redmine_search

0

Описание

A modern semantic search tool for your Redmine wikis and issues, powered by local AI embeddings.

Языки

  • Python100%
10 месяцев назад
10 месяцев назад
10 месяцев назад
10 месяцев назад
10 месяцев назад
README.md

Redmine Semantic Search (AI-powered, Multi-Project)

A modern semantic search tool for your Redmine wikis and issues, powered by local AI embeddings.
Features include:

  • Fast semantic (meaning-based) search across wiki pages and issues
  • Multi-project support (switch projects in the web UI)
  • Embedding and index caching for fast re-runs
  • Streamlit web UI, easy to use
  • Cache clear button for any project/model

Requirements

  • Python 3.8+
  • uv package manager (for fast and reproducible installs)
  • Redmine with API access enabled

Quick Start

  1. Clone the repo and move in:

  2. Copy and edit the sample configuration:

    Then edit

    config.yaml
    and set your Redmine URL, API key, and project(s).

  3. Install dependencies with

    uv
    :

  4. Run the app:

  5. Open your browser to

    http://localhost:8501


Configuration

Example

config.yaml
for multiple projects:

  • redmine_url
    : Root URL of your Redmine instance
  • api_key
    : Your Redmine user API key
  • projects
    : List of project identifiers to make searchable
  • count
    : How many results to display per search (default 5)
  • language
    : Embedding model (see SBERT models)

Features

  • AI-based semantic search: Find the most relevant Redmine wiki pages and issues for your query
  • Multi-project support: Select any configured project in the web UI
  • Fast: Embeddings and index are cached for instant startup after the first run
  • Cache clear button: In the UI, clear cache for the current project/model and force a refresh if Redmine data changed
  • No OpenAI API needed: Runs fully locally and privately

Data & Caching

  • All embedding and index files are stored in the
    data/
    directory.
  • To clear cache for a specific project/model, use the UI "Clear cache" button.
  • To clear all cache manually, simply delete the
    data/
    directory.

Troubleshooting

  • If you update Redmine data and want to force a re-embed, use the cache clear button or remove the relevant files in
    data/
    .
  • If you see warnings about PyTorch or Streamlit watcher, they are safe to ignore as long as the UI works.

Credits


License

MIT


Sample
requirements.txt
for use with
uv