Build with E2E Cloud: Step-by-Step Guide to Build Advanced Multistage AI Agents Using DSPy

June 24, 2024

Benefits of Using DSPy for Building Multistage Agents

DSPy is a handy framework that reduces the effort involved in prompt engineering for building AI applications. It achieves this by providing modules to structure the prompts and optimizers that tweak the prompts based on example data.

When building multistage LLM-based agents, DSPy makes it extremely convenient to design the workflow. Multistage agents typically require extensive prompt-tweaking to work well, and without DSPy, this can become complicated and messy. Each step requires manually engineering the prompts until the desired results are achieved, and then ensuring that these steps work together.

DSPy’s modules and optimizers automatically handle the prompting steps. Moreover, this approach is LLM-agnostic, meaning that if the LLM changes, you don’t have to rewrite your prompts to fit the new LLM’s template. Instead, DSPy will automatically adjust for you.

This is extremely useful because it allows you to use any Large Language Model you prefer without worrying about re-engineering the prompt templates.

In this blog, we’ll design a multistage Personal Styling Agent that gives users styling advice based on their preferences.

Personalized Styling Agent - Workflow

We’ll break our agent into two modules. The first module will generate a list of clothing items and accessories based on user preferences.

We’ll also store a list of products separately in the Qdrant Vector Database. The second module will fetch products from this vector database that are similar to the items generated by the first module. It will then generate styling advice based on the fetched list of products.

This way, users receive advice based on the available list of products. This feature can be useful for e-commerce companies that are looking to provide product recommendations to users based on their preferences.

Build with E2E Cloud

This blog is a part of the ‘Build with E2E Cloud’ series. E2E Networks is an NSE-listed, AI-first hyperscaler that provides accelerated cloud computing solutions, including cutting-edge cloud GPUs like the A100 and H100. 

E2E Cloud also provides a comprehensive and robust AI development platform called TIR for various machine learning and artificial intelligence operations. The TIR platform provides a suite of tools to develop, deploy, and manage AI models effectively.

Head over to to know more about E2E’s services!

Let’s Code

Since we are going to be using Llama3 for this project, we recommend that you use the A100 or H100 node on E2E cloud. 

Begin by installing the dependencies:


Initialize the vector database. You can also use the Qdrant instance offered by TIR. 

from qdrant_client import QdrantClient
# Initialize the client
client = QdrantClient(":memory:")

We’ll use a fashion product dataset from Kaggle. More specifically, we’ll be vectorizing one column from this dataset. This column contains the names of different products. This is how it looks like:

import pandas as pd

# Read the CSV file
df = pd.read_csv('styles.csv', on_bad_lines='skip')
# Extract the 'productDisplayName' column into a list
product_display_names = df['productDisplayName'].astype(str).tolist()

We’ll now insert the vector data in batches of 1000.

# Function to split list into batches of specified size
def batch_list(lst, batch_size):
    for i in range(0, len(lst), batch_size):
        yield lst[i:i + batch_size]

# Define batch size
batch_size = 1000

# Process the batches
for idx, batch in enumerate(batch_list(product_display_names, batch_size)):
    start_id = idx * batch_size + 1
    end_id = start_id + len(batch) - 1
    doc_ids = list(range(start_id, end_id + 1))

Now we’ll configure DSPy to include the Qdrant Retriever and the LLM Llama3.

from dspy.retrieve.qdrant_rm import QdrantRM
import dspy

qdrant_retriever_model = QdrantRM("product_list", client, k=3)

ollama_model = dspy.OllamaLocal(model="llama3",model_type='text',
                                top_p=0.8, frequency_penalty=1.17, top_k=40)

dspy.settings.configure(lm= ollama_model, rm=qdrant_retriever_model)

Make sure you have Ollama installed on your system. You can do so by following the instructions here

curl -fsSL | sh

You can start the server by running:

ollama serve

Now pull llama3.

ollama pull llama3

Alternatively, you can deploy Llama3 endpoint on TIR, and use that as well. 

