feat: Collapse Compaction maintains agent intelligence beyond compaction by preserving recent interactions and curating knowledge collapsing the older stale messages

[ryanwyler]

Summary

Adds a new "collapse" compaction mode that preserves recent context while summarizing older messages.

The problem: Standard compaction compresses the entire conversation into a tiny summary, causing massive context loss (100% -> 3%). The AI forgets everything.

The solution: Collapse only compresses the OLD, stale parts while keeping current work pristine:

• Extracts oldest 65% of tokens and summarizes them at a breakpoint
• Uses newest 15% as reference context for the summary LLM
• Leaves 35% of recent conversation completely untouched
• Result: Compresses 100% context down to 15-25%, leaving 75% remaining context window while preserving all critical knowledge

Changes

• packages/opencode/src/session/compaction.ts - Main collapse logic with processCollapse(), token estimation, and markdown conversion
• packages/opencode/src/id/id.ts - Add createLike() for inserting messages at past timestamps, detectFormat() for backward compatibility
• packages/opencode/src/session/message-v2.ts - Update filterCompacted() for breakpoint detection
• packages/opencode/src/config/config.ts - New config options (method, trigger, extractRatio, recentRatio)
• packages/opencode/src/cli/cmd/tui/routes/session/ - TUI toggle and sidebar display
• packages/opencode/src/server/server.ts - Unified flow for both modes
• packages/opencode/src/cli/cmd/tui/ui/link.tsx - Fix underline prop error from fix(tui): make auth URLs clickable regardless of line wrapping #6317

Configuration

{
  "compaction": {
    "method": "collapse",
    "trigger": 0.85,
    "extractRatio": 0.65,
    "recentRatio": 0.15
  }
}

Default remains "standard" - users can opt-in to collapse via config or TUI toggle (Ctrl+P -> "Use collapse compaction").

Testing

1. Set compaction.method to "collapse" in config or toggle via TUI
2. Have a long conversation that triggers context overflow
3. Verify collapse message appears at breakpoint position
4. Verify recent messages remain untouched
5. Verify no infinite compaction loops

[ariane-emory]

Not really in favour of changing the defaults, new behaviour ought be opt-in IMO.

[ryanwyler]

Not really in favour of changing the defaults, new behaviour ought be opt-in IMO.

Fixed it. Thank you for the feedback.

[ryanwyler]

Before Compaction: 170,424 Tokens
After Compaction: 67,289 Tokens

60% Reduction in token consumption and 100% MEMORY RETENTION

After collapse compact, AI Agent able to produce pages and pages of details about the project on many topics without the use of any tools and no reading of documents.