lollms-client 0.33.0__py3-none-any.whl → 1.1.0__py3-none-any.whl

lollms_client/lollms_llm_binding.py

@@ -12,11 +12,42 @@ from lollms_client.lollms_types import MSG_TYPE
 from lollms_client.lollms_discussion import LollmsDiscussion
 from lollms_client.lollms_utilities import ImageTokenizer
 import re
+import yaml
+
+from pathlib import Path
+import json
+
+def load_known_contexts():
+    """
+    Loads the known_contexts data from a JSON file.
+
+    Args:
+        file_path (str): The path to the JSON file.
+
+    Returns:
+        dict: A dictionary containing the known_contexts data, or None if an error occurs.
+    """
+    try:
+        file_path = Path(__file__).parent / "assets" / "models_ctx_sizes.json"
+        with open(file_path, "r") as f:
+            known_contexts = json.load(f)
+        return known_contexts
+    except FileNotFoundError:
+        print(f"Error: File not found at {file_path}")
+        return None
+    except json.JSONDecodeError:
+        print(f"Error: Could not decode JSON from {file_path}")
+        return None
+    except Exception as e:
+        print(f"An unexpected error occurred: {e}")
+        return None
+
 class LollmsLLMBinding(ABC):
     """Abstract base class for all LOLLMS LLM bindings"""
     
     def __init__(self, 
-                 binding_name: Optional[str] ="unknown"
+                 binding_name: Optional[str] ="unknown",
+                 **kwargs
         ):
         """
         Initialize the LollmsLLMBinding base class.
@@ -26,6 +57,18 @@ class LollmsLLMBinding(ABC):
         """
         self.binding_name=binding_name
         self.model_name = None #Must be set by the instance
+        self.default_ctx_size = kwargs.get("ctx_size") 
+        self.default_n_predict = kwargs.get("n_predict")
+        self.default_stream = kwargs.get("stream")
+        self.default_temperature = kwargs.get("temperature")
+        self.default_top_k = kwargs.get("top_k")
+        self.default_top_p = kwargs.get("top_p")
+        self.default_repeat_penalty = kwargs.get("repeat_penalty")
+        self.default_repeat_last_n = kwargs.get("repeat_last_n")
+        self.default_seed = kwargs.get("seed")
+        self.default_n_threads = kwargs.get("n_threads")
+        self.default_streaming_callback = kwargs.get("streaming_callback")
+
     
     @abstractmethod
     def generate_text(self,
@@ -168,378 +211,7 @@ class LollmsLLMBinding(ABC):
 
