mirableio / chat-history

ChatGPT history exploration

ChatGPT history

UI for navigating and organizing OpenAI's ChatGPT conversations.

Important: This project is 100% unaffiliated with OpenAI.

Features

• See activity graph and useful statistics
• Quickly browse and open the chats
• Search chats (semantic and "strict")
• List of favorite chats
• Open conversations on the ChatGPT site

Setup

Currently can only be installed locally. Requires Python 3.10+

1. Export ChatGPT history
2. Unzip the download, place conversations.json in the data folder
3. make install
4. make run
5. Open http://127.0.0.1:8080 in your browser
6. Optional - copy secrets.template.toml to data/secrets.toml and update OpenAI API key, then restart the server. First run will take a while to create embeddings. 10MB JSON: ~30 min, ~$0.10 cost.

chat-history's Issues

Knowledge graph view

Like https://help.obsidian.md/Plugins/Graph+view

Automatic node naming / tag conversations

Semantically related chats

By vector similarity

Mark success/failure

Hide conversations

Annotations

Quick sidebar filter

Custom titles

Add custom titles to conversations

Semantic search

Hash-based navigation

Hash links to:

• separate conversations
• search results
• statistics

Memory optimization

Remember filters

Or at least dropdown value

Show metadata

            "message": {
                "id": "---",
                "author": {
                    "role": "system",
                    "name": null,
                    "metadata": {}
                },
                "create_time": null,
                "update_time": null,
                "content": { "content_type": "text", "parts": [""] },
                "status": "finished_successfully",
                "end_turn": null,
                "weight": 1.0,
                "metadata": {
                    "is_user_system_message": true,
                    "user_context_message_data": {
                        "about_user_message": "XXXX",
                        "about_model_message": "YYYYY"
                    }
                },
                "recipient": "all"
            },

The current code doesn't accept the previous version of the conversion.json file since the Chatpt began until 20/12/2022

I have noticed an issue on your project the current code hasn't accepted the previous version of the conversion.json file since the Chatpt began until 20/12/2022 there is an explanation:

Differences between my modification and your current code In history.py:

1. Pydantic Usage:

   • Difference: my modification uses Pydantic's BaseModel and Field for defining the Part class, whereas your current code doesn't.
   • Reason: Pydantic is a data validation and serialization library. The usage of BaseModel suggests that my modification is likely aiming for better data validation and easier serialization/deserialization.

2. Typing Imports:

   • Difference: my modification has a separate import statement for typing functionalities compared to your current code.
   • Reason: Different import statements can change how types are referred to in the code. The one in my modification might be more standard or clearer.

3. Attributes in Part Class:

   • Difference: my modification defines several attributes (like asset_pointer, size_bytes, etc.) that are not present in your current code.
   • Reason: These attributes likely represent important data fields for the Part class. Their absence in your current code suggests it might be an incomplete or older version.

4. Parts Attribute:

   • Difference: Both files have a parts attribute, but they are defined differently. your current code has a more complex type hint involving OrderedDict.
   • Reason: This suggests that the structure of the parts attribute was changed or is being considered for a change. The usage of OrderedDict might indicate an attempt to maintain the order of elements.

5. Commented parent_id:

   • Difference: your current code has the parent_id attribute commented out.
   • Reason: It's possible that there was uncertainty about whether this attribute should be included or not. Commenting it out might have been a temporary measure.

Conclusion:

my modification seems to be a more refined and updated version with better data validation using Pydantic and clearer type hints. It's advisable to consider it as the base and merge any necessary changes from history.py if needed.

Favorite single messages

Bookmarks inside long conversations

Embeddings error handling

If error happens during embedding creation, handle it without losing all the progress

Single chat statistics

• chat length
• highlight significant delays between messages

Show plugins

Check for chat forks

When user regenerates the input or edits original message

Export to Markdown

https://github.com/daugaard47/ChatGPT_Conversations_To_Markdown/

Cross-chats

When viewing a conversation, highlight if I have messages in another conversation with similar timestamp

Favorite chats