Now we’ll write a signature for our first module to generate a list of clothing items and accessories.

class Items(dspy.Signature):
    Extract clothing items and accessories given user preference
    prompt = dspy.InputField()
    terms = dspy.OutputField(format=list)

In DSPy, prompts are structured as signatures. The doc string contains the instruction of the prompt. The variable names (‘prompt’ and ‘terms’) are important as they will be passed onto the LLM. So make sure your variable names are meaningful.

The call to the LLM happens in the modules. We’ll use a module called dspy.Predict which simply predicts the outcome given a prompt. You can check out the list of DSPy modules here

class FindItems(dspy.Module):

    def __init__(self):
        self.item_extractor = dspy.Predict(Items)

    def forward(self, question):
        instruction = (
            "Return a list of clothing items and accessories to be worn by a user, given their preferences"
            f"Preferences: {question}"
        prediction = self.item_extractor(
        return prediction.terms

Now that we have this module written, let’s optimize it. We’ll first create a trainset which contains a sample of how we want our outputs to look like.

trainset_dict= {
    "desert_wear": "* Lightweight, breathable long-sleeve shirt\n* Wide-brimmed hat\n* Sunglasses with UV protection\n* Cargo pants\n* Lightweight scarf or shemagh\n* Hiking boots with good ventilation\n* Moisture-wicking socks\n* Sunscreen\n* Hydration pack or water bottle\n* Lightweight jacket for cooler evenings",
    "caribbean_party": "* Flowy sundress with tropical print\n* Beach hat with colorful feathers\n* Statement necklace with seashells and beads\n* Flip flops in bright colors\n* Sunglasses with brightly colored frames\n* Lightweight, short-sleeve shirt with floral patterns\n* Comfortable shorts\n* Colorful bangles or bracelets\n* Straw tote bag\n* Lightweight scarf or wrap",
    "award_function": "* Elegant evening gown or tuxedo\n* Statement jewelry (earrings, necklace, bracelet)\n* Dress shoes or heels\n* Clutch or small evening bag\n* Cufflinks and tie (for tuxedo)\n* Evening makeup and hairstyle\n* Silk scarf or shawl\n* Formal watch\n* Perfume or cologne\n* Pocket square (for tuxedo)"

The keys of this dictionary represent the user preferences and the values represent the desired output by the LLM. 

Let’s wrap it around in a format that is suitable to work with DSPy.

trainset = [dspy.Example(question=key, answer=value) for key, value in trainset_dict.items()]

Then we optimize it.

from dspy import teleprompt
teleprompter = teleprompt.LabeledFewShot()
optimized_Find_Items = teleprompter.compile(student=FindItems(), trainset=trainset)

Let’s put our optimized module to test.

response = optimized_Find_Items("party wear")

Here is the extracted list of clothing items and accessories:

**Party Wear**

* Sequined dress

* High heels (black or red)

* Statement necklace with rhinestones

* Earrings with dangling crystals

* Sparkly clutch purse (silver or gold)

* Cocktail ring with gemstones

As we can see, we are getting our output in the format we’d asked for. Now, we’ll write a parsing function that collects the names of these items into a list.

def parse_clothing_items(text):
    # Split the text into lines
    lines = text.split('\n')

    # Initialize an empty list to hold the items
    items = []

    # Loop through each line
    for line in lines:
        # Check if the line starts with '*'
        if line.strip().startswith('* '):
            # Remove the '*' and any leading/trailing whitespace, then add to the list
            item = line.strip()[2:].strip()

    return items

Next we write the final module, which receives the list of items from the first module, does a vector search to find a list of similar products, and then generates styling advice based on the list. 

class Style(dspy.Signature):
    Generate styling advice, given these products
    products = dspy.InputField()
    answer = dspy.OutputField()

class Styling_Advise(dspy.Module):
    def __init__(self):
        self.generate_response = dspy.Predict(Style)
        self.retrieve = dspy.Retrieve(k=2)

    def forward(self, question):
        item_list= parse_clothing_items(optimized_Find_Items(question))
        product_list = []
        for item in item_list:
            product_list += self.retrieve(item).passages

        response = self.generate_response(products= ', '.join(product_list))
        return response.answer

This final module serves as a complete agent - with just one call, it does the entire job of coming up with a list of items, retrieving similar products from the vector DB, and finally generating the styling advice.

