@루이스모네다 luismoneda lgmoneda #이맥스 #시맨틱검색

Building applications with LLMs through composability

contacts::Luis Moneda - lgmoneda - lgmoneda.github.io

관련링크

Luis Moneda 2024 “A messy version of an ai assistant in Emacs I’ve been iterating to find out what is useful.

(Luis Moneda 2024a)

A messy version of an ai assistant in Emacs I’ve been iterating to find out what is useful. Notes in https://lgmoneda.github.io/ - assistant-settings.el

Luis Moneda 2023 “A relevance revolution for knowledge-workers”

(Luis Moneda 2023a)

Cognition augmentation caused by systems based on Large Language Models will change collaborative knowledge work.

Luis Moneda 2024 “lgmoneda/dot-emacs”

(Luis Moneda [2017] 2024)

My Emacs config files

TODO Luis Moneda 2023 “Semantic Search for Org roam”

(Luis Moneda 2023b)

Creating a Semantic Search functionality to improve search, thinking, and writing with Org roam.

Moneda, Luis 2024 “Cognitive joy tooling #1: the Catalyst AI assistant, design review and 90-days retrospective”

(Luis Moneda 2024b)

My experience using a catalyst agent that engages me when I use Org and Org Roam Cognitive joy tooling #1

Summary

본 글은 저자가 개발한 Emacs 기반의 AI 어시스턴트 “Catalyst Assistant” 90일 사용 후기를 다룹니다. 주요 기능은 개인 노트(Org Roam), 할 일 목록, 일정 관리에 대한 심층적인 질의응답 및 연관 노트 제안입니다.

주요 내용은 다음과 같습니다:

1 세 가지 주요 기능:

1 의미 기반 검색 (Semantic Search): 4895개 노드에 대한 효과적인 검색 및 관련 노트 제시 2 질의응답 (Q&A): 사용빈도가 낮아 유용성이 낮다고 판단 3 Catalyst Assistant: 생각을 촉진하고 자기 성찰을 돕는 AI 어시스턴트. 개인의 일정, 할 일, Org Roam 노트 변경 사항을 모니터링하고, 관련 정보를 제공하며, 질문에 답하고, 글쓰기 과정에서 피드백을 제공하는 기능

2 Catalyst Assistant 구조: Org 파일 변경 로그를 모니터링하여 AI 모델에 입력하고, 프롬프트 기반으로 다양한 루틴을 실행합니다. (아침/저녁 루틴, 지속적 검토 등)

3 상호작용 방식: 채팅, 수동 루틴 실행, 사전 설정된 루틴을 통한 자동 피드백 등 세 가지 방식을 제공합니다.

4 장점: 자동 피드백 기능은 글쓰기 과정에서 유용하며, 생각을 정리하고 잊기 쉬운 부분을 상기시켜주는 데 효과적입니다.

5 단점: 연관 노트 제안 기능의 정확도가 낮고, 환각(hallucination) 현상이 간헐적으로 발생하며, 마크다운 포맷팅 및 메모리 기능에 문제가 있습니다.

6 결론: Semantic Search 만큼 중추적인 역할은 아니지만, 향후 개선을 통해 더욱 유용한 도구가 될 가능성을 보여줍니다. 향후 버전에서는 에이전트 기반 시스템으로 개선할 계획입니다.

Sharing Org Roam content | lgmoneda

(Luis Moneda 2025a)

Cognitive joy tooling #2: textual topography

(Luis Moneda 2025b) Cognitive joy tooling \#2 Moneda, Luis Profiling a text’s characteristics using embeddings and plots 2025

Cognitive joy tooling #3: Closing the creation-application gap of my knowledge base using auto-segmentation, colors, and semantic search

(Luis Moneda 2025c) Cognitive joy tooling \#3 Moneda, Luis Segmenting what I read or write to make me inspect my personal notes more often 2025

blog | lgmoneda

[2025-04-06 Sun 14:33]

