Jump to content
Main menu
Main menu
move to sidebar
hide
Navigation
Main page
Recent changes
Random page
freem
Search
Search
Appearance
Create account
Log in
Personal tools
Create account
Log in
Pages for logged out editors
learn more
Contributions
Talk
Editing
Openai/691c325a-b61c-8003-b613-87e7ee20e0bf
(section)
Add languages
Page
Discussion
English
Read
Edit
Edit source
View history
Tools
Tools
move to sidebar
hide
Actions
Read
Edit
Edit source
View history
General
What links here
Related changes
Special pages
Page information
Appearance
move to sidebar
hide
Warning:
You are not logged in. Your IP address will be publicly visible if you make any edits. If you
log in
or
create an account
, your edits will be attributed to your username, along with other benefits.
Anti-spam check. Do
not
fill this in!
==== ผมใส่เป็น mermaid sequence diagram ให้ดูภาพรวม (อ่านง่าย) — agent = Claude, MemoryLayer = mem0 API, VectorDB = Weaviate, Encoder = embedding model: ==== <syntaxhighlight lang="mermaid">sequenceDiagram participant User participant Claude as Agent (Claude) participant MemoryLayer as Memory API (mem0 / custom) participant Encoder as Embedding service participant Weaviate as Vector DB User->>Claude: ส่งข้อความ/คำสั่ง Claude->>MemoryLayer: query_memory(user_id, query/intent) MemoryLayer->>Weaviate: vector_search(query_embedding, filters) Weaviate-->>MemoryLayer: return top-k memory objects MemoryLayer-->>Claude: return relevant memories (plus metadata) Claude->>Agent: build response using retrieved memories Note over Claude,MemoryLayer: During or after response, Claude may emit "insight" Claude->>MemoryLayer: write_memory({text, metadata, importance}) MemoryLayer->>Encoder: embed(text) Encoder-->>MemoryLayer: embedding MemoryLayer->>Weaviate: store(object, embedding, schema fields) Weaviate-->>MemoryLayer: ack MemoryLayer-->>Claude: ack/write result </syntaxhighlight> จุดสำคัญทางออกแบบ * What to store: raw transcripts, extracted facts, user preferences, action history, summaries. * How to store: store both text, summary (condensed), embedding, timestamp, source, ttl/priority. * Retrieval policy: combine semantic similarity + structured filters (recency, importance) เพื่อควบคุม context size. Weaviate สามารถทำ hybrid search/filters. docs.weaviate.io<ref>{{cite web|title=docs.weaviate.io|url=https://docs.weaviate.io/integrations/llm-agent-frameworks/mem0|publisher=docs.weaviate.io|access-date=2025-11-19}}</ref> * Summarization & consolidation: เมื่อ memory ยาวเกิน ควรสรุป/merge objects เพื่อลด noise และ save tokens. * Access control & privacy: บังคับนโยบายว่า memory บางประเภท (PII, sensitive) ไม่ควรถูกส่งกลับไปยัง LLM หรือถูกเก็บโดยไม่ได้ consent. * MCP compatibility: ถ้าทำให้ Claude/other tools ใช้ memory ร่วมกัน ให้ expose MCP-style endpoints / real-time API (ชุมชนมีการทดสอบการแชร์ memory ผ่าน MCP). Reddit<ref>{{cite web|title=Reddit|url=https://www.reddit.com/r/claude/comments/1owx8ry/testing_a_shared_longterm_memory_layer_for_claude/|publisher=reddit.com|access-date=2025-11-19}}</ref> (อ้างอิงหลัก: JP Hwang โพสต์การทดลอง และเอกสาร/ตัวอย่าง Weaviate + mem0 ที่ชุมชนใช้อ้างอิง). LinkedIn<ref>{{cite web|title=LinkedIn|url=https://www.linkedin.com/posts/jphwang_i-tried-building-an-agentic-memory-layer-activity-7395784323339243520-p4XB|publisher=linkedin.com|access-date=2025-11-19}}</ref>
Summary:
Please note that all contributions to freem are considered to be released under the Creative Commons Attribution-ShareAlike 4.0 (see
Freem:Copyrights
for details). If you do not want your writing to be edited mercilessly and redistributed at will, then do not submit it here.
You are also promising us that you wrote this yourself, or copied it from a public domain or similar free resource.
Do not submit copyrighted work without permission!
Cancel
Editing help
(opens in new window)