gmakstutis/LightRAG
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update Ollama context length configuration

Browse source 
- Rename OLLAMA_NUM_CTX to OLLAMA_LLM_NUM_CTX
- Increase default context window size
- Add requirement for minimum context size
- Update documentation examples
This commit is contained in:
yangdx 2025-07-29 09:53:37 +08:00
parent 645f81f7c8
commit 75d1b1e9f8
4 changed files with 9 additions and 8 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
env.example
View file

@ -118,8 +118,9 @@ LLM_MODEL=gpt-4o
LLM_BINDING_HOST=https://api.openai.com/v1 LLM_BINDING_HOST=https://api.openai.com/v1
LLM_BINDING_API_KEY=your_api_key LLM_BINDING_API_KEY=your_api_key
### Set as num_ctx option for Ollama LLM ### Set as num_ctx option for Ollama LLM (Must be larger than MAX_TOTAL_TOKENS+2000)
# OLLAMA_NUM_CTX=32768 ### see also env.ollama-binding-options.example for fine tuning ollama
# OLLAMA_LLM_NUM_CTX=32768
### Optional for Azure ### Optional for Azure
# AZURE_OPENAI_API_VERSION=2024-08-01-preview # AZURE_OPENAI_API_VERSION=2024-08-01-preview

4
lightrag/api/README-zh.md
View file

@ -69,8 +69,8 @@ LLM_BINDING=ollama
LLM_MODEL=mistral-nemo:latest LLM_MODEL=mistral-nemo:latest
LLM_BINDING_HOST=http://localhost:11434 LLM_BINDING_HOST=http://localhost:11434
# LLM_BINDING_API_KEY=your_api_key # LLM_BINDING_API_KEY=your_api_key
### Ollama 服务器上下文 token 数（基于您的 Ollama 服务器容量） ### Ollama 服务器上下文 token 数（必须大于 MAX_TOTAL_TOKENS+2000）
OLLAMA_NUM_CTX=8192 OLLAMA_LLM_NUM_CTX=8192
EMBEDDING_BINDING=ollama EMBEDDING_BINDING=ollama
EMBEDDING_BINDING_HOST=http://localhost:11434 EMBEDDING_BINDING_HOST=http://localhost:11434

4
lightrag/api/README.md
View file

@ -69,8 +69,8 @@ LLM_BINDING=ollama
LLM_MODEL=mistral-nemo:latest LLM_MODEL=mistral-nemo:latest
LLM_BINDING_HOST=http://localhost:11434 LLM_BINDING_HOST=http://localhost:11434
# LLM_BINDING_API_KEY=your_api_key # LLM_BINDING_API_KEY=your_api_key
### Ollama Server context length ### Ollama Server context length (Must be larger than MAX_TOTAL_TOKENS+2000)
OLLAMA_NUM_CTX=8192 OLLAMA_LLM_NUM_CTX=16384
EMBEDDING_BINDING=ollama EMBEDDING_BINDING=ollama
EMBEDDING_BINDING_HOST=http://localhost:11434 EMBEDDING_BINDING_HOST=http://localhost:11434

4
lightrag/llm/binding_options.py
View file

@ -240,7 +240,7 @@ class _OllamaOptionsMixin:
"""Options for Ollama bindings.""" """Options for Ollama bindings."""
# Core context and generation parameters # Core context and generation parameters
num_ctx: int = 4096 # Context window size (number of tokens) num_ctx: int = 32768 # Context window size (number of tokens)
num_predict: int = 128 # Maximum number of tokens to predict num_predict: int = 128 # Maximum number of tokens to predict
num_keep: int = 0 # Number of tokens to keep from the initial prompt num_keep: int = 0 # Number of tokens to keep from the initial prompt
seed: int = -1 # Random seed for generation (-1 for random) seed: int = -1 # Random seed for generation (-1 for random)
@ -438,7 +438,7 @@ if __name__ == "__main__":
# test LLM options # test LLM options
ollama_options = OllamaLLMOptions.options_dict(args) ollama_options = OllamaLLMOptions.options_dict(args)
print(ollama_options) print(ollama_options)
print(OllamaLLMOptions(num_ctx=32768).asdict()) print(OllamaLLMOptions(num_ctx=30000).asdict())
# test embedding options # test embedding options
embedding_options = OllamaEmbeddingOptions.options_dict(args) embedding_options = OllamaEmbeddingOptions.options_dict(args)