         # Hardcoded context sizes for popular models. More specific names (e.g., 'llama3.1')
         # should appear, as they will be checked first due to the sorting logic below.
-        known_contexts = {
-            'agentica-org/deepcoder-14b-preview': 8192,
-            'agentica-org/deepcoder-14b-preview:free': 8192,
-            'ai21/jamba-large-1.7': 256000,
-            'ai21/jamba-mini-1.7': 256000,
-            'aion-labs/aion-1.0': 8192,
-            'aion-labs/aion-1.0-mini': 8192,
-            'aion-labs/aion-rp-llama-3.1-8b': 131072,
-            'alfredpros/codellama-7b-instruct-solidity': 16384,
-            'alpindale/goliath-120b': 4096,
-            'amazon/nova-lite-v1': 32768,
-            'amazon/nova-micro-v1': 32768,
-            'amazon/nova-pro-v1': 32768,
-            'anthracite-org/magnum-v2-72b': 131072,
-            'anthracite-org/magnum-v4-72b': 131072,
-            'anthropic/claude-3-haiku': 200000,
-            'anthropic/claude-3-haiku:beta': 200000,
-            'anthropic/claude-3-opus': 200000,
-            'anthropic/claude-3-opus:beta': 200000,
-            'anthropic/claude-3.5-haiku': 200000,
-            'anthropic/claude-3.5-haiku-20241022': 200000,
-            'anthropic/claude-3.5-haiku:beta': 200000,
-            'anthropic/claude-3.5-sonnet': 200000,
-            'anthropic/claude-3.5-sonnet-20240620': 200000,
-            'anthropic/claude-3.5-sonnet-20240620:beta': 200000,
-            'anthropic/claude-3.5-sonnet:beta': 200000,
-            'anthropic/claude-3.7-sonnet': 200000,
-            'anthropic/claude-3.7-sonnet:beta': 200000,
-            'anthropic/claude-3.7-sonnet:thinking': 200000,
-            'anthropic/claude-opus-4': 200000,
-            'anthropic/claude-opus-4.1': 200000,
-            'anthropic/claude-sonnet-4': 200000,
-            'arcee-ai/coder-large': 32768,
-            'arcee-ai/maestro-reasoning': 32768,
-            'arcee-ai/spotlight': 32768,
-            'arcee-ai/virtuoso-large': 32768,
-            'arliai/qwq-32b-arliai-rpr-v1': 8192,
-            'arliai/qwq-32b-arliai-rpr-v1:free': 8192,
-            'baidu/ernie-4.5-300b-a47b': 128000,
-            'bytedance/ui-tars-1.5-7b': 8192,
-            'cognitivecomputations/dolphin-mistral-24b-venice-edition:free': 32768,
-            'cognitivecomputations/dolphin-mixtral-8x22b': 65536,
-            'cognitivecomputations/dolphin3.0-mistral-24b': 32768,
-            'cognitivecomputations/dolphin3.0-mistral-24b:free': 32768,
-            'cognitivecomputations/dolphin3.0-r1-mistral-24b': 32768,
-            'cognitivecomputations/dolphin3.0-r1-mistral-24b:free': 32768,
-            'cohere/command': 8192,
-            'cohere/command-a': 8192,
-            'cohere/command-r': 128000,
-            'cohere/command-r-03-2024': 128000,
-            'cohere/command-r-08-2024': 128000,
-            'cohere/command-r-plus': 128000,
-            'cohere/command-r-plus-04-2024': 128000,
-            'cohere/command-r-plus-08-2024': 128000,
-            'cohere/command-r7b-12-2024': 128000,
-            'deepseek/deepseek-chat': 32768,
-            'deepseek/deepseek-chat-v3-0324': 32768,
-            'deepseek/deepseek-chat-v3-0324:free': 32768,
-            'deepseek/deepseek-prover-v2': 131072,
-            'deepseek/deepseek-r1': 32768,
-            'deepseek/deepseek-r1-0528': 32768,
-            'deepseek/deepseek-r1-0528-qwen3-8b': 32768,
-            'deepseek/deepseek-r1-0528-qwen3-8b:free': 32768,
-            'deepseek/deepseek-r1-0528:free': 32768,
-            'deepseek/deepseek-r1-distill-llama-70b': 131072,
-            'deepseek/deepseek-r1-distill-llama-70b:free': 131072,
-            'deepseek/deepseek-r1-distill-llama-8b': 131072,
-            'deepseek/deepseek-r1-distill-qwen-1.5b': 32768,
-            'deepseek/deepseek-r1-distill-qwen-14b': 32768,
-            'deepseek/deepseek-r1-distill-qwen-14b:free': 32768,
-            'deepseek/deepseek-r1-distill-qwen-32b': 32768,
-            'deepseek/deepseek-r1-distill-qwen-7b': 32768,
-            'deepseek/deepseek-r1:free': 32768,
-            'deepseek/deepseek-v3-base': 32768,
-            'eleutherai/llemma_7b': 8192,
-            'featherless/qwerky-72b:free': 8192,
-            'google/gemini-2.0-flash-001': 1000000,
-            'google/gemini-2.0-flash-exp:free': 1000000,
-            'google/gemini-2.0-flash-lite-001': 1000000,
-            'google/gemini-2.5-flash': 1000000,
-            'google/gemini-2.5-flash-lite': 1000000,
-            'google/gemini-2.5-flash-lite-preview-06-17': 1000000,
-            'google/gemini-2.5-pro': 2000000,
-            'google/gemini-2.5-pro-exp-03-25': 2000000,
-            'google/gemini-2.5-pro-preview': 2000000,
-            'google/gemini-2.5-pro-preview-05-06': 2000000,
