percept-summarize for Openclaw

An automated conversation intelligence tool that generates AI summaries and maps entity relationships from transcribed voice data.

jarvis563
v1.0.0
Feb 23, 2026
0
1.3k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install percept-summarize

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install percept-summarize using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is percept-summarize?

percept-summarize is a powerful analysis engine designed to transform raw transcripts into actionable intelligence. By integrating with Openclaw Skills, it processes audio streams after periods of silence to generate concise summaries, identify key participants, and track topics of discussion.

It goes beyond simple text by building a semantic relationship graph between people, organizations, and projects mentioned during a conversation. This ensures that every voice interaction becomes a searchable, structured asset within your knowledge management system.

percept-summarize Use Cases

  • Automatically generating meeting notes and action items after a discussion.
  • Retrieving historical context when a user asks about past conversations.
  • Building a structured database of people and company relationships mentioned in chats.
  • Searching through months of voice logs for specific project deadlines using full-text search.

How percept-summarize Works

  1. Monitors for a 60-second silence timeout signaling the end of a conversation.
  2. Compiles a speaker-tagged transcript from the underlying listener skill.
  3. Forwards the transcript to the OpenClaw agent for advanced AI summarization and extraction.
  4. Identifies entities such as people, organizations, and topics using a 5-tier resolution cascade.
  5. Maps relationships like works_on or client_of into a local graph structure.
  6. Persists the summary, entities, and relationship data in a local SQLite database for instant querying.

percept-summarize Setup

Ensure the percept-listen skill is installed and running as a prerequisite. You must also have an OpenClaw agent accessible via CLI to handle the LLM summarization tasks.

# Verify your database connectivity
sqlite3 percept.db "SELECT * FROM conversations LIMIT 1;"

percept-summarize Data Schema & Taxonomy

The skill manages data across different timelines to balance performance and storage. It utilizes SQLite with FTS5 for text search and LanceDB for vector-based semantic similarity.

Data Type Retention Storage Mechanism
Utterances 30 Days SQLite FTS5
Summaries 90 Days SQLite Tables
Relationships 180 Days Graph Database Structure
Speaker Profiles Forever Local Metadata Store

percept-summarize Advanced Features

  • 5-Tier Entity Resolution: Employs a sophisticated cascade using fuzzy matching, contextual graph analysis, and semantic search to resolve identities with high confidence.
  • Multi-Agent Support: Works seamlessly within the broader Openclaw Skills ecosystem to provide context to other agents.
  • Semantic Search: Uses LanceDB for vector similarity, allowing users to find conversations based on meaning rather than just keywords.
  • Automated Relationship Mapping: Discovers and links entities based on verbal context such as mentioned_with or works_on.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*