gardel/lightrag
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Adjust concurrency limits more LLM friendly settings for new comers

Browse Source 
- Lowered max async LLM processes to 4
- Enabled LLM cache for entity extraction
- Reduced max parallel insert to 2
This commit is contained in:
yangdx
2025-03-16 23:56:34 +08:00
parent 9d971e5889
commit c2ba7f33ff
5 changed files with 7 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@@ -1061,7 +1061,7 @@ Valid modes are:
| **llm\_model\_func** | `callable` | Function for LLM generation | `gpt_4o_mini_complete` |
| **llm\_model\_name** | `str` | LLM model name for generation | `meta-llama/Llama-3.2-1B-Instruct` |
| **llm\_model\_max\_token\_size** | `int` | Maximum token size for LLM generation (affects entity relation summaries) | `32768`（default value changed by env var MAX_TOKENS) |
| **llm\_model\_max\_async** | `int` | Maximum number of concurrent asynchronous LLM processes | `16`（default value changed by env var MAX_ASYNC) |
| **llm\_model\_max\_async** | `int` | Maximum number of concurrent asynchronous LLM processes | `4`（default value changed by env var MAX_ASYNC) |
| **llm\_model\_kwargs** | `dict` | Additional parameters for LLM generation | |
| **vector\_db\_storage\_cls\_kwargs** | `dict` | Additional parameters for vector database, like setting the threshold for nodes and relations retrieval. | cosine_better_than_threshold: 0.2（default value changed by env var COSINE_THRESHOLD) |
| **enable\_llm\_cache** | `bool` | If `TRUE`, stores LLM results in cache; repeated prompts return cached responses | `TRUE` |

3
env.example
View File

@@ -50,7 +50,8 @@
# MAX_TOKEN_SUMMARY=500 # Max tokens for entity or relations summary
# SUMMARY_LANGUAGE=English
# MAX_EMBED_TOKENS=8192
# ENABLE_LLM_CACHE_FOR_EXTRACT=false # Enable LLM cache for entity extraction, defaults to false
# ENABLE_LLM_CACHE_FOR_EXTRACT=true # Enable LLM cache for entity extraction
# MAX_PARALLEL_INSERT=2 # Maximum number of parallel processing documents in pipeline
### LLM Configuration (Use valid host. For local services installed with docker, you can use host.docker.internal)
LLM_BINDING=ollama

2
lightrag/api/README.md
View File

@@ -224,7 +224,7 @@ LightRAG supports binding to various LLM/Embedding backends:
Use environment variables `LLM_BINDING` or CLI argument `--llm-binding` to select LLM backend type. Use environment variables `EMBEDDING_BINDING` or CLI argument `--embedding-binding` to select LLM backend type.
### Entity Extraction Configuration
* ENABLE_LLM_CACHE_FOR_EXTRACT: Enable LLM cache for entity extraction (default: false)
* ENABLE_LLM_CACHE_FOR_EXTRACT: Enable LLM cache for entity extraction (default: true)
It's very common to set `ENABLE_LLM_CACHE_FOR_EXTRACT` to true for test environment to reduce the cost of LLM calls.

2
lightrag/api/utils_api.py
View File

@@ -364,7 +364,7 @@ def parse_args(is_uvicorn_mode: bool = False) -> argparse.Namespace:
# Inject LLM cache configuration
args.enable_llm_cache_for_extract = get_env_value(
"ENABLE_LLM_CACHE_FOR_EXTRACT", False, bool
"ENABLE_LLM_CACHE_FOR_EXTRACT", True, bool
)
# Select Document loading tool (DOCLING, DEFAULT)

4
lightrag/lightrag.py
View File

@@ -214,7 +214,7 @@ class LightRAG:
llm_model_max_token_size: int = field(default=int(os.getenv("MAX_TOKENS", 32768)))
"""Maximum number of tokens allowed per LLM response."""
llm_model_max_async: int = field(default=int(os.getenv("MAX_ASYNC", 16)))
llm_model_max_async: int = field(default=int(os.getenv("MAX_ASYNC", 4)))
"""Maximum number of concurrent LLM calls."""
llm_model_kwargs: dict[str, Any] = field(default_factory=dict)
@@ -238,7 +238,7 @@ class LightRAG:
# Extensions
# ---
max_parallel_insert: int = field(default=int(os.getenv("MAX_PARALLEL_INSERT", 20)))
max_parallel_insert: int = field(default=int(os.getenv("MAX_PARALLEL_INSERT", 2)))
"""Maximum number of parallel insert operations."""
addon_params: dict[str, Any] = field(