Analysis & Exports

When you share your chatbot as a learning tool, Illinois Chat shows you how people use it — helping you understand your audience's needs and improve your content. Only the owner and administrators can see the Analysis page and run exports.

The Analysis page

Open Analysis in your chatbot to see statistics and charts:

• Usage over time — conversation and message volume, including weekly trends.
• Model usage — which LLMs are being used and how often.

Use it to spot gaps: questions that come up repeatedly with weak answers usually mean a document is missing from your knowledge base. To read what users actually asked, export the conversations.

Export all conversations

On the Analysis page, Download Conversation History downloads all conversations anyone has had with your chatbot as a zip containing:

• markdown export/ — one readable Markdown file per conversation,
• media_files/ — images attached to messages,
• an .xlsx spreadsheet and a .jsonl file of the same conversations,
• error.log for anything that could not be exported.

Data format

The .jsonl file is JSON Lines, one conversation per row:

• If a user was authenticated when chatting, their email address is included; otherwise null.
• Messages mirror OpenAI's chat format (role/content pairs).
• Each assistant message additionally includes the contexts that were (potentially) used to answer — up to 100 contexts per response, each with text, readable_filename, s3_path, url, and pagenumber metadata.

Reading the export

import jsonlines
import pprint

filename = 'myProject-convo_history.jsonl'
with jsonlines.open(filename) as f:
    data = list(f)

print(len(data))
pprint.pprint(data[0])

Each row looks like:

{
  "convo_id": "03a9ffb3-5bde-4766-a4eb-66dff42ed8ac",
  "course_name": "my-project",
  "user_email": "user@illinois.edu",
  "created_at": "2025-08-14T16:35:40.508062-07:00",
  "convo": {
    "model": {"id": "gpt-4o", "name": "GPT-4o"},
    "prompt": "Your system prompt...",
    "temperature": 0.4,
    "messages": [
      {"role": "user", "content": "...", "contexts": []},
      {"role": "assistant", "content": "...", "contexts": [{"text": "...", "readable_filename": "...", "url": "...", "pagenumber": ""}]}
    ]
  }
}

Export all documents

Export in the Project Files table on the Dashboard downloads a zip with one .jsonl file of the post-processed text and vector embeddings used by the LLM. To minimize data-transfer costs, exporting original files (PDFs, etc.) is only available per-document.

Large exports

When an export covers more than 500 conversations or documents it runs as a background job. A download link is emailed to the chatbot's owner and administrators when the deployment has the backend's email settings (see the Configuration reference); otherwise the export is written to object storage and nothing is sent.

A user's own history

Each user can download their own conversations with the chatbot: Export history in the chat sidebar produces a zip of Markdown files plus media. See Using a chatbot.

Privacy notes

• If a user is authenticated when chatting, their email is included in conversation logs; otherwise it is null.
• Only the owner and administrators can access conversation history and exports.