1337isnot1337/DOPPLERGEIST

选用指南

AI 根据 README 整理 · 2026/11/3 19:37:05

使用方式

通过命令行运行分析，或启动本地浏览器应用上传录音。

查看 README 依据

Use a new output directory for each experiment. Automatic CI checks installation, source compilation, CLI help and packaged assets without generated data. The [real-world data policy](AGENTS.md#real-world-data-only) supersedes older blanket pytest instructions. For pinned environments and historical clean-install evidence, see [Setup](docs/SETUP.md). Native Windows execution remains unverified. ### Analyze from the command line ```bash .venv/bin/dopplergeist analyze path/to/recording.mp3 --jobs jobs ```

Read `results/my-discovery/summary.json` and the diagnostics under `discovery/`. **These are acoustic event proposals, not coordinates or certified aircraft detections.** See [Audio discovery](docs/AUDIO_DISCOVERY.md) for settings, evidence and failure modes. ### Run the local upload application Start the browser application: ```bash .venv/bin/dopplergeist serve --jobs jobs --online --max-runtime 14400 --port 8765 ```

部署要求

需要 Python 3.11+、FFmpeg 和 Git，支持 Linux 和 macOS，原生 Windows 未验证。

查看 README 依据

Requires **Python 3.11+**, **FFmpeg** (`ffmpeg` and `ffprobe` on `PATH`) and Git. Install FFmpeg through your operating system's package manager, then run from a shell on Linux or macOS: ```bash git clone https://github.com/1337isnot1337/DOPPLERGEIST.git cd DOPPLERGEIST python -m venv .venv .venv/bin/python -m pip install . .venv/bin/dopplergeist --version ```

Use a new output directory for each experiment. Automatic CI checks installation, source compilation, CLI help and packaged assets without generated data. The [real-world data policy](AGENTS.md#real-world-data-only) supersedes older blanket pytest instructions. For pinned environments and historical clean-install evidence, see [Setup](docs/SETUP.md). Native Windows execution remains unverified. ### Analyze from the command line ```bash .venv/bin/dopplergeist analyze path/to/recording.mp3 --jobs jobs ```

已知限制

实验性软件，独立全国精度未建立；需要录音起始时间证据，否则无法进行地理搜索；数据下载可能需数小时和数十GB空间。

查看 README 依据

**Experimental software.** The local upload application and automatic audio-to-decision pipeline are available on `main`, alongside configured research replays. Independent nationwide accuracy on real recordings has not been established. [Current status](#current-status) · [Quickstart](#quickstart) · [How it works](#how-it-works) · [Results](#experimental-results) · [Documentation](#documentation-and-references) ## Current status | Area | Available now | Entry point or remaining work | | --- | --- | --- |

This scans the complete recording and saves an inspectable job. To permit historical search, add `--start-utc` and `--end-utc` for a defensible recording-start window, `--time-source` describing its evidence, and `--online` for public data acquisition. Missing clock evidence returns `needs_input`. Initial acquisition can take hours and tens of gigabytes; [the workflow guide](docs/PRODUCT.md) covers caching, runtime limits and resuming stopped jobs. ### Discover events in a recording

Open `http://127.0.0.1:8765` and upload a recording. Supply a defensible UTC window for the **first audio sample** to permit historical flight search. Without that time evidence, the application returns `needs_input / recording_time_required`. It does not estimate a location. The `--online` option can download aircraft and terrain data; archives and indexes can require tens of gigabytes or more and hours, and there is no automatic disk-space limit. See [Local upload workflow](docs/PRODUCT.md) for request fields, runtime limits and CLI alternatives. ## How it works