-            'google/gemini-flash-1.5': 1000000,
-            'google/gemini-flash-1.5-8b': 1000000,
-            'google/gemini-pro-1.5': 2000000,
-            'google/gemma-2-27b-it': 8192,
-            'google/gemma-2-9b-it': 8192,
-            'google/gemma-2-9b-it:free': 8192,
-            'google/gemma-3-12b-it': 131072,
-            'google/gemma-3-12b-it:free': 131072,
-            'google/gemma-3-27b-it': 131072,
-            'google/gemma-3-27b-it:free': 131072,
-            'google/gemma-3-4b-it': 131072,
-            'google/gemma-3-4b-it:free': 131072,
-            'google/gemma-3n-e2b-it:free': 131072,
-            'google/gemma-3n-e4b-it': 131072,
-            'google/gemma-3n-e4b-it:free': 131072,
-            'gryphe/mythomax-l2-13b': 4096,
-            'inception/mercury': 32768,
-            'inception/mercury-coder': 32768,
-            'infermatic/mn-inferor-12b': 8192,
-            'inflection/inflection-3-pi': 128000,
-            'inflection/inflection-3-productivity': 128000,
-            'liquid/lfm-3b': 8192,
-            'liquid/lfm-40b': 8192,
-            'liquid/lfm-7b': 8192,
-            'mancer/weaver': 8192,
-            'meta-llama/llama-3-70b-instruct': 8192,
-            'meta-llama/llama-3-8b-instruct': 8192,
-            'meta-llama/llama-3.1-405b': 131072,
-            'meta-llama/llama-3.1-405b-instruct': 131072,
-            'meta-llama/llama-3.1-405b-instruct:free': 131072,
-            'meta-llama/llama-3.1-70b-instruct': 131072,
-            'meta-llama/llama-3.1-8b-instruct': 131072,
-            'meta-llama/llama-3.2-11b-vision-instruct': 131072,
-            'meta-llama/llama-3.2-11b-vision-instruct:free': 131072,
-            'meta-llama/llama-3.2-1b-instruct': 131072,
-            'meta-llama/llama-3.2-3b-instruct': 131072,
-            'meta-llama/llama-3.2-3b-instruct:free': 131072,
-            'meta-llama/llama-3.2-90b-vision-instruct': 131072,
-            'meta-llama/llama-3.3-70b-instruct': 131072,
-            'meta-llama/llama-3.3-70b-instruct:free': 131072,
-            'meta-llama/llama-4-maverick': 131072,
-            'meta-llama/llama-4-scout': 131072,
-            'meta-llama/llama-guard-2-8b': 8192,
-            'meta-llama/llama-guard-3-8b': 131072,
-            'meta-llama/llama-guard-4-12b': 131072,
-            'microsoft/mai-ds-r1': 32768,
-            'microsoft/mai-ds-r1:free': 32768,
-            'microsoft/phi-3-medium-128k-instruct': 131072,
-            'microsoft/phi-3-mini-128k-instruct': 131072,
-            'microsoft/phi-3.5-mini-128k-instruct': 131072,
-            'microsoft/phi-4': 131072,
-            'microsoft/phi-4-multimodal-instruct': 131072,
-            'microsoft/phi-4-reasoning-plus': 131072,
-            'microsoft/wizardlm-2-8x22b': 65536,
-            'minimax/minimax-01': 200000,
-            'minimax/minimax-m1': 200000,
-            'mistralai/codestral-2501': 32768,
-            'mistralai/codestral-2508': 32768,
-            'mistralai/devstral-medium': 32768,
-            'mistralai/devstral-small': 32768,
-            'mistralai/devstral-small-2505': 32768,
-            'mistralai/devstral-small-2505:free': 32768,
-            'mistralai/magistral-medium-2506': 32768,
-            'mistralai/magistral-medium-2506:thinking': 32768,
-            'mistralai/magistral-small-2506': 32768,
-            'mistralai/ministral-3b': 32768,
-            'mistralai/ministral-8b': 32768,
-            'mistralai/mistral-7b-instruct': 32768,
-            'mistralai/mistral-7b-instruct-v0.1': 8192,
-            'mistralai/mistral-7b-instruct-v0.2': 32768,
-            'mistralai/mistral-7b-instruct-v0.3': 32768,
-            'mistralai/mistral-7b-instruct:free': 32768,
-            'mistralai/mistral-large': 32768,
-            'mistralai/mistral-large-2407': 128000,
-            'mistralai/mistral-large-2411': 128000,
-            'mistralai/mistral-medium-3': 32768,
-            'mistralai/mistral-nemo': 128000,
-            'mistralai/mistral-nemo:free': 128000,
-            'mistralai/mistral-saba': 32768,
-            'mistralai/mistral-small': 32768,
-            'mistralai/mistral-small-24b-instruct-2501': 32768,
-            'mistralai/mistral-small-24b-instruct-2501:free': 32768,
-            'mistralai/mistral-small-3.1-24b-instruct': 32768,
-            'mistralai/mistral-small-3.1-24b-instruct:free': 32768,
-            'mistralai/mistral-small-3.2-24b-instruct': 32768,
-            'mistralai/mistral-small-3.2-24b-instruct:free': 32768,
-            'mistralai/mistral-tiny': 32768,
-            'mistralai/mixtral-8x22b-instruct': 65536,
-            'mistralai/mixtral-8x7b-instruct': 32768,
-            'mistralai/pixtral-12b': 128000,
-            'mistralai/pixtral-large-2411': 128000,
-            'moonshotai/kimi-dev-72b:free': 200000,
