From 432ca8b7af8b8d3a7ec30151ca38308470473292 Mon Sep 17 00:00:00 2001 From: "bryan.jia" Date: Fri, 28 Feb 2025 16:00:20 +0800 Subject: [PATCH] fix: ollama model provider can not set 'num_ctx' etc. parameter #2442 langchain-openai is not compatible with parameter Settings in ollama, such as num_ctx. Therefore, you need to create model instances using langchain-ollama (cherry picked from commit 42ae7b443d7320ace464e9a12abcbb48d7fabc9c) --- .../impl/ollama_model_provider/model/llm.py | 11 +++++------ pyproject.toml | 1 + 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/apps/setting/models_provider/impl/ollama_model_provider/model/llm.py b/apps/setting/models_provider/impl/ollama_model_provider/model/llm.py index 7c98f7e5cef..6cd291ff3cc 100644 --- a/apps/setting/models_provider/impl/ollama_model_provider/model/llm.py +++ b/apps/setting/models_provider/impl/ollama_model_provider/model/llm.py @@ -10,7 +10,8 @@ from urllib.parse import urlparse, ParseResult from langchain_core.messages import BaseMessage, get_buffer_string -from langchain_openai.chat_models import ChatOpenAI +from langchain_ollama.chat_models import ChatOllama + from common.config.tokenizer_manage_config import TokenizerManage from setting.models_provider.base_model_provider import MaxKBBaseModel @@ -24,7 +25,7 @@ def get_base_url(url: str): return result_url[:-1] if result_url.endswith("/") else result_url -class OllamaChatModel(MaxKBBaseModel, ChatOpenAI): +class OllamaChatModel(MaxKBBaseModel, ChatOllama): @staticmethod def is_cache_model(): return False @@ -33,12 +34,10 @@ def is_cache_model(): def new_instance(model_type, model_name, model_credential: Dict[str, object], **model_kwargs): api_base = model_credential.get('api_base', '') base_url = get_base_url(api_base) - base_url = base_url if base_url.endswith('/v1') else (base_url + '/v1') optional_params = MaxKBBaseModel.filter_optional_params(model_kwargs) - return OllamaChatModel(model=model_name, openai_api_base=base_url, - openai_api_key=model_credential.get('api_key'), - stream_usage=True, **optional_params) + return OllamaChatModel(model=model_name, base_url=base_url, + stream=True, **optional_params) def get_num_tokens_from_messages(self, messages: List[BaseMessage]) -> int: tokenizer = TokenizerManage.get_tokenizer() diff --git a/pyproject.toml b/pyproject.toml index a5d175b855b..8134688e758 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -28,6 +28,7 @@ pycryptodome = "^3.19.0" beautifulsoup4 = "^4.12.2" html2text = "^2024.2.26" langchain-openai = "^0.1.8" +langchain-ollama = "0.1.3" django-ipware = "^6.0.4" django-apscheduler = "^0.6.2" pymupdf = "1.24.9"