Search Shortcut cmd + k | ctrl + k
flockmtl

LLM & RAG extension to combine analytics and semantic analysis

Maintainer(s): anasdorbani, queryproc

Installing and Loading

INSTALL flockmtl FROM community;
LOAD flockmtl;

Example

-- After loading, any function call will throw an error if the provider's secret doesn't exist

-- Create your provider secret by following the [documentation](https://dais-polymtl.github.io/flockmtl/docs/what-is-flockmtl/). For example, you can create a default OpenAI API key as follows:
D CREATE SECRET (TYPE OPENAI, API_KEY 'your-api-key');

-- Call an OpenAI model with a predefined prompt ('Tell me hello world') and default model ('gpt-4o-mini')
D SELECT llm_complete({'model_name': 'default'}, {'prompt_name': 'hello-world'});
┌──────────────────────────────────────────┐
 llm_complete(hello_world, default_model) 
                 varchar                  
├──────────────────────────────────────────┤
                Hello world               
└──────────────────────────────────────────┘

-- Check the prompts and supported models
D GET PROMPTS;
D GET MODELS;

-- Create a new prompt for summarizing text
D CREATE PROMPT('summarize', 'summarize the text into 1 word: {{text}}');

-- Create a variable name for the model to do the summarizing
D CREATE MODEL('summarizer-model', 'gpt-4o', {'context_window': 128000, 'max_output_tokens': 16400});

-- Summarize text and pass it as parameter 
D SELECT llm_complete({'model_name': 'summarizer-model'}, {'prompt_name': 'summarize'}, {'text': 'We support more functions and approaches to combine relational analytics and semantic analysis. Check our repo for documentation and examples.'});

About flockmtl

FlockMTL is an experimental DuckDB extension that enables seamless integration of large language models (LLMs) and retrieval-augmented generation (RAG) directly within SQL.

It introduces MODEL and PROMPT objects as first-class SQL entities, making it easy to define, manage, and reuse LLM interactions. Core functions like llm_complete, llm_filter, and llm_rerank allow you to perform generation, semantic filtering, and ranking—all from SQL.

FlockMTL is designed for rapid prototyping of LLM-based analytics and is optimized with batching and caching features for better performance.

📄 For more details and examples, see the FlockMTL documentation.

Note: FlockMTL is part of ongoing research by the Data & AI Systems (DAIS) Laboratory @ Polytechnique Montréal. It is under active development, and some features may evolve. Feedback and contributions are welcome!

Added Functions

function_name function_type description comment examples
fusion_combanz scalar NULL NULL  
fusion_combmed scalar NULL NULL  
fusion_combmnz scalar NULL NULL  
fusion_combsum scalar NULL NULL  
fusion_rrf scalar NULL NULL  
llm_complete scalar NULL NULL  
llm_complete_json scalar NULL NULL  
llm_embedding scalar NULL NULL  
llm_filter scalar NULL NULL  
llm_first aggregate NULL NULL  
llm_last aggregate NULL NULL  
llm_reduce aggregate NULL NULL  
llm_reduce_json aggregate NULL NULL  
llm_rerank aggregate NULL NULL