feat: Conversation History Export Tool

[alperents11]

Problem

Currently, there's no way to export agent conversations for:

• Debugging and analysis
• Sharing results with team members
• Creating reports or documentation
• Archiving important interactions

Users must manually copy-paste from logs or terminal output.

Proposed Solution

Create a conversation export tool that saves agent interactions in multiple formats:

python export_conversation.py --agent spot --session 2026-01-29 --format markdown

Exported to: conversations/spot_2026-01-29.md

Supported Formats:

• Markdown (.md) - Human-readable, easy to share
• JSON (.json) - Machine-readable, for analysis
• CSV (.csv) - For spreadsheet analysis
• HTML (.html) - Pretty formatted with styling

Example Output (Markdown):

# Conversation: Spot Agent
Date: 2026-01-29
Duration: 5 minutes

## Exchange 1 (14:30:45)
**User:** Pick up the red ball
**Vision:** Detected red sphere at (0.5, 0.2)
**LLM Decision:** movement(0.5, 0, 0)
**Action:** Moving forward...

## Exchange 2 (14:30:52)
...

Benefits

• Debugging: Review past conversations to find issues
• Documentation: Create reports from agent interactions
• Analysis: Export to CSV for data analysis
• Sharing: Easy to share formatted conversations
• Archival: Save important interactions

Implementation Plan

1. Parse agent logs/session data
2. Support multiple export formats
3. Add CLI interface
4. Optional: Add timestamp filtering, search

Would this be valuable for the project? Happy to implement it!

[cscguochang]

I'd like to work on this! I've already prototyped a CLI tool that exports to Markdown, JSON, CSV, and HTML. Pushing a PR shortly.

[alperents11]

Hi @Prachi1615

I've investigated the OM1 logging system and found:

• Logs are stored in logs/[agent]_[timestamp].log (when log_to_file=True)
• The main conversation loop is in src/runtime/single_mode/cortex.py::_tick()
• I see the prompt creation and LLM calls, but I'm not sure where conversation data (user input, LLM responses, actions) is being logged

Could you point me to where conversation exchanges are logged so I can properly integrate the exporter with the actual data?

Happy to update the PR once I understand the logging structure!