Sediman Browse

v0.3.6 Feature

This release adds 2 notable features for engineering teams evaluating rollout.

Published 16h AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Summary

AI summary

Updates What's Fixed, Agent Output Improvements, and Model Selection Persistence across a mixed release.

Changes in this release

Type	Severity	Summary	CVE
Feature	Low	Branding updated to Terminator/OpenSkynet in prompts while keeping Python folder as sediman Branding updated to Terminator/OpenSkynet in prompts while keeping Python folder as sediman Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix
Bugfix	Medium	Step count no longer shows previous task's steps between tasks Step count no longer shows previous task's steps between tasks Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Medium	Thinking tab labels correctly show "Expand" when collapsed and "Collapse" when expanded Thinking tab labels correctly show "Expand" when collapsed and "Collapse" when expanded Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Medium	Agent always produces non‑empty output with contextual fallbacks Agent always produces non‑empty output with contextual fallbacks Source: llm_adapter@2026-06-03 Confidence: high	—
Bugfix	Medium	Model selection now persists across restarts in ~/.terminator/tui.toml Model selection now persists across restarts in ~/.terminator/tui.toml Source: llm_adapter@2026-06-03 Confidence: low	—

Full changelog

Critical Fixes & Improvements

Model Selection Persistence (CRITICAL)

Your selected model (e.g., minimax m3) now persists across restarts
Provider and model are saved to ~/.terminator/tui.toml
CLI args can override saved config if needed

Step Count Display Fix

Fixed steps from previous task appearing in new task
Added agent_running gate to prevent step pollution between tasks
Clear streaming_phase state properly between tasks

Thinking Tab UI Fix

Fixed expand/collapse label text (was showing inverted actions)
Now correctly shows "Expand" when collapsed and "Collapse" when expanded
Press Space bar when input is empty to toggle

Agent Output Improvements

Agent always shows some output (no more empty responses)
Added contextual fallbacks for acknowledgments (good, ok, yes, no, etc.)
Improved conversational response handling with better prompts
System prompt updated to ensure engaging responses

Branding Updates

Updated branding to Terminator/OpenSkynet in prompts
TUI already correctly shows "Your Terminator" and "OpenSkynet"
Python folder remains as sediman (as requested)

Installation

curl -fsSL https://raw.githubusercontent.com/sediman-agent/OpenSkynet/main/scripts/install.sh | bash

What's Fixed

Model Selection Resetting (CRITICAL) - FIXED
Version Number Sync - FIXED
Step Count Showing Previous Steps - FIXED
Thinking Tab Expand/Collapse - FIXED
Empty Agent Responses - FIXED
Branding Consistency - FIXED

All critical UX issues have been addressed. The agent now behaves consistently
across multiple interactions with proper state management and branding.

Full Changelog: https://github.com/sediman-agent/OpenSkynet/compare/v0.3.5...v0.3.6

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track Sediman Browse

Get notified when new releases ship.

About Sediman Browse

All releases →

Related context

Related tools

Earlier breaking changes

v0.2.7 Renamed data directory from ~/.sediman to ~/.terminator across Python and Rust codebases