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

Exclude conversation history from context length calculation

Browse source
This commit is contained in:
yangdx 2025-08-25 12:43:34 +08:00
parent f688e95f56
commit 31f4f96944
1 changed files with 9 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
lightrag/operate.py
View file

@ -2499,15 +2499,15 @@ async def _build_query_context(
kg_context_tokens = len(tokenizer.encode(kg_context)) kg_context_tokens = len(tokenizer.encode(kg_context))
# Calculate actual system prompt overhead dynamically # Calculate actual system prompt overhead dynamically
# 1. Calculate conversation history tokens # 1. Converstion history not included in context length calculation
history_context = "" history_context = ""
if query_param.conversation_history: # if query_param.conversation_history:
history_context = get_conversation_turns( # history_context = get_conversation_turns(
query_param.conversation_history, query_param.history_turns # query_param.conversation_history, query_param.history_turns
) # )
history_tokens = ( # history_tokens = (
len(tokenizer.encode(history_context)) if history_context else 0 # len(tokenizer.encode(history_context)) if history_context else 0
) # )
# 2. Calculate system prompt template tokens (excluding context_data) # 2. Calculate system prompt template tokens (excluding context_data)
user_prompt = query_param.user_prompt if query_param.user_prompt else "" user_prompt = query_param.user_prompt if query_param.user_prompt else ""
@ -2542,7 +2542,7 @@ async def _build_query_context(
available_chunk_tokens = max_total_tokens - used_tokens available_chunk_tokens = max_total_tokens - used_tokens
logger.debug( logger.debug(
f"Token allocation - Total: {max_total_tokens}, History: {history_tokens}, SysPrompt: {sys_prompt_overhead}, KG: {kg_context_tokens}, Buffer: {buffer_tokens}, Available for chunks: {available_chunk_tokens}" f"Token allocation - Total: {max_total_tokens}, SysPrompt: {sys_prompt_overhead}, KG: {kg_context_tokens}, Buffer: {buffer_tokens}, Available for chunks: {available_chunk_tokens}"
) )
# Apply token truncation to chunks using the dynamic limit # Apply token truncation to chunks using the dynamic limit