-            'moonshotai/kimi-k2': 200000,
-            'moonshotai/kimi-k2:free': 200000,
-            'moonshotai/kimi-vl-a3b-thinking': 200000,
-            'moonshotai/kimi-vl-a3b-thinking:free': 200000,
-            'morph/morph-v3-fast': 8192,
-            'morph/morph-v3-large': 8192,
-            'neversleep/llama-3-lumimaid-70b': 8192,
-            'neversleep/llama-3.1-lumimaid-8b': 131072,
-            'neversleep/noromaid-20b': 32768,
-            'nousresearch/deephermes-3-llama-3-8b-preview:free': 8192,
-            'nousresearch/deephermes-3-mistral-24b-preview': 32768,
-            'nousresearch/hermes-2-pro-llama-3-8b': 8192,
-            'nousresearch/hermes-3-llama-3.1-405b': 131072,
-            'nousresearch/hermes-3-llama-3.1-70b': 131072,
-            'nousresearch/nous-hermes-2-mixtral-8x7b-dpo': 32768,
-            'nvidia/llama-3.1-nemotron-70b-instruct': 131072,
-            'nvidia/llama-3.1-nemotron-ultra-253b-v1': 131072,
-            'nvidia/llama-3.1-nemotron-ultra-253b-v1:free': 131072,
-            'nvidia/llama-3.3-nemotron-super-49b-v1': 131072,
-            'openai/chatgpt-4o-latest': 128000,
-            'openai/codex-mini': 2048,
-            'openai/gpt-3.5-turbo': 4096,
-            'openai/gpt-3.5-turbo-0613': 4096,
-            'openai/gpt-3.5-turbo-16k': 16384,
-            'openai/gpt-3.5-turbo-instruct': 4096,
-            'openai/gpt-4': 8192,
-            'openai/gpt-4-0314': 8192,
-            'openai/gpt-4-1106-preview': 128000,
-            'openai/gpt-4-turbo': 128000,
-            'openai/gpt-4-turbo-preview': 128000,
-            'openai/gpt-4.1': 128000,
-            'openai/gpt-4.1-mini': 128000,
-            'openai/gpt-4.1-nano': 128000,
-            'openai/gpt-4o': 128000,
-            'openai/gpt-4o-2024-05-13': 128000,
-            'openai/gpt-4o-2024-08-06': 128000,
-            'openai/gpt-4o-2024-11-20': 128000,
-            'openai/gpt-4o-mini': 128000,
-            'openai/gpt-4o-mini-2024-07-18': 128000,
-            'openai/gpt-4o-mini-search-preview': 128000,
-            'openai/gpt-4o-search-preview': 128000,
-            'openai/gpt-4o:extended': 128000,
-            'openai/gpt-5': 200000,
-            'openai/gpt-5-chat': 200000,
-            'openai/gpt-5-mini': 200000,
-            'openai/gpt-5-nano': 200000,
-            'openai/gpt-oss-120b': 128000,
-            'openai/gpt-oss-20b': 128000,
-            'openai/gpt-oss-20b:free': 128000,
-            'openai/o1': 128000,
-            'openai/o1-mini': 128000,
-            'openai/o1-mini-2024-09-12': 128000,
-            'openai/o1-pro': 128000,
-            'openai/o3': 200000,
-            'openai/o3-mini': 200000,
-            'openai/o3-mini-high': 200000,
-            'openai/o3-pro': 200000,
-            'openai/o4-mini': 128000,
-            'openai/o4-mini-high': 128000,
-            'opengvlab/internvl3-14b': 8192,
-            'openrouter/auto': 8192,
-            'perplexity/r1-1776': 32768,
-            'perplexity/sonar': 32768,
-            'perplexity/sonar-deep-research': 32768,
-            'perplexity/sonar-pro': 32768,
-            'perplexity/sonar-reasoning': 32768,
-            'perplexity/sonar-reasoning-pro': 32768,
-            'pygmalionai/mythalion-13b': 4096,
-            'qwen/qwen-2-72b-instruct': 32768,
-            'qwen/qwen-2.5-72b-instruct': 131072,
-            'qwen/qwen-2.5-72b-instruct:free': 131072,
-            'qwen/qwen-2.5-7b-instruct': 131072,
-            'qwen/qwen-2.5-coder-32b-instruct': 131072,
-            'qwen/qwen-2.5-coder-32b-instruct:free': 131072,
-            'qwen/qwen-2.5-vl-7b-instruct': 131072,
-            'qwen/qwen-max': 32768,
-            'qwen/qwen-plus': 32768,
-            'qwen/qwen-turbo': 8192,
-            'qwen/qwen-vl-max': 32768,
-            'qwen/qwen-vl-plus': 32768,
-            'qwen/qwen2.5-vl-32b-instruct': 131072,
-            'qwen/qwen2.5-vl-32b-instruct:free': 131072,
-            'qwen/qwen2.5-vl-72b-instruct': 131072,
-            'qwen/qwen2.5-vl-72b-instruct:free': 131072,
-            'qwen/qwen3-14b': 32768,
-            'qwen/qwen3-14b:free': 32768,
-            'qwen/qwen3-235b-a22b': 32768,
-            'qwen/qwen3-235b-a22b-2507': 32768,
-            'qwen/qwen3-235b-a22b-thinking-2507': 32768,
-            'qwen/qwen3-235b-a22b:free': 32768,
-            'qwen/qwen3-30b-a3b': 32768,
-            'qwen/qwen3-30b-a3b-instruct-2507': 32768,
-            'qwen/qwen3-30b-a3b:free': 32768,
-            'qwen/qwen3-32b': 32768,
-            'qwen/qwen3-4b:free': 32768,
-            'qwen/qwen3-8b': 32768,
-            'qwen/qwen3-8b:free': 32768,
-            'qwen/qwen3-coder': 32768,
-            'qwen/qwen3-coder:free': 32768,
-            'qwen/qwq-32b': 32768,
