{"id":49558082,"url":"https://github.com/thezupzup/nova","last_synced_at":"2026-05-03T06:17:34.846Z","repository":{"id":353464574,"uuid":"1219453630","full_name":"TheZupZup/Nova","owner":"TheZupZup","description":"A self-hosted AI assistant with intelligent model routing","archived":false,"fork":false,"pushed_at":"2026-04-24T03:07:07.000Z","size":31,"stargazers_count":0,"open_issues_count":6,"forks_count":1,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-24T03:26:37.265Z","etag":null,"topics":["ai","amd-gpu","llm","local-llm","ollama","open-source","privacy-first","rocm","self-hosted","web-interface"],"latest_commit_sha":null,"homepage":"","language":"HTML","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/TheZupZup.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-23T22:28:27.000Z","updated_at":"2026-04-24T03:07:10.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/TheZupZup/Nova","commit_stats":null,"previous_names":["thezupzup/nova"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/TheZupZup/Nova","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheZupZup%2FNova","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheZupZup%2FNova/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheZupZup%2FNova/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheZupZup%2FNova/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/TheZupZup","download_url":"https://codeload.github.com/TheZupZup/Nova/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheZupZup%2FNova/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32559760,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-03T03:21:47.309Z","status":"ssl_error","status_checked_at":"2026-05-03T03:21:43.884Z","response_time":103,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai","amd-gpu","llm","local-llm","ollama","open-source","privacy-first","rocm","self-hosted","web-interface"],"created_at":"2026-05-03T06:17:33.558Z","updated_at":"2026-05-03T06:17:34.837Z","avatar_url":"https://github.com/TheZupZup.png","language":"HTML","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Nova\n\nA self-hosted, local AI assistant built on FastAPI and Ollama.\n\n## What Nova is\n\nNova is a personal AI assistant that runs entirely on your own machine. It routes each conversation to the most appropriate local model based on complexity, maintains a persistent memory database across sessions, and serves a web interface reachable from any browser on your network.\n\nNova is under active development. Core features are functional. Some subsystems — the natural language memory retriever and the embedding pipeline — are present in the codebase but are not yet fully validated for production use.\n\n## What works today\n\n- Multi-model routing: a lightweight classifier (`gemma3:1b`) decides which model handles each request\n- Persistent memory stored in SQLite with automatic extraction from conversations\n- Manual memory commands for explicit fact storage (`Retiens ça:`, `Souviens-toi:`, `Souviens-toi de ça:`)\n- JWT-secured web interface accessible from desktop and mobile browsers on your network\n- Conversation history with a sidebar for navigation\n- Mode selector: Auto / Chat / Code / Deep\n- Real-time weather via Open-Meteo (no API key required)\n- Web search via DuckDuckGo (manual trigger)\n- Settings panel: view, add, edit, and delete stored memories; RAM budget control\n- Login rate limiting (5 attempts per 60-second window, configurable via environment variables)\n- Nova identity contract: Nova presents itself as a named assistant rather than exposing the underlying model name\n- Background RSS/web learning (disabled by default, opt-in via `NOVA_AUTO_WEB_LEARNING=true`)\n- AMD GPU acceleration via ROCm; falls back to CPU automatically\n- systemd service configuration for unattended startup\n\n## Privacy and local-first principles\n\n- Model inference runs locally through Ollama. Optional tools such as web search and weather can contact external services only when explicitly triggered.\n- The memory database is a local SQLite file under your control.\n- Credentials live in `.env` and are never committed to the repository.\n- No telemetry, no cloud sync, no third-party analytics.\n\n## Architecture overview\n\n```\nweb.py              FastAPI application and REST endpoints\nmain.py             Terminal interface (no web server required)\nconfig.py           Central configuration loaded from .env\n\ncore/\n  router.py         Model selection via gemma3:1b classifier\n  chat.py           Conversation logic and message assembly\n  memory.py         SQLite memory: facts, conversations, settings\n  memory_command.py Manual memory command parser\n  identity.py       Nova identity contract injected into the system prompt\n  auth.py           JWT creation and verification\n  rate_limiter.py   Per-IP sliding-window rate limiter for the login endpoint\n  learner.py        Background RSS feed ingestion (opt-in)\n  weather.py        Open-Meteo integration\n  search.py         DuckDuckGo integration\n  updater.py        Model version