Let’s test it out.

module = Styling_Advise()

response = module("Corporate Award Ceremony")


Products: Aneri Women Black & Maroon Salwar Suit, Aneri Women Blue & White Salwar Suit, United Colors of Benetton Women White Shirt, Park Avenue Red Patterned Tie, Park Avenue Blue Patterned Tie, Basics Men Black Blazer Jacket, Basics Men Beige Blazer Jacket, Black Coffee Brown Formal Trousers, Basics Men Navy Trousers, ADIDAS Men Black Socks, Parx Men Black Socks, Woodland Men Brown Leather Shoes, Clarks Brown Leather Casual Shoes, Being Human Men Black Dial Orange Strap Watch, Being Human Men Black Dial Steel Plated Strap Watch, Revv Men Black and White Cufflinks, Belmonte Men Silver Cufflinks


For the women's products (Aneri Salwar Suits):

* Pair the black & maroon salwar suit with a simple white or light-colored blouse to let the outfit shine.

* For a more formal look, pair the blue & white salwar suit with a crisp white shirt and minimal jewelry.

For men's accessories:

* The Basics Men Black Blazer Jacket is versatile - wear it with navy trousers for a classic look or beige trousers for a more relaxed vibe. Pair it with either ADIDAS or Parx black socks.

* For the Park Avenue ties, pair the red patterned tie with white dress shirt and dark-colored trousers (black coffee brown formal trousers work well). The blue patterned tie looks great with light-colored shirts and navy trousers.

For footwear:

* Woodland Men Brown Leather Shoes are perfect for a more casual look. Pair them with beige or light-colored pants.

* Clarks Brown Leather Casual Shoes can be dressed up or down, depending on the occasion. Try

Let’s try out another prompt:

response = module("wedding occasion")



Products: Avirate Black Formal Dress, French Connection Teal Dress, s.Oliver Women's White Blouse Top, United Colors of Benetton Men Woven Black Tie, Enroute Women Casual Taupe Heels, Enroute Women Casual Taupe Heels, Ivory Tag Women Colour Interplay Multicolour Necklace, Ivory Tag Women Colour Beam Multicolour Necklace, French Connection Teal Dress, FNF Red Bridal Collection Lehenga Choli, Streetwear Satin Smooth Brides Mate Lipcolor 05, Jacques M Women Be Gracious Deo


For a stylish and cohesive look with these products:

* Pair the Avirate Black Formal Dress with Enroute Women Casual Taupe Heels for a chic evening outfit. Add Ivory Tag Women Colour Interplay Multicolour Necklace to add some sparkle.

* The French Connection Teal Dress looks stunning on its own, but you can also pair it with United Colors of Benetton Men Woven Black Tie as an unexpected yet stylish combination.

* For a more casual look, team the s.Oliver Women's White Blouse Top with Enroute Women Casual Taupe Heels and Ivory Tag Women Colour Beam Multicolour Necklace for a chic everyday outfit.

* The FNF Red Bridal Collection Lehenga Choli is perfect for a special occasion. Pair it with Streetwear Satin Smooth Brides Mate Lipcolor 05 to complete the look.

* For men, the United Colors of Benetton Men Woven Black Tie can be paired with any formal attire.

Remember, accessorizing wisely and mixing high and low pieces can create a unique and stylish outfit that reflects your personal taste. 

Final Words

DSPy’s signatures, modules, and optimizers provide a powerful method for building a multistage agent that moves away from traditional prompt engineering. In this blog, we learned how to build a multistage Personal Styling Agent using DSPy’s unique framework.

By taking inspiration from the above technique, you can design similar agents to provide assistance with activities like financial advising, travel planning, healthcare advice, and a variety of other use cases.