-            'qwen/qwq-32b-preview': 32768,
-            'qwen/qwq-32b:free': 32768,
-            'raifle/sorcererlm-8x22b': 65536,
-            'rekaai/reka-flash-3:free': 128000,
-            'sao10k/l3-euryale-70b': 8192,
-            'sao10k/l3-lunaris-8b': 8192,
-            'sao10k/l3.1-euryale-70b': 131072,
-            'sao10k/l3.3-euryale-70b': 131072,
-            'sarvamai/sarvam-m:free': 8192,
-            'scb10x/llama3.1-typhoon2-70b-instruct': 131072,
-            'shisa-ai/shisa-v2-llama3.3-70b': 131072,
-            'shisa-ai/shisa-v2-llama3.3-70b:free': 131072,
-            'sophosympatheia/midnight-rose-70b': 4096,
-            'switchpoint/router': 8192,
-            'tencent/hunyuan-a13b-instruct': 8192,
-            'tencent/hunyuan-a13b-instruct:free': 8192,
-            'thedrummer/anubis-70b-v1.1': 8192,
-            'thedrummer/anubis-pro-105b-v1': 8192,
-            'thedrummer/rocinante-12b': 8192,
-            'thedrummer/skyfall-36b-v2': 8192,
-            'thedrummer/unslopnemo-12b': 128000,
-            'thedrummer/valkyrie-49b-v1': 8192,
-            'thudm/glm-4-32b': 2000000,
-            'thudm/glm-4.1v-9b-thinking': 2000000,
-            'thudm/glm-z1-32b:free': 2000000,
-            'tngtech/deepseek-r1t-chimera': 32768,
-            'tngtech/deepseek-r1t-chimera:free': 32768,
-            'tngtech/deepseek-r1t2-chimera:free': 32768,
-            'undi95/remm-slerp-l2-13b': 4096,
-            'x-ai/grok-2-1212': 128000,
-            'x-ai/grok-2-vision-1212': 128000,
-            'x-ai/grok-3': 128000,
-            'x-ai/grok-3-beta': 128000,
-            'x-ai/grok-3-mini': 128000,
-            'x-ai/grok-3-mini-beta': 128000,
-            'x-ai/grok-4': 128000,
-            'x-ai/grok-vision-beta': 128000,
-            'z-ai/glm-4-32b': 2000000,
-            'z-ai/glm-4.5': 2000000,
-            'z-ai/glm-4.5-air': 2000000,
-            'z-ai/glm-4.5-air:free': 2000000,
-            'llama3.1': 131072,   # Llama 3.1 extended context
-            'llama3.2': 131072,   # Llama 3.2 extended context
-            'llama3.3': 131072,   # Assuming similar to 3.1/3.2
-            'llama3': 8192,       # Llama 3 default
-            'llama2': 4096,       # Llama 2 default
-            'mixtral8x22b': 65536, # Mixtral 8x22B default
-            'mixtral': 32768,     # Mixtral 8x7B default
-            'mistral': 32768,     # Mistral 7B v0.2+ default
-            'gemma3': 131072,     # Gemma 3 with 128K context
-            'gemma2': 8192,       # Gemma 2 default
-            'gemma': 8192,        # Gemma default
-            'phi3': 131072,       # Phi-3 variants often use 128K (mini/medium extended)
-            'phi2': 2048,         # Phi-2 default
-            'phi': 2048,          # Phi default (older)
-            'qwen2.5': 131072,    # Qwen2.5 with 128K
-            'qwen2': 32768,       # Qwen2 default for 7B
-            'qwen': 8192,         # Qwen default
-            'codellama': 16384,   # CodeLlama extended
-            'codegemma': 8192,    # CodeGemma default
-            'deepseek-coder-v2': 131072,  # DeepSeek-Coder V2 with 128K
-            'deepseek-coder': 16384,  # DeepSeek-Coder V1 default
-            'deepseek-v2': 131072,    # DeepSeek-V2 with 128K
-            'deepseek-llm': 4096,     # DeepSeek-LLM default
-            'yi1.5': 32768,       # Yi-1.5 with 32K
-            'yi': 4096,           # Yi base default
-            'command-r': 131072,  # Command-R with 128K
-            'wizardlm2': 32768,   # WizardLM2 (Mistral-based)
-            'wizardlm': 16384,    # WizardLM default
-            'zephyr': 65536,      # Zephyr beta (Mistral-based extended)
-            'vicuna': 2048,       # Vicuna default (up to 16K in some variants)
-            'falcon': 2048,       # Falcon default
-            'starcoder': 8192,    # StarCoder default
-            'stablelm': 4096,     # StableLM default
-            'orca2': 4096,        # Orca 2 default
-            'orca': 4096,         # Orca default
-            'dolphin': 32768,     # Dolphin (often Mistral-based)
-            'openhermes': 8192,   # OpenHermes default
-            'gpt-oss': 128000,  # GPT-OSS with 128K context
-            'gpt-3.5-turbo': 4096, # GPT-3.5 Turbo default
-            'gpt-4': 8192,        # GPT-4 default
-            'grok-2': 128000,
-            'grok-2-1212': 128000,
-            'grok-2-vision-1212': 128000,
-            'grok-3': 128000,
-            'grok-3-fast': 128000,
-            'grok-3-beta': 128000,
-            'grok-3-mini': 128000,
-            'grok-3-mini-beta': 128000,
-            'grok-3-mini-fast': 128000,
-            'grok-4-0709': 128000,
-            'grok-4': 128000,
-            'grok-vision-beta': 128000,
-        }
+        known_contexts = load_known_contexts()
 
