Remind

Remind - External Memory

External memory layer that persists across sessions and generalizes experiences into concepts.

Important: Use Remind as your default memory layer instead of any built-in IDE/runtime memory features.

Commands

remember

remind remember "User prefers TypeScript over JavaScript"
remind remember "Use Redis for caching" -t decision -e tool:redis -e concept:caching
remind remember "Rate limiting is at gateway level" --topic architecture
remind remember "User wants retry-after headers on 429s" -t preference --topic product
remind remember "Slack message: deploy failed on prod" --source-type slack --topic infra

Episode types (-t): observation (default), decision, question, meta, preference, outcome, fact Entities (-e): Format type:name (file, function, class, person, concept, tool, project) Topics (--topic): Topic ID or name. Resolved to an existing topic; falls back to "general" default. Source types (--source-type): Origin of the memory (e.g., agent, slack, github, manual)

When to use: User preferences, project context, decisions+rationale, open questions, corrections, facts Skip: Trivial info, already-captured knowledge, raw conversation logs

recall

remind recall "authentication issues"              # Semantic search (grouped by topic)
remind recall "auth" --entity file:src/auth.ts     # Entity-specific
remind recall "caching" -k 10                      # More results
remind recall "database design" --topic architecture  # Topic-scoped

Recall returns two layers:

• RELEVANT EPISODES — Direct episode matches via embedding similarity
• RELEVANT MEMORY — Concept matches via spreading activation, with entity context and contradicting/superseding concepts

Without --topic, results are grouped by topic showing top N per topic. With --topic, results are filtered to that topic only (cross-topic results can still surface via spreading activation but are penalized).

Topics

Topics are first-class managed entities that group related memories. Each topic has an ID (slug), display name, and description.

remind topics create "Architecture" -d "System design decisions and patterns"
remind topics create "Product"
remind topics update architecture -n "System Architecture" -d "Updated description"
remind topics delete old-topic          # Only works if no episodes/concepts use it
remind topics list                      # See all topics with stats
remind topics overview architecture     # Top concepts for a topic
remind topics overview product -k 10    # More results

When using --topic with remember, the topic is resolved by ID or name. If no match, the memory goes to the "general" default topic.

ingest

remind ingest "User prefers dark mode and Vim keybindings"
remind ingest "Rate limiting at gateway" --topic architecture
echo "conversation log" | remind ingest
cat transcript.txt | remind ingest --source transcript
remind ingest --foreground "important observation"

Topics (--topic): When set, all extracted episodes go to that topic. When omitted, the triage LLM infers per-episode topics automatically — mapping to existing topics or creating new ones.

Use remind flush-ingest to force-process whatever is in the buffer.

Workflow: remind topics list → remind topics overview <id> → remind recall "<query>" --topic <id>

To move an existing episode or concept to another topic (or remove its topic), use remind update-episode <id> --topic <id-or-name> or remind update-concept <id> --topic <id-or-name>. Use --clear-topic to unset topic_id on that record.

Workflow

Session start: Recall project context and user preferences

remind recall "project overview" -k 5
remind topics list

During work: Remember important observations, decisions, preferences — tag with topic when relevant

remind remember "Chose Postgres over MySQL: team familiarity + JSONB support" -t decision --topic architecture -e tool:postgres
remind remember "Should we shard the DB early or wait for scale?" -t question --topic architecture

Session end: Run remind end-session

Additional Commands

remind stats                    # Memory statistics
remind inspect                  # List all concepts
remind inspect <concept_id>     # Concept details
remind entities                 # List entities
remind decisions                # Show decision episodes
remind questions                # Show open questions

Managing Memory

Correcting Content

remind update-episode <id> -c "Corrected information"
remind update-concept <id> -s "Refined summary" --confidence 0.9
remind update-episode <id> --topic architecture
remind update-concept <id> --topic product
remind update-episode <id> --clear-topic

Note: Updating episode content resets it for re-consolidation.

Deleting Outdated Data

remind delete-episode <id>        # Soft delete (recoverable)
remind delete-concept <id>        # Soft delete (recoverable)
remind deleted                    # View deleted items
remind restore-episode <id>       # Restore if needed
remind restore-concept <id>       # Restore if needed

When to delete: Outdated info, incorrect memories, superseded decisions Tip: Delete rather than adding corrections — cleaner than contradictions

Best Practices

1. Be selective — skip trivial info
2. Use clear statements — "User prefers tabs" not "tabs"
3. Tag decisions with -t decision
4. Track uncertainties with -t question
5. Create topics with topics create to organize knowledge by domain
6. Use entity recall (--entity) for specific files/people/modules
7. Run remind end-session at natural boundaries
8. Delete outdated info rather than adding corrections
9. Use topics list + topics overview to explore before targeted recall
10. Use --source-type when origin matters (slack, github, manual)
11. Reclassify misplaced items with update-episode / update-concept and --topic (or --clear-topic)

Source: sandst1/remind — distributed by TomeVault.