Repository navigation
feat(memory): add durable conversation memory - #3
Merged
Merged
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
T3 Code currently loses useful preferences, decisions, project facts, and failure lessons when a conversation ends. This adds server-owned memory that is enabled by default, recalls a bounded relevant context on later turns, and remains available across web, desktop, mobile, remote connections, and server restarts.
On first use, the server reviews existing completed conversations from oldest to newest, including archived and settled conversations, with one bounded transcript per chat. Its queue, cursor, retries, quarantined failures, and processing lease live in SQLite; generated note bodies use content-addressed Markdown files under T3 home. Daily consolidation and a deeper weekly review run when the environment is idle, with overdue work resuming after a stopped server starts again. The dedicated memory model is configurable and defaults to Codex
gpt-5.6-luna.Recall is injected at the shared provider boundary and exposed through scoped
memory_search,memory_read,memory_remember, andmemory_forgettools. Codex, Claude, and OpenCode run unattended generation in isolated no-tool sessions; unsupported providers fail explicitly. Settings expose environment and conversation controls on web/desktop and mobile.Fork migrations now use
Fork_NNN_*names and an independenteffect_sql_fork_migrationsledger. A compatibility bridge adopts the known earlier fork migration IDs without consuming later upstream IDs.Validation:
Implemented with GPT-5.6 Sol in Codex.