management\n\nmemory/\n  store.py          Natural language memory store\n  retriever.py      Semantic memory retrieval\n  extractor.py      Memory extraction pipeline\n  schema.py         Memory data schema\n  policy.py         Retention and cleanup policy\n\nstatic/\n  index.html        Web interface\n\ntests/              Pytest test suite\n```\n\n## Getting started\n\n### Requirements\n\n- Linux (tested on Fedora)\n- Python 3.11+\n- [Ollama](https://ollama.com) installed and running\n- AMD GPU with ROCm (optional — falls back to CPU)\n\n### 1. Clone the repository\n\n```bash\ngit clone https://github.com/TheZupZup/Nova.git\ncd Nova\n```\n\n### 2. Create a virtual environment and install dependencies\n\n```bash\npython -m venv .venv\nsource .venv/bin/activate\npip install -r requirements.txt\n```\n\n### 3. Pull the required Ollama models\n\n```bash\nollama pull gemma3:1b\nollama pull gemma4\nollama pull deepseek-coder-v2\nollama pull qwen2.5:32b\n```\n\n`qwen2.5:32b` requires significant disk space and RAM. If your hardware is constrained, you can skip it; the router falls back to `gemma4` for advanced requests.\n\n### 4. Configure credentials\n\n```bash\ncp .env.example .env\n```\n\nEdit `.env` and set at minimum:\n\n```\nNOVA_USERNAME=your_username\nNOVA_PASSWORD=your_password\nNOVA_SECRET_KEY=a-long-random-string\n```\n\nThe defaults in `.env.example` are intentionally weak placeholders. Change them before any deployment, especially if Nova is exposed beyond localhost.\n\n### 5. Start Nova\n\n```bash\npython web.py\n```\n\nNova is available at `http://localhost:8080`.\n\n## Running as a systemd service\n\nCreate `/etc/systemd/system/nova.service`:\n\n```ini\n[Unit]\nDescription=Nova AI Assistant\nAfter=network.target ollama.service\n\n[Service]\nType=simple\nUser=yourusername\nWorkingDirectory=/path/to/Nova\nExecStart=/path/to/Nova/.venv/bin/uvicorn web:app --host 0.0.0.0 --port 8080\nRestart=always\nRestartSec=5\nEnvironment=\"PATH=/path/to/Nova/.venv/bin:/usr/bin:/usr/local/bin\"\n\n[Install]\nWantedBy=multi-user.target\n```\n\n```bash\nsudo systemctl daemon-reload\nsudo systemctl enable nova\nsudo systemctl start nova\n```\n\n## Configuration\n\nAll configuration is read from `.env` at startup. Key variables:\n\n| Variable | Default | Description |\n|---|---|---|\n| `NOVA_USERNAME` | — | Login username |\n| `NOVA_PASSWORD` | — | Login password |\n| `NOVA_SECRET_KEY` | — | JWT signing secret |\n| `OLLAMA_HOST` | `http://localhost:11434` | Ollama API base URL |\n| `NOVA_AUTO_WEB_LEARNING` | `false` | Enable background RSS/web learning |\n| `LOGIN_RATE_LIMIT_MAX` | `5` | Max login attempts per window |\n| `LOGIN_RATE_LIMIT_WINDOW` | `60` | Rate limit window in seconds (sliding) |\n| `LOGIN_RATE_LIMIT_TRUSTED_PROXIES` | — | Comma-separated proxy IPs to trust for `X-Forwarded-For` |\n\nModel assignments are defined in `config.py` in the `MODELS` dictionary. To swap a model, update that dictionary and restart Nova.\n\nA note on language: the default system prompt and Nova's persona are written in French. Nova auto-detects the language of each message and replies in kind, so English conversations work without any configuration change.\n\n## Development workflow\n\n```bash\n# Run the full test suite\npytest\n\n# Run a specific test file with verbose output\npytest tests/test_router.py -v\n```\n\nThe test suite covers model routing, memory storage and parsing, manual memory commands, rate limiting, the identity contract, and weather integration.\n\nA `Dockerfile` is included in the repository for containerised setups. It is not the primary supported deployment method at this stage.\n\n## Contributing\n\nSee [CONTRIBUTING.md](CONTRIBUTING.md) for branch and pull request rules.\n\nShort version:\n\n- Branch from `main` with a descriptive name (`feature/…`, `fix/…`, `refactor/…`)\n- One change per PR\n- Avoid modifying unrelated files\n- Keep changes small and readable\n\n## Good first issues\n\nCheck the [open issues](https://github.com/TheZupZup/Nova/issues) on GitHub, particularly those labelled `good first issue`. Current tracked issues suitable for new contributors:\n\n- [#23](https://github.com/TheZupZup/Nova/issues/23) Persist the selected model mode across sessions\n- [#49](https://github.com/TheZupZup/Nova/issues/49) Add `max_length` constraints to relevant input fields\n- [#66](https://github.com/TheZupZup/Nova/issues/66) Replace deprecated `datetime.utcnow()` calls\n\n## Roadmap\n\nThe following are areas of active interest, not commitments:\n\n- Hardening the natural language memory pipeline for production use\n- Multi-user support (currently single-user via one set of credentials)\n- Improved model fallback and error reporting in the web UI\n- Docker Compose setup for simpler deployment\n- Broader test coverage\n\n## License\n\n[Mozilla Public License 2.0](LICENSE)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fthezupzup%2Fnova","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fthezupzup%2Fnova","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fthezupzup%2Fnova/lists"}