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

Refactor embedding function initialization and remove start-server.sh

Browse Source 
- Simplified RAG initialization logic by deduplicating embedding function
- Removed start-server.sh script which is not needed
- No functional changes to the application
This commit is contained in:
yangdx
2025-01-19 05:19:02 +08:00
parent 8ea179a98b
commit 387be31f09
4 changed files with 28 additions and 978 deletions
Download Patch File Download Diff File Expand all files Collapse all files

78
lightrag/api/lightrag_server.py
View File

@@ -615,6 +615,32 @@ def create_app(args):
**kwargs,
)
embedding_func = EmbeddingFunc(
embedding_dim=args.embedding_dim,
max_token_size=args.max_embed_tokens,
func=lambda texts: lollms_embed(
texts,
embed_model=args.embedding_model,
host=args.embedding_binding_host,
)
if args.embedding_binding == "lollms"
else ollama_embed(
texts,
embed_model=args.embedding_model,
host=args.embedding_binding_host,
)
if args.embedding_binding == "ollama"
else azure_openai_embedding(
texts,
model=args.embedding_model, # no host is used for openai
)
if args.embedding_binding == "azure_openai"
else openai_embedding(
texts,
model=args.embedding_model, # no host is used for openai
),
)
# Initialize RAG
if args.llm_binding in ["lollms", "ollama"] :
rag = LightRAG(
@@ -630,31 +656,7 @@ def create_app(args):
"timeout": args.timeout,
"options": {"num_ctx": args.max_tokens},
},
embedding_func=EmbeddingFunc(
embedding_dim=args.embedding_dim,
max_token_size=args.max_embed_tokens,
func=lambda texts: lollms_embed(
texts,
embed_model=args.embedding_model,
host=args.embedding_binding_host,
)
if args.embedding_binding == "lollms"
else ollama_embed(
texts,
embed_model=args.embedding_model,
host=args.embedding_binding_host,
)
if args.embedding_binding == "ollama"
else azure_openai_embedding(
texts,
model=args.embedding_model, # no host is used for openai
)
if args.embedding_binding == "azure_openai"
else openai_embedding(
texts,
model=args.embedding_model, # no host is used for openai
),
),
embedding_func=embedding_func,
)
else :
rag = LightRAG(
@@ -662,31 +664,7 @@ def create_app(args):
llm_model_func=azure_openai_model_complete
if args.llm_binding == "azure_openai"
else openai_alike_model_complete,
embedding_func=EmbeddingFunc(
embedding_dim=args.embedding_dim,
max_token_size=args.max_embed_tokens,
func=lambda texts: lollms_embed(
texts,
embed_model=args.embedding_model,
host=args.embedding_binding_host,
)
if args.embedding_binding == "lollms"
else ollama_embed(
texts,
embed_model=args.embedding_model,
host=args.embedding_binding_host,
)
if args.embedding_binding == "ollama"
else azure_openai_embedding(
texts,
model=args.embedding_model, # no host is used for openai
)
if args.embedding_binding == "azure_openai"
else openai_embedding(
texts,
model=args.embedding_model, # no host is used for openai
),
),
embedding_func=embedding_func,
)
async def index_file(file_path: Union[str, Path]) -> None: