This project is a LangGraph-based conversational assistant designed to help users make short-term decisions about their cryptocurrency portfolio. It combines LLM-powered reasoning with tool integrations for real-time crypto news retrieval, cost management of token use, and optional conversation memory storage in SQLite.

• Portfolio Advice: Uses ChatOpenAI (gpt-4o) to provide guidance on short-term portfolio actions.
• Crypto News Tool: Fetches the latest cryptocurrency news from trusted sources via the Tavily API.
• Conversation Summarization: Summarizes long conversations to keep context and cost manageable.
• Optional Memory: Persist state and summaries using a local SQLite database.
• LangGraph Workflow:
  • assistant → Handles user queries and portfolio advice.
  • tools → Fetches crypto news when needed.
  • summarise_conversation → Periodically summarizes the dialogue to reduce context length.

1. Clone this repository:

   git clone <your-repo-url>
   cd <your-repo-name>
   

2. Create and activate a Python virtual environment:

python -m venv .venv source .venv/bin/activate # macOS/Linux .venv\Scripts\activate # Windows

3. Install dependencies:

pip install -r requirements.txt

Example requirements.txt:

langchain-openai langchain-core langgraph requests

You’ll need to set the following environment variables in a .env file:

OPENAI_API_KEY – your OpenAI API key (for ChatOpenAI).

TAVILY_API_KEY – your Tavily API key (for crypto news retrieval).

See also the .env_example file for the other variables that must be set.

Run the assistant by importing the graph in your Python session or script:

from assistant import graph, State

state = State(messages=[])
output = graph.invoke(state)
print(output)

The assistant will:

• Take the user’s portfolio as input.
• Retrieve crypto news (via Tavily API).
• Provide short-term advice based on the combination of portfolio and news.
• Summarise conversation history if it is too long to manage token costs.
• Optionally, save conversation history if memory is enabled.

See the testing_assistant.py script for further examples.

If you set with_memory = True in assistant.py, the assistant will store state and conversation summaries in a local SQLite database.

Default location: ./state_db/example.db

Database is created automatically if it doesn’t exist.

assistant.py         # Main assistant code
state_db/            # (Optional) Folder for SQLite memory
requirements.txt     # Python dependencies
README.md            # Project documentation
testing_assistant.py # Examples of how to use the assistant
other_tools.py       # Other tools in the process of being developed
query_db.py          # Examples of how to query the database to check past conversations
studio/              # Everything required to deploy LangGraph Studio

• LangGraph Studio is a custom IDE for viewing and testing agents.
• Studio can be run locally and opened in your browser on Mac, Windows, and Linux.
• Graphs for LangGraph Studio are in the studio/ folder.
• To start the local development server, run the following command in your terminal in the /studio directory:

langgraph dev

You should see the following output:

- 🚀 API: http://127.0.0.1:2024
- 🎨 Studio UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
- 📚 API Docs: http://127.0.0.1:2024/docs

Open your browser and navigate to the Studio UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024.

I created this assistant to learn the following concepts of LangGraph:

• Agent
• State (multiple states, reducers)
• Memory (checkpoints, external)
• Summarisation
• Humman-in-the-loop (streaming, (breakpoints, edit state (not implemented))
• Controlability (parallelisation, subgraphs, map-reduce) >> not yet implemented.
• Observability

• Parallelise Tavily calls when the LLM decides to search for information about multiple coins.
• Use multiple tools (e.g. a tool that downloads the past performance of a coin, makes a forcast or set of forecasts, and the LLM uses that in addition to the internet search to make a recommendation).
• Traceability and observability with LangSmith.

⚠## ️ Disclaimer

This assistant provides educational and informational guidance only. It is not financial advice. Always do your own research before making investment decisions.