Article

• 11 Jan 2025 » Sharing Org Roam content
• 04 Jan 2025 » Cognitive joy tooling #2: textual topography
• 19 Oct 2024 » Cognitive joy tooling #1: the Catalyst AI assistant, design review and 90-days retrospective
• 06 Jul 2024 » Invariance in Machine Learning #4: Invariant Random Forest
• 28 Oct 2023 » Spotting hallucination in LLMs using similarity variance
• 15 Apr 2023 » Q&A with Org roam using ChatGPT
• 08 Apr 2023 » Semantic Search for Org roam
• 25 Jan 2023 » A relevance revolution for knowledge-workers
• 28 Oct 2022 » On Project management for Data Science
• 06 Jul 2022 » A Random Time Robust Forest
• 13 Jun 2022 » Time Robust Tree experiments on real datasets
• 26 Mar 2022 » Continuous Performance Calibration
• 22 Mar 2022 » Ang Li’s example on the Unit Selection problem with counterfactual logic
• 12 Mar 2022 » The Manager-report expectations doc
• 23 Feb 2022 » Defining a starter project in Data Science
• 03 Dec 2021 » Introducing the Time Robust Tree - invariance in Machine Learning #3
• 27 May 2021 » Exploring invariance in Machine Learning #2: Invariant Risk Minimization
• 23 Feb 2021 » On “Why greatness cannot be planned” and how data science happens in the industry
• 21 Feb 2021 » Exploring invariance in Machine Learning #1: Invariant Causal Prediction
• 19 Feb 2021 » An introduction to Independent Causal Mechanisms
• 12 Jan 2021 » Spurious correlation, machine learning, and causality
• 07 Dec 2020 » Temporal feature selection with SHAP values
• 29 Feb 2020 » A sane internet experience
• 12 Jan 2020 » An one on one guide when talking to leadership
• 15 Oct 2019 » Book summary: The Power of Less
• 17 Feb 2019 » The pessimist and the optmistic view of bias in Machine Learning
• 01 Jun 2018 » Judea Pearl’s ladder of causation
• 06 Mar 2018 » Using Orange: a visual tool for Data Science!
• 21 Jan 2018 » Find yourself a job!
• 15 Mar 2017 » A brief introduction to Emacs Lisp for people with programming background
• 19 Feb 2017 » A solution for the reload modules problem in Emacs Python Shell
• 02 Apr 2016 » hello world!

(Luis Moneda [2017] 2024)

2023-07-10 Survey

[2023-07-10 Mon 09:56]

LangChain 관련 조사. 아래 lgmoneda 님의 블로그에서 엄청난 인사이트를 확인한 바 있다. 나의 투두 중에 하나이다. ¹

• #랭체인

조만간 아래 문서는 정리를 해서 올릴 예정이다.

(Luis Moneda 2023b)

• “Semantic Search for Org roam” Luis Moneda 2023
• Creating a Semantic Search functionality to improve search, thinking, and writing with Org roam.

몇 가지 관련 링크만 넣어둔다.

• LangChain JS/TS Doc ²
• LangChain + Org-mode ³
• LangChain Github Clone ⁴

caption=“<span class=“figure-number”>Figure 1: lgmoneda.github.io” width=“100%” >

Related-Notes

BIBLIOGRAPHY

Luis Moneda. 2023a. “A Relevance Revolution for Knowledge-Workers.” lgmoneda. January 25, 2023. http://lgmoneda.github.io/2023/01/25/relevance-revolution-knowledge-work.html.

———. 2023b. “Semantic Search for Org Roam.” lgmoneda. April 8, 2023. http://lgmoneda.github.io/2023/04/08/semantic-search-for-org-roam.html.

———. 2024a. “A Messy Version of an Ai Assistant in Emacs I’ve Been Iterating to Find out What Is Useful. Notes in https://lgmoneda.github.io/.” Gist. 2024. https://gist.github.com/lgmoneda/6764fd712660b9bd0eb9870f66778ee6.

———. (2017) 2024. “Lgmoneda/Dot-Emacs.” https://github.com/lgmoneda/dot-emacs.

———. 2024b. “Cognitive Joy Tooling #1: The Catalyst Ai Assistant, Design Review and 90-Days Retrospective.” lgmoneda. October 19, 2024. https://lgmoneda.github.io/2024/10/19/catalyst-assistant-90-days.html.

———. 2025a. “Sharing Org Roam Content | Lgmoneda.” 2025. http://lgmoneda.github.io/2025/01/11/sharing-org-roam-content.html.

———. 2025b. “Cognitive Joy Tooling #2: Textual Topography.” lgmoneda. January 4, 2025. https://lgmoneda.github.io/2025/01/04/intro-textual-topography.html.

———. 2025c. “Cognitive Joy Tooling #3: Closing the Creation-Application Gap of My Knowledge Base Using Auto-Segmentation, Colors, and Semantic Search.” lgmoneda. February 9, 2025. https://lgmoneda.github.io/2025/02/09/cognitive-joy-tooling-3-closing-the-creation-application-gap-of-my-knowledge-base-using-auto-segmentation-colors-and-semantic-search.html.

Footnotes

1. http://lgmoneda.github.io/2023/04/08/semantic-search-for-org-roam.html ↩

2. https://js.langchain.com/docs ↩

3. https://python.langchain.com/docs/modules/data_connection/document_loaders/integrations/org_mode ↩

4. https://github.com/junghan0611/langchainjs ↩