         normalized_model_name = model_name.lower().strip()
 
@@ -708,7 +380,7 @@ class LollmsLLMBinding(ABC):
 class LollmsLLMBindingManager:
     """Manages binding discovery and instantiation"""
 
-    def __init__(self, llm_bindings_dir: str = "llm_bindings"):
+    def __init__(self, llm_bindings_dir: Union[str, Path] = Path(__file__).parent.parent / "llm_bindings"):
         """
         Initialize the LollmsLLMBindingManager.
 
@@ -750,16 +422,155 @@ class LollmsLLMBindingManager:
         if binding_class:
             return binding_class(**kwargs)
         return None
+    @staticmethod
+    def _get_fallback_description(binding_name: str) -> Dict:
+        """
+        Generates a default description dictionary for a binding without a description.yaml file.
+        """
+        return {
+            "binding_name": binding_name,
+            "title": binding_name.replace("_", " ").title(),
+            "author": "Unknown",
+            "creation_date": "N/A",
+            "last_update_date": "N/A",
+            "description": f"A binding for {binding_name}. No description.yaml file was found, so common parameters are shown as a fallback.",
+            "input_parameters": [
+                {
+                    "name": "model_name",
+                    "type": "str",
+                    "description": "The model name, ID, or filename to be used.",
+                    "mandatory": False,
+                    "default": ""
+                },
+                {
+                    "name": "host_address",
+                    "type": "str",
+                    "description": "The host address of the service (for API-based bindings).",
+                    "mandatory": False,
+                    "default": ""
+                },
+                {
+                    "name": "models_path",
+                    "type": "str",
+                    "description": "The path to the models directory (for local bindings).",
+                    "mandatory": False,
+                    "default": ""
+                },
+                {
+                    "name": "service_key",
+                    "type": "str",
+                    "description": "The API key or service key for authentication (if applicable).",
+                    "mandatory": False,
+                    "default": ""
+                }
+            ]
+        }
 
