Using lingoose with RAG produces 2 system messages in the Thread (+sometimes returns random output/answer, depending on the chosen RAG/DB/Model, but that might be some other issue).
Additionally the .withModel() method on the embedder does not seem to have any effect or returns empty.
Link to Issue: henomis/lingoose#208
go test -v ./...
This will output a thread with 2 system messages (one from lingoose/assistant/prompt.go and one from the user defined system message).
TestOllamaAssistantWithRag
ollama_test.go:34: ragQueryData: [this is a little side story in paris about a little mermaid]
ollama_test.go:47: Thread:
system:
Type: text
Text: You are a helpful AI Assistant. Count the words and always return the answer as a number without verbose information. Answer 0 if you are unsure.
system:
Type: text
Text: You name is AI assistant, and you are a helpful and polite assistant . Your task is to assist humans with their questions.
user:
Type: text
Text: Use the following pieces of retrieved context to answer the question.
Question: It's about a story of a little mermaid.
Context:
this is a little side story in paris about a little mermaid
assistant:
Type: text
Text:
ollama_test.go:50:
Answer needs to be a number!