
Name	GradientGenerator
Source	https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/gradient
Most common position in a pipeline	After a `PromptBuilder`
Mandatory input variables	“prompt”: A string containing the prompt for the LLM
Output variables	“replies”: A list of strings with all the replies generated by the LLM

GradientGenerator enables text generation using generative models hosted by Gradient AI. You can either use one of the base models provided through the platform, or models that you’ve fine-tuned and are hosting through the platform.

You can currently use the following models, with more becoming available soon. Check out the Gradient documentation for the full list.

bloom-560m
llama2-7b-chat
nous-hermes2

📘
For an example showcasing this component, check out this article and the related 🧑‍🍳 Cookbook.

Parameters Overview

GradientGenerator needs an access_token and workspace_id. It also needs either a base_model_slug or model_adapter_id. You can provide these in one of the following ways:

For the access token and workspace_id, do one of the following:

Provide the access_token and workspace_id init parameter.
Set GRADIENT_ACCESS_TOKEN and GRADIENT_WORKSPACE_ID environment variables.

For the model you would like to use, do one of the following :

Provide the base_model_slug. Check the available base models on the Gradient documentation
If you’ve deployed a model (fine-tuned or not) on Gradient, provide the model_adapter_id for that model.

Usage

You need to install gradient-haystack package to use the GradientGenerator:

pip install gradient-haystack

On its own

Basic usage (with a base model). You can replace the base_model_slug with a model_adapter_id to use your own deployed models in your Gradient workspace:

import os
from haystack_integrations.components.generators.gradient import GradientGenerator

os.environ["GRADIENT_ACCESS_TOKEN"]="YOUR_GRADIENT_ACCESS_TOKEN"
os.environ["GRADIENT_WORKSPACE_ID"]="GRADIENT_WORKSPACE_ID"

generator = GradientGenerator(base_model="llama2-7b-chat",
                              max_generated_token_count=350)
generator.warm_up()
generator.run(prompt="What is the meaning of life?")

In a pipeline

Here’s an example of this generator in a RAG Pipeline. In this Pipeline, we are using the GradientTextEmbedder and the GradientDocumentEmbedder as well. You can replace these with any other embedder. It assumes that you have an InMemoryDocumentStore that has Documents in it:

import os
from haystack import Pipeline
from haystack.components.retrievers.in_memory import InMemoryEmbeddingRetriever
from haystack.components.builders import PromptBuilder
from haystack_integrations.components.embedders.gradient import GradientTextEmbedder
from haystack_integrations.components.generators.gradient import GradientGenerator
from haystack.document_stores.in_memory import InMemoryDocumentStore

os.environ["GRADIENT_ACCESS_TOKEN"]="YOUR_GRADIENT_ACCESS_TOKEN"
os.environ["GRADIENT_WORKSPACE_ID"]="GRADIENT_WORKSPACE_ID"

document_store = InMemoryDocumentStore()

prompt = """ Answer the query, based on the
content in the documents.

Documents:
{% for doc in documents %}
  {{doc.content}}
{% endfor %}

Query: {{query}}
"""
text_embedder = GradientTextEmbedder()
retriever = InMemoryEmbeddingRetriever(document_store=document_store)
prompt_builder = PromptBuilder(template=prompt)
generator = GradientGenerator(model_adapter_id="YOUR_MODEL_ADAPTER_ID",
                              max_generated_token_count=350)
rag_pipeline = Pipeline()

rag_pipeline.add_component(instance=text_embedder, name="text_embedder")
rag_pipeline.add_component(instance=retriever, name="retriever")
rag_pipeline.add_component(instance=prompt_builder, name="prompt_builder")
rag_pipeline.add_component(instance=generator, name="generator")

rag_pipeline.connect("text_embedder", "retriever")
rag_pipeline.connect("retriever.documents", "prompt_builder.documents")
rag_pipeline.connect("prompt_builder", "generator")

question = "What are the steps for creating a custom component?"
result = rag_pipeline.run(data={"text_embedder":{"text": question},
                                "prompt_builder":{"query": question}})