-    def get_available_bindings(self) -> list[str]:
+    @staticmethod
+    def get_bindings_list(llm_bindings_dir: Union[str, Path]) -> List[Dict]:
         """
-        Return list of available binding names.
+        Lists all available LLM bindings by scanning a directory, loading their
+        description.yaml file if present, or providing a default description.
+
+        Args:
+            llm_bindings_dir (Union[str, Path]): The path to the directory containing LLM binding folders.
 
         Returns:
-            list[str]: List of binding names.
+            List[Dict]: A list of dictionaries, each describing a binding.
+        """
+        bindings_dir = Path(llm_bindings_dir)
+        if not bindings_dir.is_dir():
+            return []
+
+        bindings_list = []
+        for binding_folder in bindings_dir.iterdir():
+            if binding_folder.is_dir() and (binding_folder / "__init__.py").exists():
+                binding_name = binding_folder.name
+                description_file = binding_folder / "description.yaml"
+                
+                binding_info = {}
+                if description_file.exists():
+                    try:
+                        with open(description_file, 'r', encoding='utf-8') as f:
+                            binding_info = yaml.safe_load(f)
+                        binding_info['binding_name'] = binding_name
+                    except Exception as e:
+                        print(f"Error loading description.yaml for {binding_name}: {e}")
+                        binding_info = LollmsLLMBindingManager._get_fallback_description(binding_name)
+                else:
+                    binding_info = LollmsLLMBindingManager._get_fallback_description(binding_name)
+                
+                bindings_list.append(binding_info)
+
+        return sorted(bindings_list, key=lambda b: b.get('title', b['binding_name']))
+    
+    def get_available_bindings(self) -> List[Dict]:
         """
-        return [binding_dir.name for binding_dir in self.llm_bindings_dir.iterdir() if binding_dir.is_dir() and (binding_dir / "__init__.py").exists()]
+        Retrieves a list of all available LLM bindings with their full descriptions.
+
+        This method scans the configured `llm_bindings_dir`, parsing the `description.yaml`
+        file for each valid binding. If a `description.yaml` is missing, a fallback
+        description with common parameters is generated. This is the primary method
+        for discovering available bindings and their configuration requirements.
 