You can always rent a cloud GPU on E2ENetworks or you can deploy your entire stack on their TIR AI platform, which comes with a set of tools, containers, and readymade endpoints that can assist you with your AI development. 


Latest Blogs
This is a decorative image for: A Complete Guide To Customer Acquisition For Startups
October 18, 2022

A Complete Guide To Customer Acquisition For Startups

Any business is enlivened by its customers. Therefore, a strategy to constantly bring in new clients is an ongoing requirement. In this regard, having a proper customer acquisition strategy can be of great importance.

So, if you are just starting your business, or planning to expand it, read on to learn more about this concept.

The problem with customer acquisition

As an organization, when working in a diverse and competitive market like India, you need to have a well-defined customer acquisition strategy to attain success. However, this is where most startups struggle. Now, you may have a great product or service, but if you are not in the right place targeting the right demographic, you are not likely to get the results you want.

To resolve this, typically, companies invest, but if that is not channelized properly, it will be futile.

So, the best way out of this dilemma is to have a clear customer acquisition strategy in place.

How can you create the ideal customer acquisition strategy for your business?

  • Define what your goals are

You need to define your goals so that you can meet the revenue expectations you have for the current fiscal year. You need to find a value for the metrics –

  • MRR – Monthly recurring revenue, which tells you all the income that can be generated from all your income channels.
  • CLV – Customer lifetime value tells you how much a customer is willing to spend on your business during your mutual relationship duration.  
  • CAC – Customer acquisition costs, which tells how much your organization needs to spend to acquire customers constantly.
  • Churn rate – It tells you the rate at which customers stop doing business.

All these metrics tell you how well you will be able to grow your business and revenue.

  • Identify your ideal customers

You need to understand who your current customers are and who your target customers are. Once you are aware of your customer base, you can focus your energies in that direction and get the maximum sale of your products or services. You can also understand what your customers require through various analytics and markers and address them to leverage your products/services towards them.

  • Choose your channels for customer acquisition

How will you acquire customers who will eventually tell at what scale and at what rate you need to expand your business? You could market and sell your products on social media channels like Instagram, Facebook and YouTube, or invest in paid marketing like Google Ads. You need to develop a unique strategy for each of these channels. 

  • Communicate with your customers

If you know exactly what your customers have in mind, then you will be able to develop your customer strategy with a clear perspective in mind. You can do it through surveys or customer opinion forms, email contact forms, blog posts and social media posts. After that, you just need to measure the analytics, clearly understand the insights, and improve your strategy accordingly.

Combining these strategies with your long-term business plan will bring results. However, there will be challenges on the way, where you need to adapt as per the requirements to make the most of it. At the same time, introducing new technologies like AI and ML can also solve such issues easily. To learn more about the use of AI and ML and how they are transforming businesses, keep referring to the blog section of E2E Networks.

Reference Links

This is a decorative image for: Constructing 3D objects through Deep Learning
October 18, 2022

Image-based 3D Object Reconstruction State-of-the-Art and trends in the Deep Learning Era

3D reconstruction is one of the most complex issues of deep learning systems. There have been multiple types of research in this field, and almost everything has been tried on it — computer vision, computer graphics and machine learning, but to no avail. However, that has resulted in CNN or convolutional neural networks foraying into this field, which has yielded some success.

The Main Objective of the 3D Object Reconstruction

Developing this deep learning technology aims to infer the shape of 3D objects from 2D images. So, to conduct the experiment, you need the following:

  • Highly calibrated cameras that take a photograph of the image from various angles.
  • Large training datasets can predict the geometry of the object whose 3D image reconstruction needs to be done. These datasets can be collected from a database of images, or they can be collected and sampled from a video.

By using the apparatus and datasets, you will be able to proceed with the 3D reconstruction from 2D datasets.

State-of-the-art Technology Used by the Datasets for the Reconstruction of 3D Objects

The technology used for this purpose needs to stick to the following parameters:

  • Input

Training with the help of one or multiple RGB images, where the segmentation of the 3D ground truth needs to be done. It could be one image, multiple images or even a video stream.

