graphiti-memory
npx machina-cli add skill ggozad/haiku.skills/graphiti-memory --openclawGraphiti Memory
Use the remember tool to store facts, observations, and context into the knowledge graph. Use the recall tool to search the knowledge graph for relevant memories. Use the forget tool to remove outdated or incorrect memories.
Workflow
- Use
rememberto store important information as it comes up. - Use
recallto retrieve relevant context before answering questions. - Use
forgetto clean up memories that are no longer accurate.
Guidelines
- Store information that might be useful for future conversations.
- Be specific when recalling — use targeted queries rather than broad ones.
Source
git clone https://github.com/ggozad/haiku.skills/blob/main/skills/graphiti-memory/haiku_skills_graphiti_memory/graphiti-memory/SKILL.mdView on GitHub Overview
Graphiti Memory lets you persist important facts, observations, and context as memories in a knowledge graph. It uses remember to store, recall to retrieve, and forget to prune outdated data. This keeps conversations coherent and context-aware across sessions.
How This Skill Works
Memories are created by the remember tool as you encounter information. Before answering, use recall to fetch relevant memories with targeted queries, then answer with context. When memories become wrong or outdated, use forget to remove them.
When to Use It
- Before answering a question, to pull in related memories for accuracy and context.
- To maintain long term context across multiple conversations or sessions.
- When you need to attach or cite facts to support a response with sources from memory.
- To prune or update memories after new information changes the context.
- For complex tasks that rely on evolving domain knowledge beyond a single chat.
Quick Start
- Step 1: Remember to store important information as it comes up.
- Step 2: Recall to fetch relevant memories before answering.
- Step 3: Forget to remove memories that are outdated or incorrect.
Best Practices
- Capture precise facts with metadata such as source and timestamp.
- Use targeted recall queries to minimize irrelevant results.
- Update or prune memories regularly with the forget tool as information changes.
- Avoid duplicate memories by checking for existing entries before remember.
- Periodically audit the knowledge graph for accuracy and completeness.
Example Use Cases
- Remember a user's language preference to tailor tone and style.
- Recall prior product details during a customer support chat to avoid repetition.
- Store event dates or deadlines so they are recalled when scheduling.
- Recall a user's location or preferences to personalize recommendations.
- Forget outdated pricing or policy details after updates.