← 返回 Skills 市场

GitHub Semantic Search

Name: GitHub Semantic Search
Author: ultracold-molecule

作者 Ultracold-molecule · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install github-semantic

功能描述

AI-Native GitHub Assistant powered by Embedder+Qdrant+LLM architecture. Index repos, semantic search across issues/PRs/code, proactive monitoring with Feishu...

使用说明 (SKILL.md)

🦞 Super GitHub — AI-Native GitHub Assistant

Powered by the same Embedder + Qdrant + LLM architecture as elite memory systems. Index repos, search semantically, monitor proactively — all with natural language.

Architecture

Query → [LLM: understand intent] → [Embedder: vectorize] → [Qdrant: semantic search] → [gh CLI: act]

Three-layer system (same as production memory pipelines):

Layer	Component	Role
Embedder	Ollama `nomic-embed-text`	Converts text → 768-dim vectors
Vector Store	Qdrant (local)	Stores & searches vectors by similarity
Action Layer	`gh` CLI	Executes GitHub operations

Prerequisites

gh CLI authenticated (gh auth status)
Ollama running with nomic-embed-text:latest
Qdrant running at localhost:6333

Quick Start

# 1. Initialize Qdrant collection
python scripts/github_indexer.py init

# 2. Index a repo
python scripts/github_indexer.py add owner/repo --all

# 3. Search with natural language
python scripts/github_search.py "memory search failing in agent" --limit 10

# 4. Monitor for keywords
python scripts/github_monitor.py watch owner/repo --events issues,ci --keywords bug,broken,urgent

Scripts

Script	Purpose
`github_indexer.py`	Index repos (issues, PRs, metadata) into Qdrant
`github_search.py`	Natural language semantic search
`github_monitor.py`	Proactive monitoring with keyword alerts

Detailed Commands

Index (github_indexer.py)

python github_indexer.py init                    # Create Qdrant collection
python github_indexer.py add owner/repo --all     # Index everything
python github_indexer.py add owner/repo --issues # Issues only
python github_indexer.py add owner/repo --prs    # PRs only
python github_indexer.py add owner/repo --repo   # Repo metadata
python github_indexer.py status                   # Show indexed data
python github_indexer.py rm owner/repo           # Remove from index

Search (github_search.py)

python github_search.py "query"                            # Search all
python github_search.py "query" --repo owner/repo         # Filter by repo
python github_search.py "query" --type issue              # Filter by type
python github_search.py "query" --limit 20               # More results
python github_search.py "query" --repo owner/repo --ci    # Show CI runs

Monitor (github_monitor.py)

python github_monitor.py watch owner/repo                  # Start watching
python github_monitor.py watch owner/repo --events issues,ci
python github_monitor.py status                            # Show watches
python github_monitor.py check                            # Run checks
python github_monitor.py unwatch owner/repo               # Stop watching

Memory System Analogy

Component	GitHub Skill	Memory System
Data	Issues, PRs, code	Conversations
Embedder	nomic-embed-text	nomic-embed-text
Vector Store	Qdrant	Qdrant
Add	github_indexer.py	mem0 add
Search	github_search.py	mem0 search

Why Vector Search vs Keyword?

Approach	"memory problems" query
Keyword	Exact match only
Vector (this)	"memory leak", "OOM", "out of memory"

Setup Checklist

gh auth login — authenticate GitHub CLI
ollama pull nomic-embed-text:latest — download embedder
Start Qdrant: qdrant --storage-path ./qdrant-data
python github_indexer.py init — create collection

安全使用建议

This skill performs GitHub indexing, semantic search, and proactive monitoring and will use your local GitHub CLI auth and local Ollama/Qdrant services. Before installing or running it: (1) review and if needed change hard-coded paths (GH_EXE, OLLAMA_MODELS, STATE_FILE) so it matches your OS and does not write to unexpected locations; (2) understand that the monitor sends alerts via 'openclaw message send' to a hard-coded Feishu user id — check/replace that recipient and ensure you consent to sending repo content to Feishu; (3) confirm you want the skill to access your gh-authenticated account (it will read issues/PRs and CI data via the gh CLI, including private repo content if your gh session allows it); (4) run the scripts in an isolated environment or on a test repo first; and (5) ask the publisher to declare required binaries and any credentials explicitly (gh, Ollama running locally, Qdrant, openclaw) and to remove hard-coded IDs/paths for safer, cross-platform usage.

能力评估

ℹ Purpose & Capability

Functionality (indexing, semantic search, monitoring) matches the name/description and SKILL.md. However the package metadata declares no required binaries/env but the code requires the GitHub CLI, a local Ollama embeddings endpoint, Qdrant, and the openclaw CLI. Also the scripts hard-code Windows paths (GH_EXE and D:\ChatAI paths), which is an implementation detail that wasn't declared.

ℹ Instruction Scope

SKILL.md instructions cover index/search/monitor flows and list high-level prerequisites (gh, Ollama, Qdrant). The runtime scripts, however, also: write state to a specific local file (D:\ChatAI\OpenClaw\github_monitor_state.json), call a local embedding HTTP endpoint (http://localhost:11434), and invoke 'openclaw message send' to post Feishu alerts to a hard-coded user id. Those behaviors are consistent with monitoring but are not fully documented in the metadata and grant the skill the ability to transmit indexed content to an external chat channel.

✓ Install Mechanism

Instruction-only with included scripts; there is no install spec that downloads or extracts remote archives. Risk from install mechanism is low. The runtime requires external services (Ollama, Qdrant) and system CLIs which are not installed by the skill.

⚠ Credentials

No required env vars are declared, but the code will use the user's GitHub credentials via the gh CLI, the user's Ollama/Qdrant instances, and the openclaw CLI to send Feishu messages. The monitor will transmit alert content (issue/PR text, potentially private data) to a hard-coded Feishu user via openclaw. The lack of declared credentials/permissions and the hard-coded external recipient make privilege/credential access proportionality unclear.

✓ Persistence & Privilege

always:false and user-invocable are appropriate. The skill does create/modify a local state file under a hard-coded path, but it does not attempt to change other skills or global agent configuration.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install github-semantic
安装完成后，直接呼叫该 Skill 的名称或使用 /github-semantic 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Initial release: AI-native GitHub with Embedder+Qdrant+LLM architecture. Index repos, semantic search, proactive monitoring.

元数据

Slug github-semantic

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题