The testing will also be done on the same parameters, which will also help to create a uniform, cluttered background, or both.

  • Output

The volumetric output will be done in both high and low resolution, and the surface output will be generated through parameterisation, template deformation and point cloud. Moreover, the direct and intermediate outputs will be calculated this way.

  • Network architecture used

The architecture used in training is 3D-VAE-GAN, which has an encoder and a decoder, with TL-Net and conditional GAN. At the same time, the testing architecture is 3D-VAE, which has an encoder and a decoder.

  • Training used

The degree of supervision used in 2D vs 3D supervision, weak supervision along with loss functions have to be included in this system. The training procedure is adversarial training with joint 2D and 3D embeddings. Also, the network architecture is extremely important for the speed and processing quality of the output images.

  • Practical applications and use cases

Volumetric representations and surface representations can do the reconstruction. Powerful computer systems need to be used for reconstruction.

Given below are some of the places where 3D Object Reconstruction Deep Learning Systems are used:

  • 3D reconstruction technology can be used in the Police Department for drawing the faces of criminals whose images have been procured from a crime site where their faces are not completely revealed.
  • It can be used for re-modelling ruins at ancient architectural sites. The rubble or the debris stubs of structures can be used to recreate the entire building structure and get an idea of how it looked in the past.
  • They can be used in plastic surgery where the organs, face, limbs or any other portion of the body has been damaged and needs to be rebuilt.
  • It can be used in airport security, where concealed shapes can be used for guessing whether a person is armed or is carrying explosives or not.
  • It can also help in completing DNA sequences.

So, if you are planning to implement this technology, then you can rent the required infrastructure from E2E Networks and avoid investing in it. And if you plan to learn more about such topics, then keep a tab on the blog section of the website

Reference Links

This is a decorative image for: Comprehensive Guide to Deep Q-Learning for Data Science Enthusiasts
October 18, 2022

A Comprehensive Guide To Deep Q-Learning For Data Science Enthusiasts

For all data science enthusiasts who would love to dig deep, we have composed a write-up about Q-Learning specifically for you all. Deep Q-Learning and Reinforcement learning (RL) are extremely popular these days. These two data science methodologies use Python libraries like TensorFlow 2 and openAI’s Gym environment.

So, read on to know more.

What is Deep Q-Learning?

Deep Q-Learning utilizes the principles of Q-learning, but instead of using the Q-table, it uses the neural network. The algorithm of deep Q-Learning uses the states as input and the optimal Q-value of every action possible as the output. The agent gathers and stores all the previous experiences in the memory of the trained tuple in the following order:

State> Next state> Action> Reward

The neural network training stability increases using a random batch of previous data by using the experience replay. Experience replay also means the previous experiences stocking, and the target network uses it for training and calculation of the Q-network and the predicted Q-Value. This neural network uses openAI Gym, which is provided by taxi-v3 environments.

Now, any understanding of Deep Q-Learning   is incomplete without talking about Reinforcement Learning.

What is Reinforcement Learning?

Reinforcement is a subsection of ML. This part of ML is related to the action in which an environmental agent participates in a reward-based system and uses Reinforcement Learning to maximize the rewards. Reinforcement Learning is a different technique from unsupervised learning or supervised learning because it does not require a supervised input/output pair. The number of corrections is also less, so it is a highly efficient technique.

Now, the understanding of reinforcement learning is incomplete without knowing about Markov Decision Process (MDP). MDP is involved with each state that has been presented in the results of the environment, derived from the state previously there. The information which composes both states is gathered and transferred to the decision process. The task of the chosen agent is to maximize the awards. The MDP optimizes the actions and helps construct the optimal policy.

For developing the MDP, you need to follow the Q-Learning Algorithm, which is an extremely important part of data science and machine learning.

What is Q-Learning Algorithm?

The process of Q-Learning is important for understanding the data from scratch. It involves defining the parameters, choosing the actions from the current state and also choosing the actions from the previous state and then developing a Q-table for maximizing the results or output rewards.

