Configuration

Both the central server and the remote agent take a JSON config file. Examples live in configs/*.json.example; customize and save as configs/server.json or configs/agent.json.

Server (configs/server.json)

Two top-level keys:

{
  "meshlink": { ... },      // optional — enables C2 over Meshtastic
  "captures": [ ... ],      // one entry per capture source
  "web_port": 8080          // optional — starts embedded web UI on this port
}

meshlink

Optional. Opens a Meshtastic serial link, attaches an AgentManager, and routes incoming HELLO/STAT/DET messages from agents into output/agents_YYYYMMDD_HHMMSS.db.

Field	Type	Default	Notes
port	string	—	Serial path, prefer /dev/serial/by-id/... for stability
channel_index	int	0	Meshtastic channel index to transmit on

captures

Each entry is one capture source. The type field selects the backend.

Common fields across all types:

Field	Type	Notes
name	string	Label for dashboard + logs (e.g. "pmr", "uhf", "ble")
type	string	One of: hackrf, rtlsdr, rtlsdr_sweep, ble, wifi, standalone

type: "hackrf"

Wideband IQ capture fed through a channelizer.

{
  "name": "pmr",
  "type": "hackrf",
  "serial": "0000000000000000a06063c8244f6b5f",
  "ppm": 0,
  "center_freq_mhz": 446.1,
  "sample_rate_mhz": 4,
  "lna_gain": 32,
  "vga_gain": 40,
  "transcribe": false,
  "channels": []
}

• serial — HackRF serial (from hackrf_info); matters only when multiple are attached.
• center_freq_mhz, sample_rate_mhz — capture center and bandwidth.
• lna_gain (0–40, step 8) + vga_gain (0–62, step 2) — receive gains.
• transcribe — global default; each voice channel can override.
• channels: [] — if empty, the server auto-discovers FM voice bands that fit within the capture (PMR446, PMR446 Digital, 70cm, etc.). Otherwise, provide explicit entries:

"channels": [
  {
    "name": "keyfob",
    "freq_mhz": 433.92,
    "bandwidth_mhz": 2,
    "parsers": ["keyfob", "tpms"]
  }
]

Channel fields: name, freq_mhz, bandwidth_mhz, parsers (list of parser names), plus optional transcribe, whisper_model, language.

type: "rtlsdr"

Fixed-frequency RTL-SDR capture.

{
  "name": "ais",
  "type": "rtlsdr",
  "device_index": 0,
  "center_freq_mhz": 162.0,
  "sample_rate_mhz": 2.4,
  "parsers": ["ais"]
}

type: "rtlsdr_sweep"

RTL-SDR in sweep mode, covering a band by hopping.

{
  "name": "ism",
  "type": "rtlsdr_sweep",
  "device_index": 0,
  "band_start_mhz": 433,
  "band_end_mhz": 435,
  "parsers": ["keyfob", "tpms"]
}

type: "ble"

{
  "name": "ble",
  "type": "ble",
  "adapter": "hci1",
  "parsers": ["apple_continuity", "remoteid_ble"]
}

Requires sudo (HCI access).

type: "wifi"

{
  "name": "wifi",
  "type": "wifi",
  "interface": "wlan1",
  "band": "all",
  "parsers": ["probe_request", "beacon", "remoteid_wifi", "deauth", "evil_twin"]
}

band: "2.4", "5", or "all". Custom channel list via "channels": [1, 6, 11, 36, 48]. Requires sudo + a monitor-mode capable adapter. deauth flags deauth/disassoc floods (default ≥ 5 frames / 10s per source). evil_twin flags SSID collisions across different BSSID prefixes — both surface in the dashboard's Rogue sub-tab.

type: "standalone"

Wrap an existing sdr.py <subcommand> as a managed subprocess (stdout drained in the background so the pipe buffer doesn't block). Useful for scanners that aren't yet capture/parser-based.

{
  "name": "adsb",
  "type": "standalone",
  "scanner_type": "adsb",
  "device_index": 0,
  "gain": 40,
  "args": ["--transcribe"]
}

Available parser names

• fm_voice (implicit from FM band auto-discovery)
• keyfob, tpms
• gsm, lte
• apple_continuity, remoteid_ble
• probe_request, beacon, remoteid_wifi, deauth, evil_twin
• lora, elrs, meshtastic

Example configs in the repo

• configs/server.json.example — HackRF UHF (keyfob/TPMS), HackRF ISM 868 (LoRa/ELRS), ADS-B, BLE, WiFi
• configs/server_voice.json.example — adds FM voice demod (PMR + 70cm) via channelizer, with transcription
• configs/multi_3band.json.example — three HackRFs covering separate bands
• configs/hackrf_pmr.json.example — single HackRF, PMR-only

Agent (configs/agent.json)

Used by sdr.py agent. Flat JSON:

{
  "agent_id": "N01",
  "meshtastic_port": "/dev/serial/by-id/usb-RAKwireless_WisCore_RAK4631_Board_XXXX-if00",
  "mesh_channel_index": 0,
  "state_dir": "/var/lib/sigint",
  "gps_port": "/dev/serial/by-id/usb-u-blox_AG_-_www.u-blox.com_u-blox_7_-_GPS_GNSS_Receiver-if00"
}

Field	Type	Notes
agent_id	string	Short identifier, ideally N01…N99. Shows up in the Agents tab and as device_id on every detection.
meshtastic_port	string	Serial path to the Meshtastic radio.
mesh_channel_index	int	Must match the server's meshlink.channel_index. Default 0.
state_dir	string	Where state.json, outbox.db, and the scanner's session DB live. /var/lib/sigint for systemd installs.
gps_port	string | null	NMEA GPS serial (u-blox 7/8). Handed to the scanner subprocess as --gps --gps-port <path>.

Environment-variable overrides (if not set in JSON):

• SIGINT_AGENT_ID
• SIGINT_MESHTASTIC_PORT

Example: configs/agent.json.example.

Environment file (.env)

At project root:

OPENAI_API_KEY=sk-...
WHISPER_LANGUAGES=en,es
TAK_HOST=tak.example.com
TAK_PORT=8089

See .env.example for the full list.