Athena was an internal project from Oxari that aimed, while riding the AI boom, to provide ESG Analysts and Leaders with a way to effortlessly access data and perform tasks such as benchmarking or materiality assessments using natural language without exiting a chat interface. The project never left the prototype phase.
The answers provided by Athena were split into two major distinct categories:
Quantitative — which would perform RAG over content extracted from corporate reports (annual reports, sustaibalitiy reports, responsability reports, compnay policies, etc.)
Quantitative — which would interpret the request and construct a mongodb aggregation pipeline query to perform the necessary operations and return the result together with a visualization.
The file db_document_example_apple.json provides an example of the document structure present in the numerical database.
Check the python version inside runtime.txt
conda create -n athenaenv pyhton=3.12.7
conda activate athenaenv
pip install -r requirements.txt .env
OPENAI_API_KEY='YOUR_OPENAI_KEY'
# text-embedding-3-small was previously used
EMBEDDINGS_MODEL='YOUR_CHOSEN_EMBEDDINGS'
# MongoDB free cluster containing numerical data
ATLAS_CLUSTER_URI='mongodb+srv://<user>:<password>@prototypingcluster.ibemdxn.mongodb.net/?retryWrites=true&w=majority&appName=PrototypingCluster'
# The project has a flavour of email verification that made use of smtppro.zoho.eu
# Optional if email verification wants to be kept; this can be bypassed by passing a valid verification code from the start
EMAIL_SENDER_ADDR=
EMAIL_SENDER_PWD=For the Oxari guest read-only access athenaguest can be used as the user, together with zbcb8YlifqxbOOsY as the password.
streamlit run main.pyImportant
After the application starts on http://localhost:8501/, TEST can be provided as a valid verification code in order to skip the email verification step, which was left active.
The following questions were tested during the development stage:
- What are the scope 2 emissions of Meta, Apple, Lenovo, and Netflix since 2020?
- Why are the scope 2 emissions of Apple so low since 2021?
- Provide scope 2 to net income ratio for the companies in 2023
- What is the average scope 2 emissions for the companies in the technology sector compared to the communication services sector?
- What are the ESG metrics in your database?
- Which companies are in the database?
- Give me the number of companies in the database.
- How many distinct companies are in the db?
- Which water related metrics do you have access to?
- Summarize the main points of this chat
-
Give me the operating profit, the revenue and the total assets of meta
-
What are scope 2 for meta and apple?
-
Give me the revenue of both meta and apple over the last 4 years
-
What is the average water consumption for the companies in the technology sector compared to the communication services sector?
-
Can you compare Apple and Netflix in terms of scope 2 intensities?
-
What was the increase in water consumption for Meta in the last 3 years?
Follow-up:
-
What was the reason for this negative increase? [Obs: the fact that we are still talking about Meta is inferred from the history]
-
What is the year when Apple polluted the most?
-
Please give me the combined scope 1 and 2 emissions location based for apple.
-
Give me the profit of Meta and Apple in 2023.
-
Provide me the newest operating cash flow of Meta and Apple.
-
Can you tell me the number of females in the supervisory board of Netflix?
-
Can you tell me the gender distribution in the supervisory board of Netflix?
- Can you tell me a couple of things about the diversity and inclusion of Netflix?
- What is the decarbonization policy of Meta?
Prototype developed at Oxari B.V. (now a closed entity)
If this is of any interest, please feel free to contact one of the main contributors: