AsterMem
AsterMem is a self-hosted personal memory service that uniformly stores Markdown, SQLite, keyword index and vector index in its own lazy cat micro service, and provides the same long-term memory to Web UI, AI Agents and MCP clients.
Lazy cat OIDC login has been integrated. ##Functional characteristics - Use Markdown + SQLite to save personal memories and support creation, editing, archiving, recovery, labeling and priority. - Use Whoosh keyword search and have Chroma document-level and paragraph-level semantic retrieval links; the Embedding Provider will immediately refresh the vector runtime after saving it. - Provide Explore (Memory Roaming), personal portrait, usage, API log, vector status and security management pages;2.0.2 New image retest has been completed, and see "Real Return after Deployment" for current blocking items. - Built-in multiple model Provider configurations, and you can access local LM Studio / Ollama or external services such as OpenAI, DeepSeek, and Bailian. - Provides scoped API tokens, Agent tool entrances and true Streamable HTTP MCP endpoints. ###First use process 1. Open the app and log in using your lazy cat account. 2. Enter "Settings" and select the Dialogue Provider; if you need semantic search, select the Embedding Provider. 3. Enter "Hand it to AI" and click "I will write it myself this time" or "New memory" to create the first memory. It is recommended to include a clear date for the first time. For example, fill in the title "West Lake Night Run and Supply Arrangements", and fill in the text: `Plan to start from Broken Bridge at 19:30 pm on Saturday, October 17, 2026, and complete an 8-kilometer jog along Beishan Street.` 4. Open "Memory" and verify that the data you just created is visible. Segments, tags, and time events are generated by background tasks; do not judge the failure immediately after submitting or clicking "Re-segment", wait for about one minute before refreshing. 5. Enter "Atlas" → "Timeline". The default range covers the past 30 days to the future 365 days, and this memory should be seen in `2026-10-17`; the "Saturday" after the specific date will not be broken into another date. ###Daily usage 1. View, search, edit, archive and recover memories in Memory. After modifying the date, plan or the entire paragraph, enter the details and click "Re-segment", wait for the background task to complete before viewing the map. 2. Natural language issues are raised in Exploration; session metadata is written to backend SQLite and can be read from other clients. 3. Import ZIP, JSON, pictures or pasted text in Import/Export, or export a full portable ZIP. 4. View associations, Timeline, and vector visualizations in Atlas. The Timeline displays the next year by default, and the start and end dates can be adjusted; check the event to mark completion, and click the event to return to the source memory. 5. If two events with the same title and content appear in the Timeline, go to "Memory" first to confirm whether there are two duplicate source memories with different IDs. Memories from different sources generate events separately and do not belong to repeated writing in the same segment. 6. Rename, merge, and organize tags in Tags. 7. Maintain personal data in "Portrait" and view model calls and API activities in "Usage" and "Log". 8. Manage local accounts, API Tokens, Login Protection, and factory resets in "Management". 9. AI agents can share the same memory through `/api/agent/call` or `/mcp`.
