Add cross-run memory system with SQLite+FTS5 storage#2
Merged
antonio2368 merged 8 commits intomasterfrom Mar 19, 2026
Merged
Conversation
Introduces a persistent memory system that recalls relevant context from previous runs and extracts new memories (decision, observation, summary, principle) post-run via LLM. Includes TUI management screen (M key), headless support (--no-memory), configurable recall/extraction, resume awareness, and per-run artifacts (_recalled_memories.md, _memories.json). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Introduce a separate info_modal (green) for success messages like config save, replacing the misleading red error_modal. Consolidate duplicate draw_error_modal functions from home.rs and pipeline.rs into a single draw_modal in screen::mod.rs rendered globally on every screen. Hide cursors when any modal is shown to prevent bleed-through. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds a Memory section to the edit popup (Tab-accessible alongside Agents and Timeouts) allowing users to view and modify all memory config fields at runtime. Toggle fields (enabled, disable_extraction) use Space; numeric/string fields use inline text editing. Session overrides are tracked separately and merged on save, with the save handler now returning the exact saved config to eliminate races. Also updates agent rename/remove to follow extraction_agent references, locks the popup during in-flight saves, and adds comprehensive tests. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
antaljanosbenjamin
approved these changes
Mar 18, 2026
…gement UI - Add recall tracking: memories record how often and when they were last recalled, with two-phase commit to avoid false positives on setup failure - Rebuild FTS5 index with Porter stemming for better recall quality; hyphenated terms split into segments with prefix wildcards for longer tokens - Add schema migration system (v0→v5) with transactional DDL migrations - Memory screen: bulk delete (D, two-press confirm), "never recalled" filter (r), DB size display, recall count/last-recalled in detail view - Unify extraction agent resolution across TUI, headless, and home screen - Headless: structured JSON warnings for memory store errors Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Previously clear_history() was only called between iterations but not between loop pass advances (loopback), causing providers to accumulate conversation history across all loop passes even with keep_session=false. Also fix pre-existing clippy::useless_format lint in memory/store.rs and update pre-commit hook to match CI flags (--all-targets --all-features). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…earing The keep_session_across_iterations flag is now iteration-scoped only, as originally designed. A new keep_across_loop_passes flag (default: true) independently controls whether provider history is cleared between loop pass advances within a single iteration. UI updated with two toggle columns (Iter/Loop) navigable with h/l keys. Help screen and README updated to document the new setting. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add archived column (schema v6) with auto-archival of stale permanent memories (decisions/principles) after configurable days without recall - Prioritize finalization/consolidation files in extraction prompt budget - Return extraction count from headless mode for better reporting - Increase default max_recall (15→20), max_recall_bytes (8K→16K), observation_ttl_days (90→120) - Add archived view toggle (a) and unarchive (u) keys to memory management - Auto-unarchive decisions when re-extracted as duplicates Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Introduces a persistent memory system that recalls relevant context from previous runs and extracts new memories (decision, observation, summary, principle) post-run via LLM. Includes TUI management screen (M key), headless support (--no-memory), configurable recall/extraction, resume awareness, and per-run artifacts (_recalled_memories.md, _memories.json).