Project Info

Winner

Best AI x Crypto with Infinity

Midas

This project did not submit a demo video on Devpost.

Inspiration

The intersection between human emotion and technology has been rapidly expanding, providing Data Science and Computer Science majors such as us to explore and push the boundaries of our collective knowledge through new technologies that grow and improve everyday.

What it does

Midas uses a webcam to process data about traders' facial expressions in real-time, creating and contributing to a user database of processed statistical information about their emotions during their different trades. By pairing information about the trades, their outcomes, and the emotions exhibited throughout, a LLM then generates custom feedback for each user to improve their performance based on their personal emotional data.

How we built it

Midas was built using a combination of Hume.AI and LLM training. We used Python for the backend as well as front end using REFLEX. Hume was used to get emotional data and face tracking. MindsDB was use for SQL database integration into our LLM.

Challenges we ran into

Developing an accurate reading of someone's emotions over a period of time and how exactly to train the LLm based on these emotions. Additionally, ensuring user privacy and data security was a concern.

Accomplishments we're proud of

We successfully created a functional system to analyze traders' emotions during their trades. Integrating emotional data with trade-related information is a unique feature of our project. We're proud of our LLM's ability to generate personalized feedback based on a user's emotional data.

What we learned

We gained a deeper understanding of emotion tracking, Natural Languge Processing, and data analysis techniques. Additionally, we learned about the challenges and ethical considerations involved in working with user data and privacy.

What's next

There are several potential future directions for Midas. This might include expanding the types of emotional data collected, conducting user studies to assess the impact of emotional feedback on trading performance, and enhancing the overall user experience. We may also explore opportunities to integrate with trading platforms or provide more detailed analytics for traders. Further development could involve ensuring regulatory compliance in the finance industry and addressing potential privacy concerns. Additionally, we could expand the use of the technology beyond trading, such as in education or mental health applications.

Analysis

Compare with all teams

View

Metric

Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.

Technology

Found in codeClaimed only

CSSIn code
JavaScriptIn code
Next.jsIn code
PythonIn code
ReactIn code
SQLIn code
Tailwind CSSIn code

7 of 7 appear in the indexed code.

AI coding agents

No AI coding agent signals were found in this repository.

Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.

Codebase size

Source size

170 KB

Source files

Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.

Feature verification

2 Verified6 Code-supported2 Claimed only

psycopg2 Postgres integrationVerified
Built with psycopg2 (listed in Built With)
Claimed on Devposthigh confidence
- antitilt.py:2 — imports psycopg2 and uses psycopg2.connect(**connection_params) across create_crypto_tables, insert_into_table, filter_table_by_emotion, to_df
Reflex-based frontend/backend web appVerified
Built using Python for the backend as well as front end using REFLEX
Claimed on Devposthigh confidence
- website/website/website.py — Reflex app entrypoint file present
- website/rxconfig.py — Reflex project config
- website/requirements.txt:2 — reflex==0.3.1 declared as a dependency
- website/website/pages/index.py:1 — Reflex page components imported and rendered (rx.hstack, rx.box, rx.text)
Displaying emotion/trade data table in the UICode-supported
Implied UI component for showing collected emotional/trade statistics to the user
Claimed on readmelow confidence
- website/website/components/humetable.py:10 — Defines an rx.data_table bound to a State with the 48 Hume emotion names as a single data row and generic 'Column N' headers, not real per-trade data, and it's unclear if this component is rendered on any page
Emotion data logged to a user/trade databaseCode-supported
Creating and contributing to a user database of processed statistical information about their emotions during their different trades
Claimed on Devpostmedium confidence
- antitilt.py:154 — insert_into_table() writes currency, emotion, weight, time_of_trade, category, delta into a Postgres table Golden_Touch via psycopg2
- webcamHume.py:108 — webcamHume.py calls insert_into_table(...) with the computed top emotion and description after Hume analysis, wiring capture to storage
Hume API facial emotion analysisCode-supported
Analyzes traders' facial expressions with Hume API to log data about their emotions
Claimed on Devpostmedium confidence
- webcamHume.py:55 — Submits recorded video to HumeBatchClient with FaceConfig(identify_faces=True) and downloads predictions to trade_emotions.json
- trade_emotions.json — Contains real Hume-format emotion prediction output, suggesting the pipeline was actually run at least once
LLM-generated personalized feedback based on emotional dataCode-supported
A LLM then generates custom feedback for each user to improve their performance based on their personal emotional data
Claimed on Devpostlow confidence
- website/Miscalgos.sql:27 — Defines a MindsDB CREATE MODEL candlelit using OpenAI gpt-4 with a prompt template that takes emotion and crypto trade data and generates short trading advice, but this SQL is a standalone script, not invoked from any Python code found in the repo
MindsDB SQL database integration into the LLMCode-supported
MindsDB was used for SQL database integration into our LLM
Claimed on Devpostlow confidence
- website/Miscalgos.sql:2 — CREATE DATABASE Dummy WITH ENGINE = postgres and a CREATE JOB binance_data joining Hume and binance data, plus the candlelit MindsDB model, but these are raw SQL snippets not called from any application code
- antitilt.py:5 — imports mindsdb_sdk and has a commented-out mindsdb_sdk.connect(...) call, but no active usage in the file
Webcam facial expression capture during tradesCode-supported
Midas uses a webcam to process data about traders' facial expressions in real-time
Claimed on Devpostmedium confidence
- webcamHume.py:26 — Opens cv2.VideoCapture(0), records webcam frames to tradefootage.mp4 until Esc is pressed
- init.py — init.py is an empty file, a duplicate/earlier webcam capture script (quicktest.py) exists but is not wired into the app
End-to-end wiring from webcam UI button to emotion captureClaimed only
Implied by 'What it does': a functioning system where starting a trade triggers webcam-based emotion tracking
Claimed on readmehigh confidence
Pairing trade outcomes with exhibited emotionsClaimed only
By pairing information about the trades, their outcomes, and the emotions exhibited throughout
Claimed on Devpostmedium confidence

An AI agent derived these features from the project’s Devpost page and readme, then searched the code for each one. Verified features are backed by cited code; claimed-only features had no supporting code, which is not by itself proof a feature is missing.