Search

This service returns the contents from the database with the most similar vector embeddings to the question and optionally also uses an LLM to construct a custom answer to the user's question using the retrieved contents.

See OpenAPI specification or SwaggerUI for more details on how to call the service.

Process flow without LLM response generation

sequenceDiagram
  autonumber
  User->>AAQ: User's question
  AAQ->>LLM: Identify language
  LLM->>AAQ: <Language>
  AAQ->>LLM: Translate text
  LLM->>AAQ: <Translated text>
  AAQ->>LLM: Paraphrase question
  LLM->>AAQ: <Paraphrased question>
  AAQ->>Vector DB: Request M most similar contents in DB
  Vector DB->>AAQ: <M contents with similarity score>
  AAQ->>Cross-encoder: Re-rank to get top N contents
  Cross-encoder->>AAQ: <N contents with similarity score>
  AAQ->>User: Return JSON of N contents

Process flow with LLM response generation

sequenceDiagram
  autonumber
  User->>AAQ: User's question
  AAQ->>LLM: Identify language
  LLM->>AAQ: <Language>
  AAQ->>LLM: Check for safety
  LLM->>AAQ: <Safety Classification>
  AAQ->>Vector DB: Request N most similar contents in DB
  Vector DB->>AAQ: <N contents with similarity score>
  AAQ->>Cross-encoder: Re-rank to get top N contents
  Cross-encoder->>AAQ: <N contents with similarity score>
  AAQ->>LLM: Given contents, construct response in user's language to question
  LLM->>AAQ: <LLM response>
  AAQ->>LLM: Check if LLM response is consistent with contents
  LLM->>AAQ: <Consistency score>
  AAQ->>User: Return JSON of LLM response and N contents

Optional components

Align Score

In the Process Flow above, Step 12: Check if answer is consistent with content can be done using a custom AlignScore model instead of another LLM call.

sequenceDiagram
  autonumber
  User->>AAQ: User's question
  AAQ->>LLM: Identify language
  LLM->>AAQ: <Language>
  AAQ->>LLM: Check for safety
  LLM->>AAQ: <Safety Classification>
  AAQ->>Vector DB: Request N most similar contents in DB
  Vector DB->>AAQ: <N contents with similarity score>
  AAQ->>LLM: Construct response to question given contents
  LLM->>AAQ: <LLM response>
  AAQ->>Custom AlignScore Model: Check if LLM response is consistent with contents
  Custom AlignScore Model ->>AAQ: <Consistency score>
  AAQ->>User: Return JSON of LLM response and N contents

To use the custom AlignScore model, AAQ needs access to the AlignScore service. See documentation for how to setup the service and configure AAQ to call it.