PowerOn/gateway
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

neues text model

Browse source
This commit is contained in:
patrick-motsch 2026-02-06 13:34:50 +01:00
parent bb10a46cd5
commit 8dfb7caf92
1 changed files with 13 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

26
modules/aicore/aicorePluginPrivateLlm.py
View file

@ -217,35 +217,35 @@ class AiPrivateLlm(BaseConnectorAi):
# Define all models with their Ollama backend names # Define all models with their Ollama backend names
# Actual model specs (for 32GB RAM server): # Actual model specs (for 32GB RAM server):
# - deepseek-ocr: 3.34B params, 8K context, ~6.7GB RAM # - qwen2.5:7b: 7.6B params, 128K context, ~4.7GB RAM (Text)
# - qwen2.5vl:7b: 8.29B params, 125K context, ~6GB RAM # - qwen2.5vl:7b: 8.29B params, 125K context, ~6GB RAM (Vision)
# - granite3.2-vision: 2B params, 16K context, ~2.4GB RAM # - granite3.2-vision: 2B params, 16K context, ~2.4GB RAM (Vision)
modelDefinitions = [ modelDefinitions = [
# OCR Text Model (deepseek-ocr: 3.34B, 8K context) # Text Model (qwen2.5:7b: 7.6B, 128K context)
{ {
"model": AiModel( "model": AiModel(
name="poweron-ocr-general", name="poweron-text-general",
displayName="PowerOn OCR General", displayName="PowerOn Text General",
connectorType="privatellm", connectorType="privatellm",
apiUrl=f"{self.baseUrl}/api/analyze", apiUrl=f"{self.baseUrl}/api/analyze",
temperature=0.1, temperature=0.1,
maxTokens=4096, maxTokens=8192,
contextLength=8192, # deepseek-ocr actual context: 8K contextLength=128000, # qwen2.5:7b actual context: 128K
costPer1kTokensInput=0.0, # Flat rate pricing costPer1kTokensInput=0.0, # Flat rate pricing
costPer1kTokensOutput=0.0, # Flat rate pricing costPer1kTokensOutput=0.0, # Flat rate pricing
speedRating=8, # Fast due to smaller model speedRating=8, # Fast and efficient
qualityRating=8, qualityRating=9, # High quality text model
functionCall=self.callAiText, functionCall=self.callAiText,
priority=PriorityEnum.COST, priority=PriorityEnum.COST,
processingMode=ProcessingModeEnum.BASIC, processingMode=ProcessingModeEnum.BASIC,
operationTypes=createOperationTypeRatings( operationTypes=createOperationTypeRatings(
(OperationTypeEnum.DATA_EXTRACT, 9), (OperationTypeEnum.DATA_EXTRACT, 9),
(OperationTypeEnum.DATA_ANALYSE, 7), (OperationTypeEnum.DATA_ANALYSE, 9),
), ),
version="deepseek-ocr", version="qwen2.5:7b",
calculatepriceCHF=lambda processingTime, bytesSent, bytesReceived: PRICE_TEXT_PER_CALL calculatepriceCHF=lambda processingTime, bytesSent, bytesReceived: PRICE_TEXT_PER_CALL
), ),
"ollamaModel": "deepseek-ocr" "ollamaModel": "qwen2.5:7b"
}, },
# Vision General Model (qwen2.5vl:7b: 8.29B, 125K context) # Vision General Model (qwen2.5vl:7b: 8.29B, 125K context)
{ {