Product frame
The project is not about transcript export. It is about returning to a project and recovering its concepts, architecture shifts, engineering decisions, recurring patterns, and unresolved questions.
Exploration · Conversation history as project memory
chatgpt-history
chatgpt-history explores a practical problem: when project thinking is scattered across many ChatGPT sessions, how can it be rebuilt into readable, traceable, reusable project memory instead of preserved as raw chat logs?
Narrative
Design and engineering judgement
Engineering frame
The pipeline reads exported ChatGPT markdown, creates session-level summaries, uses turn-pair chunks as evidence, then runs embedding, clustering, knowledge synthesis, timeline generation, and deterministic report rendering.
Design frame
This is knowledge-interface design: the output should read like a technical retrospective, showing how a project formed and which conversations support each theme.
System
Components and visibility
| Component | Type | Visibility | Role |
|---|---|---|---|
| ChatGPT markdown corpus | Input corpus | Local/private data | Keeps project conversations and metadata as read-only local analysis input. |
| Session analysis | Topic layer | Public implementation | Turns each conversation into a structured summary for topic discovery. |
| Turn-pair chunks | Evidence layer | Public implementation | Adds fine-grained evidence, recurring concepts, and traceability. |
| Project report renderer | Knowledge artifact | Public implementation | Renders project_knowledge, timeline, and cluster records into readable Markdown/PDF reports. |
Public URLs