Update README.md
Browse files
README.md
CHANGED
|
@@ -44,7 +44,7 @@ requirements.txt
|
|
| 44 |
# Design Considerations and Constraints
|
| 45 |
* A Streamlit app
|
| 46 |
* A free LLM API which allows us to test without running into concerns with running out of credits - gemini-2.5-flash-preview-04-17-thinking
|
| 47 |
-
* A way to send relevant information to the LLM
|
| 48 |
* We tried sending the full dataset in a dataframe containing approx 4.1m tokens. This failed because Gemini only accepts a maximum of 1m tokens as input
|
| 49 |
* We tried splitting the full dataset into 5 equal chunks (worked), uploaded each chunk for the chat interface (worked), but unable to use all 5 chunks (ie the full dataset) in the chat (failed)
|
| 50 |
* Since we were unable to send the huge dataset to Gemini, we explored sending a summary of the dataset
|
|
@@ -69,4 +69,24 @@ requirements.txt
|
|
| 69 |
|
| 70 |
# Architecture
|
| 71 |
* With the above design considerations and constraints, we created the chatbot with the following architecture
|
| 72 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
# Design Considerations and Constraints
|
| 45 |
* A Streamlit app
|
| 46 |
* A free LLM API which allows us to test without running into concerns with running out of credits - gemini-2.5-flash-preview-04-17-thinking
|
| 47 |
+
* A way to send relevant information/grounding context to the LLM
|
| 48 |
* We tried sending the full dataset in a dataframe containing approx 4.1m tokens. This failed because Gemini only accepts a maximum of 1m tokens as input
|
| 49 |
* We tried splitting the full dataset into 5 equal chunks (worked), uploaded each chunk for the chat interface (worked), but unable to use all 5 chunks (ie the full dataset) in the chat (failed)
|
| 50 |
* Since we were unable to send the huge dataset to Gemini, we explored sending a summary of the dataset
|
|
|
|
| 69 |
|
| 70 |
# Architecture
|
| 71 |
* With the above design considerations and constraints, we created the chatbot with the following architecture
|
| 72 |
+
|
| 73 |
+
* Preparation steps
|
| 74 |
+
* For dataset and model
|
| 75 |
+
* Download dataset
|
| 76 |
+
* Exploratory data analysis
|
| 77 |
+
* Prepare grounding information
|
| 78 |
+
* Prepare system prompt
|
| 79 |
+
* Gemini API chat configuration
|
| 80 |
+
* Tool call - GoogleSearch
|
| 81 |
+
* For Streamlit app
|
| 82 |
+
* Configure chat interface
|
| 83 |
+
* During normal usage of the chatbot
|
| 84 |
+
* User loads the chat interface and picks a question from the 'Example questions' section
|
| 85 |
+
* User types their question in the 'Ask a question' box and submits their question to Gemini (our LLM)
|
| 86 |
+
* The system prompt (which contains the dataframe summary, i.e. grounding information) also gets sent to Gemini
|
| 87 |
+
* Gemini can
|
| 88 |
+
* Analyse the user's question and respond directly, sending the response back to our Streamlit app, which processes the response and shows on the chat box
|
| 89 |
+
* Alternatively assess whether a tool call is required
|
| 90 |
+
* If Gemini decides that a tool call is required to answer an 'out of domain' question', Gemini make 1 or more calls to the GoogleSearch tool
|
| 91 |
+
* Then Gemini will formulate a response, sends that response to our Streamlit app, to process and show on the chat box
|
| 92 |
+
* We also extract the search links and show below the LLM's answer, for users to verify the LLM's response if they wish
|