The 4 steps that are involved in Q-Learning:

  1. Initializing parameters – The RL (reinforcement learning) model learns the set of actions that the agent requires in the state, environment and time.
  2. Identifying current state – The model stores the prior records for optimal action definition for maximizing the results. For acting in the present state, the state needs to be identified and perform an action combination for it.
  3. Choosing the optimal action set and gaining the relevant experience – A Q-table is generated from the data with a set of specific states and actions, and the weight of this data is calculated for updating the Q-Table to the following step.
  4. Updating Q-table rewards and next state determination – After the relevant experience is gained and agents start getting environmental records. The reward amplitude helps to present the subsequent step.  

In case the Q-table size is huge, then the generation of the model is a time-consuming process. This situation requires Deep Q-learning.

Hopefully, this write-up has provided an outline of Deep Q-Learning and its related concepts. If you wish to learn more about such topics, then keep a tab on the blog section of the E2E Networks website.

Reference Links

This is a decorative image for: GAUDI: A Neural Architect for Immersive 3D Scene Generation
October 13, 2022

GAUDI: A Neural Architect for Immersive 3D Scene Generation

The evolution of artificial intelligence in the past decade has been staggering, and now the focus is shifting towards AI and ML systems to understand and generate 3D spaces. As a result, there has been extensive research on manipulating 3D generative models. In this regard, Apple’s AI and ML scientists have developed GAUDI, a method specifically for this job.

An introduction to GAUDI

The GAUDI 3D immersive technique founders named it after the famous architect Antoni Gaudi. This AI model takes the help of a camera pose decoder, which enables it to guess the possible camera angles of a scene. Hence, the decoder then makes it possible to predict the 3D canvas from almost every angle.

What does GAUDI do?

GAUDI can perform multiple functions –

  • The extensions of these generative models have a tremendous effect on ML and computer vision. Pragmatically, such models are highly useful. They are applied in model-based reinforcement learning and planning world models, SLAM is s, or 3D content creation.
  • Generative modelling for 3D objects has been used for generating scenes using graf, pigan, and gsn, which incorporate a GAN (Generative Adversarial Network). The generator codes radiance fields exclusively. Using the 3D space in the scene along with the camera pose generates the 3D image from that point. This point has a density scalar and RGB value for that specific point in 3D space. This can be done from a 2D camera view. It does this by imposing 3D datasets on those 2D shots. It isolates various objects and scenes and combines them to render a new scene altogether.
  • GAUDI also removes GANs pathologies like mode collapse and improved GAN.
  • GAUDI also uses this to train data on a canonical coordinate system. You can compare it by looking at the trajectory of the scenes.

How is GAUDI applied to the content?

The steps of application for GAUDI have been given below:

  • Each trajectory is created, which consists of a sequence of posed images (These images are from a 3D scene) encoded into a latent representation. This representation which has a radiance field or what we refer to as the 3D scene and the camera path is created in a disentangled way. The results are interpreted as free parameters. The problem is optimized by and formulation of a reconstruction objective.
  • This simple training process is then scaled to trajectories, thousands of them creating a large number of views. The model samples the radiance fields totally from the previous distribution that the model has learned.
  • The scenes are thus synthesized by interpolation within the hidden space.
  • The scaling of 3D scenes generates many scenes that contain thousands of images. During training, there is no issue related to canonical orientation or mode collapse.
  • A novel de-noising optimization technique is used to find hidden representations that collaborate in modelling the camera poses and the radiance field to create multiple datasets with state-of-the-art performance in generating 3D scenes by building a setup that uses images and text.

To conclude, GAUDI has more capabilities and can also be used for sampling various images and video datasets. Furthermore, this will make a foray into AR (augmented reality) and VR (virtual reality). With GAUDI in hand, the sky is only the limit in the field of media creation. So, if you enjoy reading about the latest development in the field of AI and ML, then keep a tab on the blog section of the E2E Networks website.

Reference Links

Build on the most powerful infrastructure cloud

A vector illustration of a tech city using latest cloud technologies & infrastructure