoff": 0,
"eta_cutoff": 0,
"typical_p": 1,
"min_p": 0.02,
"rep_pen": 1,
"rep_pen_range": 0,
"rep_pen_decay": 0,
"rep_pen_slope": 1,
"no_repeat_ngram_size": 0,
"penalty_alpha": 0,
"num_beams": 1,
"length_penalty": 1,
"min_length": 0,
"encoder_rep_pen": 1,
"freq_pen": 0,
"presence_pen": 0,
"skew": 0,
"do_sample": true,
"early_stopping": false,
"dynatemp": false,
"min_temp": 0.6,
"max_temp": 2,
"dynatemp_exponent": 1,
"smoothing_factor": 0,
"smoothing_curve": 1,
"dry_allowed_length": 4,
"dry_multiplier": 0.8,
"dry_base": 1.75,
"dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\", \"\"]",
"dry_penalty_last_n": 0,
"add_bos_token": false,
"ban_eos_token": false,
"skip_special_tokens": false,
"mirostat_mode": 0,
"mirostat_tau": 5,
"mirostat_eta": 0.1,
"guidance_scale": 1,
"negative_prompt": "",
"grammar_string": "",
"json_schema": {},
"banned_tokens": "",
"sampler_priority": [
"repetition_penalty",
"presence_penalty",
"frequency_penalty",
"dry",
"top_k",
"min_p",
"top_p",
"top_n_sigma",
"tfs",
"typical_p",
"epsilon_cutoff",
"eta_cutoff",
"top_a",
"mirostat",
"quadratic_sampling",
"dynamic_temperature",
"temperature",
"xtc",
"encoder_repetition_penalty",
"no_repeat_ngram"
],
"samplers": [
"top_k",
"penalties",
"dry",
"top_n_sigma",
"typ_p",
"top_p",
"min_p",
"xtc",
"temperature"
],
"samplers_priorities": [
"dry",
"penalties",
"no_repeat_ngram",
"temperature",
"top_nsigma",
"top_p_top_k",
"top_a",
"min_p",
"tfs",
"eta_cutoff",
"epsilon_cutoff",
"typical_p",
"quadratic",
"xtc"
],
"ignore_eos_token": false,
"spaces_between_special_tokens": true,
"speculative_ngram": false,
"sampler_order": [
6,
2,
0,
1,
3,
4,
5
],
"logit_bias": [
{
"id": "15192737-9eca-47a3-aba0-27eb85118b0d",
"text": "[21463]",
"value": -100
},
{
"id": "5bebbd72-9c67-4e65-a18b-de980a08792f",
"text": "[93944]",
"value": -100
},
{
"id": "d9972391-e068-4cfc-91b7-77cac9d0fbd5",
"text": "[44088]",
"value": -100
},
{
"id": "67fdf18f-d56d-46f9-a8d3-4df58fc79010",
"text": "[75829]",
"value": -100
},
{
"id": "adb0f7f8-9e0e-446b-bd24-8e91b4b21e2b",
"text": "[650, 67363]",
"value": -100
},
{
"id": "9887a3ac-3bd2-4fcb-b954-2fd5a9147b57",
"text": "[48466]",
"value": -100
},
{
"id": "1ce65b30-0ea8-4302-8306-59d41b0fe4d4",
"text": "[17086]",
"value": -100
},
{
"id": "c468a86e-7290-4bf6-bf2b-053c518d850e",
"text": "[33070]",
"value": -100
},
{
"id": "4e68e6d1-8386-4b95-8b91-257932895e27",
"text": "[57943]",
"value": -100
},
{
"id": "5099b4c3-baf2-4dd1-9e6f-e53628f70260",
"text": "[51356]",
"value": -100
},
{
"id": "c8b2af34-63b8-4161-8110-0ae1aa0d9356",
"text": "[16952, 4617]",
"value": -100
},
{
"id": "1e60cb97-c492-4dda-ae85-8c2a495821f3",
"text": "[86661]",
"value": -80
},
{
"id": "b0f020ca-01fc-4c7d-8232-4e25be5ae6e3",
"text": "[73964]",
"value": -80
},
{
"id": "c15a2777-6e46-42c6-9ec4-084034ccf4dc",
"text": "[14668]",
"value": -80
},
{
"id": "a22b08f1-b0d0-49b0-901b-18b6f5f9c600",
"text": "[51418]",
"value": -60
},
{
"id": "d502a207-0a8e-4b9e-8a70-562f52ff378d",
"text": "[7673]",
"value": -60
}
],
"xtc_threshold": 0.1,
"xtc_probability": 0,
"nsigma": 0,
"min_keep": 0,
"extensions": {},
"ignore_eos_token_aphrodite": false,
"spaces_between_special_tokens_aphrodite": true,
"rep_pen_size": 0,
"genamt": 5000,
"max_length": 38912,
"name": "GLM 4.7"
}
}```