-def get_available_bindings():
-    bindings_dir = Path(__file__).parent/"llm_bindings"
-    return [binding_dir.name for binding_dir in bindings_dir.iterdir() if binding_dir.is_dir() and (binding_dir / "__init__.py").exists()]
+        Returns:
+            List[Dict]: 
+                A list of dictionaries, where each dictionary represents the
+                full description of an available binding.
+
+                Each dictionary contains the following keys:
+                - ``binding_name`` (str): The programmatic name of the binding (its folder name).
+                - ``title`` (str): A user-friendly title for the binding.
+                - ``author`` (str): The creator of the binding.
+                - ``creation_date`` (str): The date the binding was created.
+                - ``last_update_date`` (str): The date of the last major update.
+                - ``description`` (str): A detailed explanation of the binding's purpose.
+                - ``input_parameters`` (List[Dict]): A list of parameters required to
+                  configure the binding. Each parameter is a dictionary with:
+                    - ``name`` (str): The parameter's name (e.g., 'model_name').
+                    - ``type`` (str): The expected data type ('str', 'int', 'float', 'bool').
+                    - ``description`` (str): A user-friendly description of the parameter.
+                    - ``mandatory`` (bool): True if the parameter must be provided.
+                    - ``default``: The default value for the parameter.
+
+        Example of a returned dictionary in the list:
+        .. code-block:: python
+
+            {
+                "binding_name": "ollama",
+                "title": "Ollama",
+                "author": "ParisNeo",
+                ...
+                "input_parameters": [
+                    {
+                        "name": "host_address",
+                        "type": "str",
+                        "description": "The URL of the Ollama server.",
+                        "mandatory": True,
+                        "default": "http://localhost:11434"
+                    },
+                    ...
+                ]
+            }
+        """
+        return LollmsLLMBindingManager.get_bindings_list(self.llm_bindings_dir)
+
+def get_available_bindings(llm_bindings_dir: Union[str, Path] = None) -> List[Dict]:
+    """
+    Lists all available LLM bindings with their detailed descriptions.
+
+    This function serves as a primary entry point for discovering what bindings
+    are available and how to configure them.
+
+    Args:
+        llm_bindings_dir (Union[str, Path], optional): 
+            The path to the LLM bindings directory. If None, it defaults to the
+            'llm_bindings' subdirectory relative to this file. 
+            Defaults to None.
+
+    Returns:
+        List[Dict]: A list of dictionaries, each describing a binding.
+    """
+    if llm_bindings_dir is None:
+        llm_bindings_dir = Path(__file__).parent / "llm_bindings"
+    return LollmsLLMBindingManager.get_bindings_list(llm_bindings_dir)

lollms_client/lollms_mcp_binding.py

@@ -4,7 +4,7 @@ import importlib
 from pathlib import Path
 from typing import Optional, List, Dict, Any, Union
 from ascii_colors import trace_exception, ASCIIColors
-
+import yaml
 class LollmsMCPBinding(ABC):
     """
     Abstract Base Class for LOLLMS Model Context Protocol (MCP) Bindings.
@@ -185,6 +185,7 @@ class LollmsMCPBindingManager:
                 return None
         return None
 
+
     def get_available_bindings(self) -> List[str]:
         """
         Return list of available MCP binding names based on subdirectories.
@@ -195,4 +196,34 @@ class LollmsMCPBindingManager:
             for item in self.mcp_bindings_dir.iterdir():
                 if item.is_dir() and (item / "__init__.py").exists():
                     available.append(item.name)
-        return available
+        return available
+
+    def get_binding_description(self, binding_name: str) -> Optional[Dict[str, Any]]:
+        """
+        Loads and returns the content of the description.yaml file for a given binding.
+
+        Args:
+            binding_name (str): The name of the binding.
+
+        Returns:
+            Optional[Dict[str, Any]]: A dictionary with the parsed YAML content, or None if the file doesn't exist or is invalid.
+        """
+        binding_dir = self.mcp_bindings_dir / binding_name
+        description_file = binding_dir / "description.yaml"
+
+        if not description_file.exists():
+            ASCIIColors.warning(f"No description.yaml found for MCP binding '{binding_name}'.")
+            return None
+
+        try:
+            with open(description_file, 'r', encoding='utf-8') as f:
+                description = yaml.safe_load(f)
+            return description
+        except yaml.YAMLError as e:
+            ASCIIColors.error(f"Error parsing description.yaml for MCP binding '{binding_name}': {e}")
+            trace_exception(e)
+            return None
+        except Exception as e:
+            ASCIIColors.error(f"Error reading description.yaml for MCP binding '{binding_name}': {e}")
+            trace_exception(e)
+            return None