Extracts topics and atoms from Telegram messages for RAG context.
npx skills add https://github.com/djimontyp/task-tracker --skill llm-pipeline
Please help me install this Agent Skill. Skill: llm-pipeline Source: https://github.com/djimontyp/task-tracker/tree/main/.claude/skills/llm-pipeline Command: npx skills add https://github.com/djimontyp/task-tracker --skill llm-pipeline
This Skill solves the challenge of extracting meaningful, structured knowledge from unstructured message streams like Telegram chats, turning conversational data into actionable insights.
Use the llm-pipeline skill to analyze the latest 50 messages from our team channel and extract key topics and insights.
High-intent search queries and answers about installing and using this skill.
Extract structured knowledge from Telegram messages using Pydantic AI agents to identify topics and atomic knowledge units. The Skill automatically scores message importance, filters noise, and outputs ExtractionResult objects containing topics and atoms for downstream RAG retrieval.
Yes. This Skill transforms conversational data into searchable embeddings by extracting key insights, embedding content to 1536-dimension vectors, and integrating with SemanticSearchService. This enables RAG-enabled context construction for intelligent information retrieval across message archives.
Semantic search works by converting extracted topics and atoms into 1536-dimension embeddings, then using SemanticSearchService to match user queries against the embedded knowledge base. This retrieval-augmented generation approach returns contextually relevant results from your Telegram-derived content.
You need Pydantic AI, OpenAI or Ollama for LLM inference, and structured message data from Telegram. The Skill requires these dependencies to initialize agent pipelines, score messages, and generate ExtractionResult outputs with topics and atoms.
Yes. The Skill supports batch embedding and threshold-based knowledge orchestration across multiple Telegram sources. Process sets of messages, extract knowledge from each, embed results, and consolidate into a unified searchable knowledge base.
Topics are high-level subject categories identified in messages, while atoms are discrete, actionable knowledge units within those topics. Together they provide both conceptual structure and granular facts, enabling flexible querying and context construction for RAG retrieval.
rawvegRun and interact with local AI models via the Ollama API.
jg-chalk-ioBuild RAG pipelines with vector databases, embeddings, and LLM generation.
kernex-devDesign and implement RAG pipelines and LLM integrations for production AI systems.
xjtulycExtract topics, entities, claims, and trends from scientific literature corpora.
